2025 BFI Economic Research Briefs by Becker Friedman Institute for Economics

2025

Economic Research Briefs The latest commentary and analysis from BFI’s leading scholars in shorter, digestible forms.

bfi.uchicago.edu

Contents JANUARY A Theory of How Workers Keep Up With Inflation

UChicago Scholar: Erik Hurst

How Much Does the U.S. Fiscal System Redistribute?

UChicago Scholar: Thomas Coleman and David A. Weisbach

Investing in Customer Capital

UChicago Scholar: Amir Sufi

Optimal Income Tax Deductions for Mixed Business and Personal Expenditures

UChicago Scholar: Jacob Goldin

Painful Bargaining: Evidence from Anesthesia Rollups

UChicago Scholars: Paulo Ramos and Thomas Wollmann

The Claiming of Children on U.S. Tax Returns

UChicago Scholar: Jacob Goldin

Toward an Understanding of the Political Economy of Using Field Experiments in Policymaking

UChicago Scholars: Guglielmo Briscese and John List

FEBRUARY Disease, Disparities, and Development: Evidence from Chagas Disease Control in Brazil

UChicago Scholar: Eduardo Montero

Disemployment Effects of Unemployment Insurance: A Meta-Analysis

UChicago Scholar: Peter Ganong

Goals, Expectations, and Performance

UChicago Scholar: Avner Strulov-Shlain

Interest Rate Risk in Banking

UChicago Scholar: Stefan Nagel

Student Loan Forgiveness

UChicago Scholar: Dmitri Koustas

Supply Chain Shocks and Firm Productivity: The Role of Reporting Quality

UChicago Scholars: Philip G. Berger and Rimmy Tomy

Talking about Words Boosts Preschool-Age Children’s Vocabulary: Evidence from a Parent Intervention

UChicago Scholars: Derek Rury, Ariel Kalil, Susan Mayer, and Daniela Bresciani Andaluz

The Anatomy of the Great Terror: A Quantitative Analysis of the 1937-38 Purges in the Red Army UChicago Scholar: Konstantin Sonin

MARCH Boosting Young Children’s Math Skill with Technology in the Home Environment; A Digital Library for Parent-Child Shared Reading Improves Literacy Skills for Young Disadvantaged Children; Priming Parental Identity: Evidence from Experimental Data

Central Bank Communication with the Polarized Public

Drive Down the Cost: Learning by Doing and Government Policies in the Global EV Battery Industry

Effects of Unemployment Insurance for Self-Employed and Marginally-Attached Workers

How Costly Are Business Cycle Volatility and Inflation? A Vox Populi Approach

Income Equality in the Nordic Countries: Myths, Facts, and Lessons

The Price of Faith: Economic Costs and Religious Adaptation in Sub-Saharan Africa

UChicago Scholars: Daniela Bresciani Andaluz, Ariel Kalil, Haoxuan “Noah” Liu, Susan E. Mayer, Rohen Shah, and Derek Rury UChicago Scholar: Michael Weber

UChicago Scholar: Hyuk-soo Kwon

UChicago Scholar: Dmitri Koustas

UChicago Scholar: Michael Weber

UChicago Scholar: Magne Mogstad

UChicago Scholar: Eduardo Montero

APRIL Credit Card Entrepreneurs

Economic Shocks and Healthcare Capital Investments

The Curious Surge of Productivity in U.S. Restaurants

The Social Construction of Race during Reconstruction

Why Has Regional Income Convergence in the U.S. Declined?

UChicago Scholars: Ufuk Akcigit, Raman S. Chhina UChicago Scholar: Maggie Shi

UChicago Scholars: Chad Syverson, Joe Tatarka

UChicago Scholars: Anjali Adukia, Richard Hornbeck, Benjamin Lualdi UChicago Scholar: Peter Ganong

MAY Evaluating Recent Crackdowns on Disability Benefits: Effects on Income and Health Care Use in Australia

Intuit QuickBooks Small Business Index: A New Employment Series for the US, Canada, and the UK

Measuring the Characteristics and Employment Dynamics of US Inventors

Non-User Utility and Market Power: The Case of Smartphones

Separation of Church and State Curricula? Examining Public and Religious Private School Textbooks

The Effect of Medicaid on Crime: Evidence from the Oregon Health Insurance Experiment

Toward an Understanding of Discrimination When Multiple Channels Exist

UChicago Scholars: Manasi Deshpande, Greg Kaplan, Tobias Leigh-Wood

UChicago Scholar: Ufuk Akcigit

UChicago Scholars: Leonardo Bursztyn, Aaron Leonard, Filip Milojević

UChicago Scholars: Anjali Adukia, Emileigh Harrison

UChicago Scholar: Katherine Baicker UChicago Scholar: John List

JUNE Authoritarian Propaganda and Social Networks

Firm Premia and Match Effects in Pay vs. Amenities

Meaning at Work

UChicago Scholar: Konstantin Sonin

UChicago Scholars: Anders Humlum and Evan K. Rose

UChicago Scholars: Virginia Minni and Luigi Zingales

Mechanism Design for Personalized Policy: A Field Experiment Incentivizing Exercise

UChicago Scholar: Rebecca Dizon-Ross

Saved by Medicaid: New Evidence on Health Insurance and Mortality from the Universe of Low-Income Adults

Stablecoin Runs and the Centralization of Arbitrage

UChicago Scholar: Bruce Meyer

UChicago Scholar: Anthony Lee Zhang

The Persistence of Female Political Power in Africa

The Reverse Cargo Cult: Why Authoritarian Governments Lie to Their People

The Role of Risk and Ambiguity Preferences on Early-Childhood Investment: Evidence from Rural India

100

UChicago Scholar: James Robinson

UChicago Scholar: Konstantin Sonin

UChicago Scholars: Michael Cuna, Min Sok Lee, John List

JULY Administrative Fragmentation in Health Care

102

Innovator Networks Within the Firm and the Quality of Innovation

104

Measuring Markets for Network Goods

106

The Benefits of Scholastic Athletics

108

The Local Root of Wage Inequality

110

UChicago Scholar: Maggie Shi

UChicago Scholar: Michael Gibbs

UChicago Scholar: Leonardo Bursztyn UChicago Scholar: James Heckman UChicago Scholar: Hugo Lhuillier

AUGUST Engineering Ukraine’s Wirtschaftswunder

112

Post-Roe Planning: The Effect of Dobbs v. Jackson on Contraceptive and Sterilization Choices

114

Reskilling and Resilience

116

UChicago Scholar: Ufuk Akcigit

UChicago Scholar: Yana Gallen

UChicago Scholar: Anders Humlum

The Personalist Penalty: Varieties of Autocracy and Economic Growth

118

Trust in Banks and Borrower Behavior: Evidence from Supervisory Actions and Local Information Quality

120

Violent Backlash to Political Reform: Evidence from Anti-Jewish Pogroms in the 1905 Russian Revolution

123

What Drives Educational Technology Adoption in Classrooms Serving Young Children? Evidence from Two Experiments

125

UChicago Scholars: Chris Blattman, Scott Gehlbach

UChicago Scholar: Rimmy Tomy

UChicago Scholar: Scott Gehlbach

UChicago Scholar: Ariel Kalil

SEPTEMBER A Tale of Two Transitions: Mobility Dynamics in China and Russia after Central Planning

127

Chat2Learn: A Proof-of-Concept Evaluation of a Technology-Based Tool to Enhance Parent-Child Language Interaction

129

Laboratories of Autocracy: Landscape of Central–Local Dynamics in China’s Policy Universe

131

UChicago Scholar: Steven Durlauf

UChicago Scholar: Ariel Kalil

UChicago Scholar: Shaoda Wang

Navigating the College Affordability Crisis: Insights from College Savings Accounts

134

UChicago Scholar: John List

Partial Language Acquisition: The Impact of Conformity

136

UChicago Scholar: Steven Durlauf

Religion in Emerging and Developing Regions UChicago Scholar: Eduardo Montero

138

OCTOBER Artificial Writing and Automated Detection

141

UChicago Scholar: Alex Imas

Firms Have Partial Knowledge: Evidence from a Reform

143

Human Capital Accumulation Across Space

145

UChicago Scholar: Avner Strulov-Shlain

UChicago Scholar: Esteban Rossi-Hansberg

Jealousy of Trade: Exclusionary Preferences and Economic Nationalism

148

UChicago Scholars: Alex Imas and Heather Sarsons

Superstar Firms Through the Generations

150

The Breakdown of the English Society of Orders: The Role of the Industrial Revolution

153

The Effects of Parental Income and Family Structure on Intergenerational Mobility: A Trajectories-Based Approach

156

The Impact of Language on Decision-Making: Auction Winners are Less Cursed in a Foreign Language

158

Why Is Manufacturing Productivity Growth So Low?

160

UChicago Scholar: Yueran Ma

UChicago Scholar: James Robinson

UChicago Scholar: Steven Durlauf

UChicago Scholars: Ali Hortaçsu and Boaz Keysar

UChicago Scholars: Ali Hortaçsu and Chad Syverson

NOVEMBER Closing Early Math Gaps by Parental Education with Technology at Home

163

UChicago Scholars: Susan Mayer and Ariel Kalil

Debt and Assets

166

The Mortgage Debt Channel of Monetary Policy when Mortgages are Liquid

169

UChicago Scholar: Raghuram Rajan

UChicago Scholar: Greg Kaplan

Who Pays for Tariffs Along the Supply Chain? Evidence from European Wine Tariffs UChicago Scholar: Ali Hortaçsu

171

DECEMBER Comparing the Impacts of Cash vs. SNAP on Consumption of Drugs and Alcohol

174

Dynamic Competition for Sleepy Deposits

176

Social Pressure Drives Parents to Adopt AI that May Harm Students

178

Stop Using Test Scores to Measure Test Results

180

UChicago Scholar: Matthew Notowidigdo UChicago Scholar: Ali Hortaçsu

UChicago Scholars: Leonardo Bursztyn and Alex Imas UChicago Scholar: Jens Ludwig

RESEARCH BRIEF • JANUARY 2025

A Theory of How Workers Keep Up With Inflation Based on BFI Working Paper No. 2024-153, “A Theory of How Workers Keep Up With Inflation,” by Hassan Afrouzi, Columbia University; Andres Blanco, Federal Reserve Bank of Atlanta; Andres Drenik, University of Texas at Austin; Erik Hurst, University of Chicago

The current inflationary period in the United States reduced worker welfare through real wage declines and the costly actions workers took to offset these losses. At the same time, inflation caused a rise in job vacancies—driven by increased job-to-job transitions—creating the appearance of a tight labor market. Between April 2021 and May 2023, the cumulative price level in the United States rose by over 14%. This inflationary period was characterized by low unemployment and historically high job vacancies. Policymakers and economists attributed these trends to a “hot” labor market, where demand for workers outpaced supply. At the same time,

however, real wages fell sharply, challenging this notion. In this paper, the authors offer a new explanation: They argue that inflation, not labor market strength, drove these dynamics, simultaneously increasing vacancies while pushing down real wages.

Real wages: the amount of money a person receives for their work after adjusting for inflation.

Vacancy-to-Unemployment Rate,Rate, Consumer Price Index RealWages Wages Figure 1 · Vacancy-to-Unemployment Consumer Price Index(CPI), (CPI), and and Real A) Vacancy-to-Unemployment Rate (2000-2024) 2

B) CPI and Real Wages (2016-2024) 1.3 CPI

1.5

13%

1.2

1 1.1

-4.4%

0.5

Atlanta Fed Real Wage Index

1 0

Jan. Jan. Jan. Jan. Jan. Jan. Jan. Jan. Jan. Jan. Jan. Jan. 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019 2021 2023

Jan. 2016

Jan. 2017

Jan. 2018

Jan. 2019

Jan. 2020

Jan. 2021

Jan. 2022

Jan. 2023

Jan. 2024

Note: Panel A shows the ratio of vacancy-to-unemployment rate from 2001 through 2024. Panel B shows real wages, measured as the Atlanta Fed Real Wage Index, along with the Consumer Price Index (CPI).

Note: The left graph shows the ratio of vacancy-to-unemployment rate from 2001 through 2024. The right graph shows real wages, measured as the Atlanta Fed Real Wage Index, along with the Consumer Price Index (CPI).

bfi.uchicago.edu

The authors construct a model to examine how inflation affects workers and labor markets. By reducing real wages, inflation prompts workers to renegotiate pay, search for better jobs, or quit, driving job-to-job transitions and increased vacancies while keeping unemployment largely unchanged. Using this framework, the authors explore the consequences of inflation for labor market outcomes and worker welfare, finding the following: •

The temporary inflation during 2021-2023 significantly reduced worker welfare across all income levels, with higher-wage workers experiencing the largest losses. The welfare losses equated to approximately 75% of monthly real income for workers in the bottom decile, 85% for the median, and 110% for those in the top decile.

•

Firms gained from inflation due to increased market power, as reflected in the historically high corporate profit-to-GDP ratios during 2021-2023.

•

While most of the welfare losses stemmed from real wage declines, workers incurred additional costs searching for jobs and renegotiating wages. However, these losses were nearly offset by the gains from reduced layoffs.

Building on these results, the authors validate their model using historical labor market data from 1950-2019. They also examine other high-inflation contexts, and find the following: •

Historical periods of high US inflation (e.g., early 1950s, mid-1970s, and late 1970s) consistently show increases in vacancy rates and vacancy-to-unemployment ratios, even when unemployment remained stable or high.

•

Inflationary periods caused upward shifts in the Curve, meaning there were more job Beveridge Curve openings relative to the number of unemployed workers. This pattern was also observed during the current inflationary episode.

•

International evidence, such as Argentina’s inflation surge in the early 2000s, reveals similar increases in vacancies and labor market tightness during inflation.

The authors conclude by distinguishing between their model’s predictions and alternate “hot labor markets” theories. They show that alternative explanations for high labor market tightness, such as productivity gains, do not align with the observed declines in real wages or the specific patterns of labor market flows that were observed during the recent inflation period. This research challenges traditional interpretations of labor market tightness during inflationary periods, demonstrating that inflation can create the illusion of an overheated labor market while eroding real wages and worker welfare. The findings are consistent with recent supply chain disruptions, energy price increases, and pandemic-related demand pressures that drove up prices without significantly increasing labor demand. Policymakers should exercise caution in interpreting high labor market tightness as a sign of economic strength and consider policies that mitigate the welfare losses caused by inflation, particularly the erosion of real wages and the costs of job transitions.

Beveridge Curve: a graphical representation of the relationship between unemployment and the job vacancy rate, the number of unfilled jobs expressed as a proportion of the labor force. It typically has vacancies on the vertical axis and unemployment on the horizontal and slopes downward, as a higher rate of unemployment normally occurs with a lower rate of vacancies.

READ THE WORKING PAPER NO. 2024-153 · DECEMBER 2024

A Theory of How Workers Keep Up With Inflation bfi.uchicago.edu/working-papers/a-theory-of-how-workerskeep-up-with-inflation

ABOUT OUR SCHOLAR

Erik Hurst

Roman Family Distinguished Service Professor of Economics and John E. Jeuck Faculty Fellow, Chicago Booth; Director, BFI

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JANUARY 2025

How Much Does the U.S. Fiscal System Redistribute? Based on BFI Working Paper 2024-147, “How Much Does the U.S. Fiscal System Redistribute?” by Thomas Coleman and David Weisbach, University of Chicago

The U.S. tax and transfer system has become more redistributive in recent decades, not less; this finding holds across multiple ways of measuring and defining income, households, and transfers. One of the most important—and politically contentious— US policy debates in recent decades involves income inequality, including whether the country’s fiscal (or tax) system has become less progressive and less redistributive over recent decades and, thus, exacerbates inequality. To the point: Many reporters, economists, and policymakers believe that the reduced top income and corporate tax rates initiated in the early 1980s and continued today are key to understanding the subsequent increase in income inequality. For some researchers, these lower tax rates—especially on income for the top 1% and 0.1%, who have experienced income gains—have upended the U.S. tax system and made it less regressive, less progressive or possibly even regressive a serious charge against a tax system founded on the principle that the more you earn, the more you pay. This paper challenges that emerging consensus by reassessing what and how data are measured to find that the US tax and transfer system redistributes more now than it did in the last several decades. The authors come to this conclusion by reviewing the methodologies of recent research, and by focusing on all income levels, not just the top 1%. The authors also describe the many ways that differing measurements and definitions can result in dissimilar conclusions. Their findings are best described via the accompanying figures,

Figure 1 · Tax and Transfer Rates, AS (1966-2019) vs. PSZ (1962-2019) A) Auten and Splinter (AS) 60% Top 1%

Top 10% 20

Middle 40%

-20

-40

-60 1960

Bottom 50%

1970

1980

1990

2000

2010

2020

2010

2020

B) Piketty, Saez and Zucman (PSZ) 60%

Progressive (tax system): A system wherein the average tax rate increases as income increases.

-20

Regressive (tax system): A system wherein the average tax rate goes down as income increases.

-40

-60 1960

1970

1980

1990

2000

Note: Details for data sources are in BFI Working Paper 2024-147.

bfi.uchicago.edu

which incorporate data from three recent research efforts that estimate changes to progressivity and redistribution and that make their data publicly available—Congressional Budget Office (CBO) 2022; Auten and Splinter (AS) 2024; and Piketty, Saez, and Zucman (PSZ) 2018). The first figure compares the tax and transfer rates produced by AS and PSZ. PSZ only provide consistent rankings for the bottom 50 percent, the 50-90 percent (which they call the middle 40 percent), the top 10 percent, the top 1 percent, and smaller groupings at the top. To compare the results to PSZ, the authors use the AS data for the same groups. Data from CBO (2022) yield the same or even stronger results when the population is divided into quintiles. The figure shows the tax and transfer rates for each of these groups from 1966 (the first year of consistent data in AS) until 2019. •

The key finding in the first figure is that both datasets show that the crucial change over the last 60 years has been the dramatic increase in transfers to the bottom half of the population. This increase swamps the changes in tax rates for the top half of the population.

The second figure helps us understand how the fiscal system has increased transfers to bottom income groups. The Panel A shows transfers as a share of national income. Panel B drills further down, showing just the transfers to the bottom quintile along with its income share. •

The second figure illustrates that the downward distribution of income is somewhat nuanced. Although total transfers have gone up over time (from 5.2% of national income in 1966 to 15.3% of national income in 2019), transfers to the bottom quintile peaked in 1975 at 5.7% of national income and declined since then to 4.7% of national income. The increase in transfers as a share of national income has instead largely accrued to the middle quintiles.

READ THE WORKING PAPER NO. 2024-147 · NOVEMBER 2024

How Much Does the U.S. Fiscal System Redistribute? bfi.uchicago.edu/working-papers/how-much-does-u-sfiscal-system-redistribute

Figure 2 · Share of Transfers for Each Quintile and for Bottom Quintile, 1960-2019 A) Transfers as Share of National Income 6% 5

Bottom Quintile

Quintile 2 Quintile 3

Quintile 4

Top Quintile

1 0

1960

1970

1980

1990

2000

2010

2020

B) Bottom Quintile Transfers vs. Pre-Tax Income Share 6% 5

Transfers

4 Income Share

3 2 1 0

1960

1970

1980

1990

2000

2010

2020

Note: Details for data sources are in BFI Working Paper 2024-147.

Bottom line: Policymakers take note—there is broad agreement among researchers that net transfer rates to bottom incomes have increased, and the size of those increases swamp any changes at the top. Relatedly, the tax and transfer system has become more redistributive over the last half century, with much of that increase occurring in the last several decades.

ABOUT OUR SCHOLARS

Thomas Coleman

Senior Lecturer, Harris School of Public Policy

David A. Weisbach

Walter J. Blum Professor of Law, University of Chicago Law School

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • JANUARY 2025

Investing in Customer Capital Based on BFI Working Paper No. 2024-144, “Investing in Customer Capital,” by Bianca He, Lauren I. Mostrom, and Amir Sufi, University of Chicago

Firms’ investments in customer capital are a key driver of their intangible capital value and are strongly linked to their overall value as well as industrywide growth. Previous research shows that corporations allocate significant resources to intangible capital capital—assets such as brand reputation, intellectual property, and customer loyalty. Intangible capital shapes key economic outcomes such as investment, profits, employee compensation, and productivity,

making it critical to accurately measure and understand. In this paper, the authors focus on customer capital, a major component of intangible capital, and provide comprehensive measures of investment in customer capital as well as explore its determinants and effects.

intangible capital: non-physical assets that contribute to a firm’s value, such as brand reputation, intellectual property, and customer relationships customer capital: a component of intangible capital representing investments in building and maintaining customer relationships, including sales, marketing, and customer data

Investing Customers Grow Faster Figure 1 · Industries Industries Investing in in Customers Grow Faster A) Share of Revenue

B) Share of Enterprise Value

40% Industry Group’s Share of Revenue

Top Tercile

50% Industry Group’s Share of Enterprise Value

45 35 Middle Tercile

35 30 30 Bottom Tercile 25 2008

25 2012

2016

2020

2008

2012

2016

2020

Note: For both figures, are sorted three groups based on the industry’s investment in sales and marketing Note: For both figures, industries are sortedindustries into three groups basedinto on the industry’s median investment in sales andmedian marketing (divided by revenue). The evolution of the shares for each group (divided by revenue). The value evolution for each group over is shown for with revenue and enterprise in the left over time is shown for revenue and enterprise in the of leftthe andshares right panels, respectively. As youtime can see, industries the largest ratio of salesvalue and marketing expense to revenue experience and right panels, respectively. As you can see, industries with the largest ratio of sales and marketing expense to revenue the largest increase in the share of both revenues and enterprise value over time. experience the largest increase in the share of both revenues and enterprise value over time.

bfi.uchicago.edu

The authors measure investments in customer capital using three data sources. First, they analyze spending on marketing and sales as reported in corporate income statements. Second, they estimate salaries paid to sales and marketing employees based on data from job postings. Third, they examine firms’ SEC filings, which often detail investments in customer capital. Together, these data sources enable the authors to construct a comprehensive dataset on customer capital investments by publicly traded firms in the United States from 2007 to 2022. The data reveal the following concerning consumer capital: •

•

On average, publicly traded firms spend 4.1% of their revenue on sales and marketing. This amount is higher than expenditures on research and development (R&D) and about two-thirds of capital expenditures. While previous research has mainly focused on advertising to measure investments in customer capital, the authors find that advertising often accounts for only a small part of this spending. Notably, 16% of annual firm records highlight advertising in their business descriptions, and it is more common for firms to focus on efforts like maintaining strong customer service (51%), building a sales force (49%), enhancing brand value (48%), and leveraging customer data (28%). There is significant variation across industries in investment in customer capital. Firms in agriculture, mining, and petroleum or coal product manufacturing spend little to nothing on sales and marketing. In contrast, the median firm in the information industry—such as software companies, digital platforms, and web search portals—spends over 20% of its revenue. Professional service firms and hightech manufacturers, including those making medical devices and computer equipment, also invest heavily in customer capital.

The authors confirm that the observed variation in industry-level investment in customer capital remains consistent over time and across measurement methods. This consistency suggests that the variation arises from fundamental differences in how industries generate revenue and profit. Motivated

by this, the authors analyze the determinants and implications of this industry-level variation. •

Three factors explain 70% of the variation in investment in customer capital across industries. First, platform business models, where firms act as intermediaries connecting buyers and sellers (e.g., eBay, Uber, Zillow), are the most significant driver, with such firms typically spending over 20% of their revenue on sales and marketing. Second, industries where firms sell products online invest heavily in customer capital, focusing on acquiring and leveraging customer data alongside brand building. Third, industries producing technical products—indicated by higher salaries for engineers—allocate substantial resources to customer capital, reflecting the need to support complex product sales with skilled teams and specialized marketing.

•

The channels through which industries invest in customer capital differ based on their business models and target markets. Industries that primarily target households focus heavily on advertising and building brand value. Industries that sell technical products prioritize developing and maintaining a strong sales force to guide customers through complex purchase decisions. Online-focused industries emphasize customer data acquisition and management, often paired with efforts to strengthen brand value. Platform-based industries, such as those facilitating digital marketplaces, invest across all channels—advertising, brand value, customer service, sales force, and customer data.

•

Turning to the implications of investing in customer capital, the authors find that industries with the largest investments in customer capital experience the greatest increases in the share of revenue and enterprise value value over the authors’ sample period.

•

Industry-level differences in customer capital investment explain much of the variation in intangible capital value across industries. Industries that invest more in customer capital relative to their revenue have higher enterprise value compared to their physical assets (QPH). QPH

enterprise value: a measure of a company’s total value, calculated as the market value of equity plus debt, minus cash, representing its overall worth to investors QPH (enterprise value to physical capital ratio): a metric assessing how much value a firm generates relative to its tangible (physical) assets

Similarly, R&D spending is strongly linked to higher intangible capital value. In contrast, residual sales, general, and administrative expenses, excluding customer capital, show no correlation with intangible capital value, highlighting the unique role of customer capital and R&D in driving value. •

Customer capital is also crucial in firm acquisitions, with higher investments linked to greater valuations of intangible assets such as brands, customer lists, and relationships. While R&D expenses primarily drive the value of other intangible assets like research and technology, residual sales, general, and administrative expenses expenses (excluding customer capital-related components) show no correlation with valuations in acquisitions.

This study highlights that spending on sales and marketing should be recognized as investment in customer capital, a critical driver of intangible asset value and firm growth. Industries that invest heavily in customer capital, such as those focused on platform business models, online sales, and hightech manufacturing, are among the fastest-growing in revenue and enterprise value. Policymakers and researchers should prioritize better measurement and understanding of customer capital investment, as it plays a pivotal role in shaping economic outcomes across industries.

READ THE WORKING PAPER NO. 2024-144 · NOVEMBER 2024

Investing in Customer Capital bfi.uchicago.edu/working-papers/investing-incustomer-capital

ABOUT OUR SCHOLARS

Bianca He

PhD Student in Finance, Chicago Booth

Lauren I. Mostrom

PhD Student in Finance, Chicago Booth

Amir Sufi

Bruce Lindsay Distinguished Service Professor of Economics and Public Policy, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JANUARY 2025

Optimal Income Tax Deductions for Mixed Business and Personal Expenditures Based on BFI Working Paper No. 2024-156, “Optimal Income Tax Deductions for Mixed Business and Personal Expenditures,” by Jacob Goldin, University of Chicago; Sebastian Koehne, Kiel University; and Nicholas Lawson, Dalhousie University

The optimal tax deduction for mixed-purpose expenditures should be proportional to the extent that the expense serves an income-generating function versus a consumption function. Business expenses are typically tax deductible, while personal expenses are not. But what happens when an expense serves both purposes? For instance, a small business owner might purchase a car used for both professional and personal activities or book an airplane ticket that combines a customer meeting with a family visit. This gray area has long challenged tax policy. As UChicago economist Henry Simons noted in 1938, “There is here an essential and insuperable difficulty, even in principle.” In practice, tax rules governing mixed-purpose expenditures are complex and inconsistent. In the United States, some such expenses are fully deductible (e.g., flying business class to meet a client), others are entirely non-deductible (e.g., buying a suit to wear at work), and some are only partially deductible (e.g., business meals, capped at 50%). In certain cases, deductibility depends on subjective factors, such as the taxpayer’s primary

intent for the expense (e.g., travel costs) or how the expense is used in practice (e.g., home offices). This paper examines the optimal taxation of mixed-purpose expenditures. Building on theoretical tax models, the authors explore how much of these expenses should be deductible based on their dual roles in generating income and providing personal benefits. They argue that mixed-purpose goods justify the use of different tax rates for different types of expenses, and outline how such distinctions should be implemented in tax policy. Key findings include: •

Under an optimal tax system, expenses solely for generating income should be fully deductible, while purely personal expenses should not be deductible. Mixed-purpose expenses should be partially deductible based on the proportion related to income generation. As expenses resemble business inputs more

bfi.uchicago.edu

closely, the deductible share should increase; the more they resemble personal consumption, the smaller the deduction. •

The optimal deduction rate for mixedpurpose goods should account for differences in income levels, as the ratio of incomegenerating to consumption benefits may vary across taxpayers. This can be implemented through income-sensitive phase-ins and phase-outs of deductions, a common feature of existing tax systems.

•

Deduction rates should be determined by marginal benefits—the additional benefits derived from extra spending—rather than average benefits. For example, while a basic phone plan may largely support work, upgrading for premium features like a highquality camera is primarily consumption-driven and should have a low or zero deduction rate.

•

For goods that provide personal utility but reduce income, the optimal policy involves a “negative deduction rate,” effectively adding a portion of the expense to taxable income to reflect its impact on earning capacity.

This research provides a framework for designing income tax deductions for mixed-purpose expenditures. The findings have significant implications for tax design: they justify partial deductions for hybrid goods and propose adjustments based on taxpayer income, marginal benefits, and administrative constraints. Furthermore, the study opens new avenues for considering “negative deductions” for goods that diminish productivity, offering a novel perspective on addressing behaviors that harm economic output.

READ THE WORKING PAPER NO. 2024-156 · DECEMBER 2024

Optimal Income Tax Deductions for Mixed Business and Personal Expenditures bfi.uchicago.edu/working-papers/optimal-income-taxdeductions-for-mixed-business-and-personal-expenditures

ABOUT OUR SCHOLAR

Jacob Goldin

Richard M. Lipton Professor of Tax Law, The Law School

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JANUARY 2025

Painful Bargaining: Evidence from Anesthesia Rollups Based on BFI Working Paper No. 2024-149, “Painful Bargaining: Evidence from Anesthesia Rollups,” by Aslihan Asil, Yale University; Paulo Ramos, University of Chicago; Amanda Starc, Northwestern University; and Thomas G. Wollmann, University of Chicago

Rollups in the anesthesia industry drive consolidation and significant price increases, particularly in already concentrated markets. Court-ordered divestitures and other remedies show potential for reducing these anti-competitive effects. Roll Up Houston In 2012, a former healthcare executive approached the private equity firm Welch, Carson, Anderson & Stowe (WCAS) with a plan to “aggressively” consolidate anesthesia markets. In each, the firm would employ a “rollup” rollup strategy: acquire an initial practice (the “platform”) and then expand by acquiring competitors (the “add-ons”). The strategy was commonly used by private equity firms at the time and has grown even more popular since then. Today, it accounts for over $1 trillion in annual deal volume. WCAS soon formed US Anesthesia Partners (USAP) and, in December 2012, acquired Greater Houston Anesthesiology, the largest practice in the region with 220 anesthesiologists. Over the next seven years, USAP acquired three additional

Figure 1 · Prices Rise Following Add-On but not Platform Acquisitions Prices Rise Following Add-On but not Platform Acquisitions 0.3

0.25 Add-on Acquisitions 0.2

0.15 Time of Event 0.1

0.05

Platform Acquisitions

-0.05

-8

-4 4 0 Time Before and After Event (Calendar Quarters)

Note: This plots the average anesthesia price on the vertical timeaxis on the horizontal axis. Note: Thisfigure figure plots the average anesthesia price onaxis theagainst vertical against time on the The vertical gray line corresponds to the quarter of the acquisition. Prices are normalized to zero in that horizontal The vertical gray line acquisitions; corresponds to theline quarter of the Prices are quarter. The axis. blue line corresponds to add-on the green corresponds to acquisition. platform acquisitions (i.e., the first acquisition by a sponsor in a market). normalized to zero in that quarter. The blue line corresponds to add-on acquisitions; the green line corresponds to platform acquisitions (i.e., the first acquisition by a sponsor in a market).

rollup: a business strategy where a firm consolidates smaller companies in the same industry to create efficiencies or gain market power, often leading to increased market concentration and higher prices

bfi.uchicago.edu

practices in Houston, as well as replicated their strategy in Dallas and Austin. Each acquisition was followed by substantial price increases. In 2023, the Federal Trade Commission (FTC) filed charges against USAP for antitrust violations, claiming that their transactions substantially reduced competition and that their conduct monopolized markets. The lawsuit, which is still ongoing, marks a significant milestone as it is the first time federal authorities have challenged a series of completed transactions and included charges against the acquirer’s financial sponsor. In its press release, the agency borrowed a term coined by one of this study’s authors, Wollmann, calling the transactions “stealth consolidation”—a reference to the fact that the acquisitions initially escaped the FTC’s notice because they were exempt from reporting requirements. In this paper, the authors study rollups in the anesthesia industry. They use data from medical claims to examine both the litigated transactions in Texas, as well as 18 similar rollups that occurred between 2012 and 2021. The authors measure changes in market structure, price, and other outcomes following the acquisitions and predict the impact of remedies and policies that combat anticompetitive rollups. They begin by documenting the following concerning rollups in the anesthesia industry: •

Rollups are a key driver of consolidation in anesthesia markets, significantly shaping market structures. The authors calculate the Herfindahl–Hirschman Index (HHI), a standard measure of market concentration, and compare actual changes in HHI to those predicted solely from add-on acquisitions. They show that rollups, rather than factors like entry, exit, or firm growth, are the primary cause of increased concentration. Large

metropolitan markets such as Phoenix, Las Vegas, Louisville, and Denver, as well as smaller ones like Trenton-Ewing and Syracuse, experienced dramatic consolidation during the authors’ sample window, with HHI increases often exceeding 1,000 points and, in some cases, surpassing 2,500 points. •

These market changes drive price increases. Within six months of an add-on acquisition, prices rise by 18 percent, and within two years, they increase by 25-30 percent. Interestingly, prices do not increase following platform acquisitions. This suggests that the price hikes are driven by reduced competition resulting from add-on acquisitions, rather than the direct involvement of financial sponsors.

•

Neither platform purchases nor add-on acquisitions improve anesthesia quality or increase the non-anesthesia prices of anesthetized procedures, suggesting that these acquisitions primarily aim to consolidate market power and increase anesthesia prices rather than enhance service quality or overall procedural costs.

•

Larger expected increases in concentration were associated with larger post-merger price changes, revealing that preacquisition market structure plays a crucial role in shaping the magnitude of price effects.

The authors use these results to construct a structural model of payor-provider bargaining. They simulate counterfactual outcomes under potential remedies and policies considered by judges, agency officials, and legislators aimed at mitigating the adverse effects of consolidation, and find the following: •

The authors first analyze the impact of court-ordered divestitures of add-on

Herfindahl–Hirschman Index: a widely used measure of market concentration in economics and antitrust analysis that is calculated by summing the squares of the market shares of all firms in a market, with the resulting index ranging from 0 (perfect competition) to 10,000 (monopoly) <1,500: low concentration 1,500–2,500: moderate concentration >2,500: high concentration change >200 points: indicates significant consolidation structural model: an economic model that estimates how agents (e.g., payors and providers) interact, used to simulate policy outcomes and predict the effects of market changes

acquisition targets by simulating the effects of “unwinding” these transactions. They find that such remedies would reduce anesthesia expenditures by $120 million annually, or 3.25%. Additionally, court-ordered divestitures—or even court-awarded damages to private plaintiffs—could deter future rollup acquisitions. The authors simulate this deterrence effect and project further annual savings of $126 million, underscoring the potential long-term benefits of these interventions. •

Next, the authors examine the impact of market entry, inspired by an alleged agreement where one firm paid another $9 million to avoid entering its market. Through their simulations, they find that while market entry does lead to reduced expenditures, its impact is relatively modest.

•

Finally, the authors assess the potential effects of recent legislation addressing ‘balance billing’ on anesthesia markets. Balance billing occurs when healthcare providers charge patients the difference between their billed amount and the amount covered by insurance, often leading to high out-of-pocket costs for patients. Consolidation in anesthesia markets exacerbates this issue, as larger, more concentrated provider groups gain increased bargaining power, potentially enabling them to charge higher out-of-network rates or demand greater reimbursement from insurers. The authors simulate the impact of limiting balance billing and find that while such

READ THE WORKING PAPER NO. 2024-149 · DECEMBER 2024

Painful Bargaining: Evidence from Anesthesia Rollups bfi.uchicago.edu/working-papers/painful-bargainingevidence-from-anesthesia-rollups

policies reduce expenditures in anesthesia markets, the effects are smaller compared to remedies like court-ordered divestitures or deterrence of future rollups. Serial acquisitions of clinician practices, often backed by investment funds, have consolidated geographically dispersed markets, raising significant antitrust concerns. While much of health policy research focuses on differences between provider types (e.g., for-profit vs. nonprofit, physician-owned vs. corporate-owned), this research underscores that competitive dynamics are the primary driver of price increases, with acquisitions frequently resulting in exceptionally large price hikes. The authors demonstrate that antitrust remedies, such as divestitures and stricter scrutiny of acquisitions, show promise in mitigating these harmful effects and preserving competition in the healthcare sector.

ABOUT OUR SCHOLARS

Paulo Ramos

PhD Student, Chicago Booth

Thomas Wollmann

Associate Professor of Economics and William Ladany Faculty Scholar, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JANUARY 2025

The Claiming of Children on U.S. Tax Returns Based on BFI Working Paper No. 2025-06, “The Claiming of Children on U.S. Tax Returns,” by Geoffrey Gee, U.S. Treasury Department; Jacob Goldin, University of Chicago; Joseph Gray-Hancuch, U.S. Treasury Department; Ithai Z. Lurie, U.S. Treasury Department; and Vedant Vohra, University of California San Diego

Approximately 95% of children in the United States are claimed on their parents’ tax returns, with lower rates among children in Black and Hispanic neighborhoods, and children from low-income families. Providing financial assistance to low-income families with children is a central policy goal. In the United States, much of this support is delivered through the tax system, requiring families to file tax returns to access child-related programs. Filing requirements often pose barriers, however, with approximately 20% of eligible households failing to claim the Earned Income Tax Credit (EITC), primarily due to non-filing.

This study examines how children are claimed on federal tax returns, using health insurance data to identify children and measure claim rates. The sample includes children reported as enrolled in health insurance for at least one month annually, representing 92-94% of U.S. children. The analysis spans tax years 2017-2021, capturing tax policy changes during that period. Key findings include:

Figure 1 · Child Claim Rate by Neighborhood Racial Composition Child Claim Rate by Neighborhood Racial Composition B) Share Hispanic

95.5 %

96 %

Claim Rate

A) Share Black

94.5

40 % Share Black

40 % Share Hispanic

Note:These These plots show binned scatterplots of the fraction of U.S. children claimed on a tax 2018 tax return, theorracial ormakeup ethnic makeup of the child’s neighborhood. Note: plots show binned scatterplots of the fraction of U.S. children claimed on a 2018 return, by the by racial ethnic of the child’s neighborhood. The solid line represents the best linear fit. The solid line represents the best linear fit.

Earned Income Tax Credit: a refundable tax credit aimed at supporting low- to moderate-income working individuals and families by reducing their tax burden and providing additional income binned scatterplot: a graphical representation that groups data points into bins based on a variable and plots the average value of another variable within each bin to identify trends or relationships

bfi.uchicago.edu

•

In recent years, most children in the U.S. have been claimed on tax returns, with 95-96% of children with health insurance reported on their parents’ returns. Using data on the fraction of the population without insurance, the authors estimate a lower bound of at least 88% for the child claim rate.

•

Parents with very low wages, as well as those whose children receive health insurance through Medicaid or the Children’s Health Insurance Program (CHIP), are less likely to claim their children on tax returns. This is likely due to reduced incentives to file, as non-earners generally did not qualify for taxadministered benefits during most years in the study, and very low-income taxpayers were eligible for only modest benefits.

•

Child claim rates are lower for children living in Black and Hispanic neighborhoods, potentially due to the smaller tax benefits these children are typically eligible to receive.

This research has important implications for designing safety-net programs. It indicates that tax-administered benefits are likely to reach most children without substantial additional cost. However, notable gaps exist for the lowestincome children and those in Black and Hispanic neighborhoods. Expanding eligibility to include traditionally excluded children could increase claim rates by providing stronger incentives for taxpayers to file returns. Without such behavioral changes, tax-administered safety net programs risk deepening existing socioeconomic inequalities.

READ THE WORKING PAPER NO. 2025-06 · DECEMBER 2024

The Claiming of Children on U.S. Tax Returns bfi.uchicago.edu/working-papers/the-claiming-of-childrenon-u-s-tax-returns

ABOUT OUR SCHOLAR

Jacob Goldin

Richard M. Lipton Professor of Tax Law, The Law School

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JANUARY 2025

Toward an Understanding of the Political Economy of Using Field Experiments in Policymaking Based on BFI Working Paper No. 2024-157, “Toward an Understanding of the Political Economy of Using Field Experiments in Policymaking,” by Guglielmo Briscese and John A. List, University of Chicago

When a field experiment shows disappointing results, policymakers update their views but show reduced demand for experiments. The public remains supportive of experiments but their trust in the implementing institutions decreases. Field experiments are widely regarded as the gold standard for uncovering the true impact of many policies. Recent advancements in experimental methods have enabled an unprecedented ability to understand what works, what doesn’t, and why. Despite their promise, however, field experiments have seen limited adoption among policymakers. In this paper, the researchers tackle this “uptake problem,” which they propose stems from two key challenges: First, concerns about scalability can erode confidence in scientific findings, as many programs that succeed in experimental settings struggle to deliver comparable results at scale. Second, the inherent uncertainty of experimentation can impose costs on policymakers, particularly when results diverge from expectations. To test these concerns, the authors design an experiment in which they study how policymakers respond to unexpected or counterintuitive findings. Their approach is twofold. First, they implement

Figure 1 · Belief Updating Among Policymakers Belief Updating Among Policymakers 1 Expectations for Trial

0.8 Treatment Control 0.6

0.4

0.2

0 0

0.2

0.4 Expectations for Pilot

0.6

0.8

Note: This figure illustrates policymakers’ predictions for the outcomes of both a pilot intervention and a full-scale

intervention. points are color-codedpolicymakers’ to distinguish whether policymakers learned about the pilot results beforea making Note: ThisThe figure illustrates predictions for the outcomes of both pilot their predictions for the full trial. As you can see, policymakers who were not exposed to the pilot results tend to maintain intervention and between a full-scale The are who color-coded toofdistinguish whether consistent expectations the pilotintervention. and trial outcomes, whilepoints policymakers were informed results overwhelmingly anticipate null results for the full-scale intervention. policymakers learned about the pilot results before making their predictions for the full trial. As you can see, policymakers who were not exposed to the pilot results tend to maintain consistent expectations between the pilot and trial outcomes, while policymakers who were informed of results overwhelmingly anticipate null results for the full-scale intervention.

a field experiment to assess the efficacy of a commonly used intervention in higher education. Second, they investigate how policymakers—and the broader public—respond to the results.

Field experiment: A study conducted in a real-world setting where researchers manipulate one or more independent variables to observe the effect on an outcome of interest. Null result: When a study or experiment does not find a statistically significant effect or relationship between the variables being tested. This means the data does not provide evidence to reject the null hypothesis (the assumption that there is no effect or relationship).

bfi.uchicago.edu

The authors’ field experiment evaluates the impact of small financial incentives designed to increase participation in college savings accounts, a policy intervention implemented in most states but not previously tested experimentally. The incentives are premised on the idea that small financial rewards can increase program attractiveness and help citizens outweigh the burden of signup costs. The results are unexpected, however: In their trial, financial incentives did not significantly improve participation in college savings accounts. The authors then conduct a survey experiment with a sample of policymakers responsible for administering college savings accounts and similar state-run savings programs, as well as with a representative sample of the US public. The survey elicits respondents’ predictions for the experimental outcomes and gathers their opinions on policy experiments. During the survey, a randomly assigned treatment group is informed of the actual experimental results. The authors compare the responses between the treatment and control groups and find the following: •

Policymakers who are randomly exposed to the trial results report a reduced focus on scaling the trial and an increased interest in funding new evaluations.

•

Learning the trial results also causes policymakers to question both the efficacy of similar interventions in other contexts, and even the scientific approach in general.

•

The authors identify similarities between their sample of US citizens and policymakers. For example, like policymakers, the public is (even

READ THE WORKING PAPER NO. 2024-157 · DECEMBER 2024

Toward an Understanding of the Political Economy of Using Field Experiments in Policymaking bfi.uchicago.edu/working-papers/toward-anunderstanding-of-the-political-economy-of-using-fieldexperiments-in-policymaking

more) optimistic about the potential for financial incentives to improve program participation rates. Additionally, when the public learns the results, they state remarkably similar preferences for how resources should be re-allocated between scaling and further evaluations. •

The authors also find dissimilarities between citizens and policymakers. For example, unlike policymakers, the public maintains their high support for policy experiments even after learning of the disappointing trial results. Their trust in the government institutions responsible for implementing the trial declines, however, particularly with regards to perceptions of competence and integrity.

•

Education can partly restore this loss of trust. Members of the public who receive information explaining the value of policy experiments and the importance of learning from unexpected results partially report smaller reductions in trust.

This research highlights a significant challenge to advancing evidence-based policymaking at scale: the need to manage both policymakers’ and citizens’ expectations of policy experimentation. The authors propose communication strategies that prepare policymakers for all potential outcomes and emphasize the valuable lessons trials can provide, regardless of their results. Finally, the authors recommend conducting future evaluations to assess the effectiveness of educational interventions that leverage policy experiments to enhance government accountability and foster trust among citizens.

ABOUT OUR SCHOLARS

Guglielmo Briscese

Postdoctoral Scholar, Harris School of Public Policy

John A. List

Kenneth C. Griffin Distinguished Service Professor of Economics, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Disease, Disparities, and Development: Evidence from Chagas Disease Control in Brazil Based on BFI Working Paper No. 2025-13, “Disease, Disparities, and Development: Evidence from Chagas Disease Control in Brazil,” by Jon Denton-Schneider, Clark University; and Eduardo Montero, University of Chicago

Brazil’s post-1983 campaign to eliminate Chagas disease increased municipalities’ GDP per capita by 11.1% and reduced their Gini coefficients by 1.1% in the long run, while reducing racial earnings gaps. Spending on circulatory disease hospital care declined by 16%, contributing to an internal rate of return of 24% and an infinite marginal value of public funds. Latin America is one of the most unequal regions in the world, with the richest 10% capturing 54% of national incomes. Racial disparities further compound this inequality, as white individuals earn at least twice as much as those with darker skin tones. One underappreciated contributor to these disparities is Chagas disease, a neglected tropical disease that afflicts 8 million people in Latin America, with another 75 million at risk of exposure. Chagas is spread primarily by triatomine bugs, which thrive in substandard housing conditions. The disease

can cause chronic heart problems, leading to longterm health deterioration that reduces labor force participation and reinforces cycles of poverty. In this paper, the authors examine the impacts of combatting Chagas on economic inequality, the intergenerational transmission of poverty, and burdens on healthcare systems. They do so in the context of Brazil’s post-1983 campaign to eliminate Chagas disease transmission through vector control. The authors exploit geographic variation in exposure to

Figure 1 ·Chagas Effect Disease of Chagas Disease Vector Control on Municipalities Effect of Vector Control on Municipalities A) GDP per Capita

B) Inequality (Gini Coefficient)

0.3 GDP per Capita (2019 R$)

0.01 Gini Coefficient

0.2 Treatment

0.01

Post-treatment average 0.1

0.02 0

-0.1

0.03

1970

1980

1990

-0.04

2000

2010

Estimate

90% Confidence Interval

1970

1980

1990

2000

2010

95% Confidence Interval

Note: These theimpact impactofofChagas Chagas disease control on GDP per capita and income inequality B). The vertical linesand show 90% and 95% confidence intervals. Note: Thesefigures figures show show the disease control on GDP per capita (Panel(Panel A) andA) income inequality (Panel B).(Panel The vertical lines show 90% 95% confidence intervals.

bfi.uchicago.edu

Chagas disease prior to the intervention, comparing municipalities that were previously infested with those that were never exposed. They find the following: •

•

Municipalities that received vector control interventions saw an 11.1% increase in GDP per capita relative to those that were never exposed to Chagas disease. Income inequality, as measured by the Gini coefficient coefficient, declined by 1.1% in treated municipalities, indicating that the intervention contributed to greater economic equity. Exposure to Chagas disease control in childhood led to higher adult incomes for nonwhite Brazilians, accelerating racial income convergence. The share of non-white adults earning above the national median increased by 1.4 percentage points (2.8%), while there was no significant change among white adults. Labor force participation among non-white adults from treatment municipalities increased by 0.9 percentage points (1.3%), suggesting that improved long-term health played a key role in these economic gains.

•

The children of non-white men from treated municipalities had a 0.44 percentage point (0.46%) increase in literacy rates, signaling that the intervention helped break cycles of intergenerational poverty.

•

Hospitalizations due to circulatory system diseases (a common long-term consequence of Chagas) decreased by 19% in treated areas compared to other causes. Public spending on hospital care for circulatory diseases declined by 16%, relieving strain on Brazil’s universal healthcare system.

•

The benefits of Chagas disease control are high compared to their costs. The internal internal rate rate of of return (IRR) for the intervention, even excluding unmeasured health benefits, was 23.9%, demonstrating strong economic justification for disease control. The Marginal marginal Value value of of public Public funds Funds (MVPF) (MVPF) was infinite, meaning the government recovered its costs entirely through hospital savings alone.

•

Eliminating Chagas disease transmission across Latin America could increase regional GDP per capita by 1.5% and reduce Gini coefficients by 0.15%. Notably, greater benefits would accrue to countries in the region with higher shares exposed to Chagas disease, which are precisely the countries that are more underdeveloped and unequal today.

These findings provide a more comprehensive understanding of the economic benefits of controlling neglected tropical diseases in developing countries, most of which can be effectively managed through environmental interventions like the campaign examined in this study. While the results presented here specifically pertain to Chagas disease, which is unique to the Americas, the broader implications for other diseases remain an open question for future research. Nevertheless, this study highlights new pathways through which health improvements can drive inclusive economic growth, reinforcing the case for investing in disease control—not only for Chagas, which affects many millions of people—but also for other diseases that contribute to chronic health burdens among the world’s poorest populations.

Gini coefficient: a measure of income inequality, where 0 represents perfect equality and 1 (or 100%) represents maximum inequality Internal rate of return (IRR): the discount rate at which the net present value of an investment is zero, indicating its profitability Marginal value of public funds (MVPF): a term coined by Nathaniel Hendren and Ben Sprung-Keyser in their 2020 paper “A Unified Welfare Analysis of Government Policies,” the benefit that a policy provides its recipients divided by its net cost

READ THE WORKING PAPER NO. 2025-13 · JANUARY 2025

Disease, Disparities, and Development: Evidence from Chagas Disease Control in Brazil bfi.uchicago.edu/working-papers/disease-disparities-anddevelopment-evidence-from-chagas-disease-control-in-brazil

ABOUT OUR SCHOLAR

Eduardo Montero

Assistant Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Disemployment Effects of Unemployment Insurance: A Meta-Analysis Based on BFI Working Paper No. 2024-96, “Disemployment Effects of Unemployment Insurance: A Meta-Analysis,” by Jonathan Cohen, Amazon; and Peter Ganong, University of Chicago

Previous studies have overestimated the impact of unemployment benefits on unemployment duration by a third due to pervasive publication bias. Using these previous studies, some economic analyses concluded that the optimal policy is to have no unemployment benefits at all. Corrected estimates imply that there are indeed welfare gains from unemployment benefits and that the optimal replacement rate is equal to about one-quarter of pre-unemployment earnings. How much do unemployment benefits extend the time people remain unemployed? While many studies examine this question, some issues—like how the impact changes with the level of benefits—require combining insights from multiple studies. Additionally, conclusions about average effects may be misleading if published research doesn’t reflect the range of actual results. To address these challenges, the authors of this research apply meta-analysis techniques to findings from 57 prior studies. They systematically review microeconomic research estimating the causal effects of unemployment insurance—specifically, the potential benefit duration and the replacement rate—on how long replacement rate individuals remain unemployed. They find the following: •

The authors reveal a notable publication bias, where statistically significant findings are eight times more likely to be published. This bias results in an overestimation of the impact of unemployment benefits on unemployment duration, as studies with insignificant results are less likely to be included in

Figure 1 · Descriptive Evidence of Publication Bias Mean Elasticity Estimates

0.5 Mean Standard Error 0.4 0.3

Top 3rd Standard Error Tercile Middle 3rd Bottom 3rd

0.58 Null Hypothesis Under No Publication Bias

0.2 0.49

0.1 0.24 0

0.25

0.5 Mean Elasticity

0.75

Note: This figure plots the relationship between estimated elasticity values and their standard errors. If there

Note: This figurebias, plots relationship between elasticity and their were no publication we the would expect the mean elasticity estimated estimates across terciles tovalues be similar. Instead, the figure shows that estimates with larger standard errors tend to report higher elasticity values, suggesting standard errors. If there were no publication bias, we would expect the mean elasticity that studies with small, statistically insignificant elasticity estimates are less likely to be published. estimates across terciles to be similar. Instead, the figure shows that estimates with larger standard errors tend to report higher elasticity values, suggesting that studies with small, statistically insignificant elasticity estimates are less likely to be published.

the body of evidence. By accounting for this bias, the authors find that the average estimated effect of benefit levels on unemployment duration is reduced by between one third and one half. Correcting for this bias helps ensure that the findings reflect a more accurate and balanced understanding of the

Replacement rate: The percentage of a worker’s previous earnings that is covered by unemployment benefits during a period of joblessness. Statistically significant: A result in data analysis or hypothesis testing that is unlikely to have occurred due to random chance, based on a predetermined significance level (e.g., 5%). A statistically significant finding provides evidence to reject the null hypothesis, suggesting that the observed effect is real and not merely a result of random variation in the data.

bfi.uchicago.edu

relationship, avoiding misleading conclusions based on incomplete data. •

•

Meta-analysis improves traditional economic models for determining optimal unemployment benefits by correcting biases and forecasting the impacts of significant policy changes. Using these advancements, the authors estimate an optimal replacement rate of 28%, balancing the costs and benefits of unemployment insurance. There is no difference between the effect of increasing benefits for a single worker (the micro effect) and for all workers (the macro effect) in traditional economic models, suggesting that general equilibrium effects either have minimal impact or offset each other.

This review confirms that expansions in unemployment insurance benefits increase unemployment duration but reveals that previous studies likely overestimated this effect due to publication bias. Adjusting for this bias indicates that the true impact is significantly smaller, offering more accurate insights for policymakers. These findings underscore the value of meta-analytical approaches in refining policy recommendations and evaluating the broader economic impacts of unemployment benefits.

Figure 2 · Optimal Replacement Rates for Unemployment Insurance Optimal Replacement Rates for Unemployment Insurance 0.8 Welfare Gain or Elasticity

0% Elasticity (simple average for US)

0.6

Elasticity (US; corrected for publication bias, account for higher distortions from higher benefits)

0.4 16% 28%

Elasticity (US; corrected for publication bias)

0.2 Welfare Gain from ConsumptionSmoothing (Gruber 1997)

50 Replacement Rates

Note: This figure illustrates the estimated optimal replacement rate for unemployment benefits, balancing the

Note: figure illustrates the estimated optimal replacement rate for unemployment costs ofThis longer unemployment durations against the benefits of financial support. The downward-sloping gray line represents the welfarethe gains from consumption smoothing, and the upward-sloping yellow line the benefits, balancing costs of longer unemployment durations against theillustrates benefits elasticity of unemployment duration to replacement rates, corrected for publication bias and adjusted for varying of financial support. downward-sloping gray line represents the welfare from baseline replacement rates.The The intersection of these two lines indicates the optimal replacement rate.gains Additional lines (green and blue) show how the predicted elasticity changes when publication bias is uncorrected or consumption smoothing, and the line illustrates the elasticity elasticity heterogeneity is not accounted for,upward-sloping leading to alternativeyellow estimates of the optimal replacement rate. of unemployment duration to replacement rates, corrected for publication bias and adjusted for varying baseline replacement rates. The intersection of these two lines indicates the optimal replacement rate. Additional lines (green, red, and blue) show how the predicted elasticity changes when publication bias is uncorrected or elasticity heterogeneity is not accounted for, leading to alternative estimates of the optimal replacement rate.

General equilibrium effects: The broad economic impacts that arise when changes in one market or policy influence other interconnected markets and agents in the economy.

READ THE WORKING PAPER NO. 2024-96 · AUGUST 2024

Disemployment Effects of Unemployment Insurance: A Meta-Analysis bfi.uchicago.edu/working-papers/disemployment-effectsof-unemployment-insurance-a-meta-analysis

100

ABOUT OUR SCHOLAR

Peter Ganong

Associate Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Goals, Expectations, and Performance Based on BFI Working Paper No. 2025-21, “Goals, Expectations, and Performance,” by Avner Strulov-Shlain, University of Chicago, and Alexandra Steiny Wellsjo, University of California San Diego

Goals mostly reflect existing expectations rather than set expectations, and while eliciting a goal can improve performance, those positive returns come from increasing motivation on a task rather than from setting a harder or easier goal. Do you set personal work goals for yourself at the beginning of each year? Does your employer ask you to list goals that improve your performance and the performance of the company? If so, do you set goals that you are likely to attain anyway (self-efficacy)? Do you set goals that only seem better because they improve on relatively poor performance (reference-dependence)? Do you “shift the goalposts” to align with expectations (changing the stakes)? In any case, how do you— or how does a company—evaluate whether goalsetting systems are successful? While research has shed light on these and related questions, our understanding of the causal relationship between goals and expectations, and how goals and expectations impact performance, remains limited. This paper employs a novel experimental design to address that gap. The authors assign participants a task (counting 1s in tables of 0s and 1s), varying participants’ self-set goals and expectations at the start of the task. The authors vary the level of practice difficulty and by that change participants performance expectations; and they ask (some) participants to set goals for the number of tables

Figure 1 Effects · Treatment Effects Goals and Expectations Treatment on Goals and on Expectations 30 Tables

25 Goal Median Expected Performance

Hard

Medium

Practice Difficulty

Easy

Attainable

Baseline

Ambitious

No Goal

Goal Treatments

Note: ThisThis figure plots the regression-adjusted average goals goals (blue) (blue) and median expected Note: figure plots the regression-adjusted average and median expected performance (red) across treatment groups, with 95% intervals. On theOn leftthe left panel, performance (red) across treatment groups, withconfidence 95% confidence intervals. panel, the authors plot the effects of Practice Difficulty assignment, and on the right the the authors plot the effects of Practice Difficulty assignment, and on the right the effects of effects of Goal Elicitation treatments. See working paper for more details.

Goal Elicitation treatments. See working paper for more details.

to solve and encourage some of them to make the goals easier or harder. Among other features, the authors’ experimental design allows them to learn whether different expectations lead to different goals; whether different goals—given the same initial

bfi.uchicago.edu

Figure 2Effects · Treatment Effects on Performance Treatment on Performance

the hard practice. These effects hold both when participants set an explicit goal or when no goal is elicited.

14 Tables Solved

•

Hard

Medium

Easy

Practice Difficulty

Attainable

Baseline

Ambitious

No Goal

Goal Treatments

Note: ThisThis figure plots regression-adjusted average performance across across treatment groups,groups, Note: figure plots regression-adjusted average performance treatment with 95% confidence intervals. On the left panel, the authors plot the effects of Practice with 95% confidence On the theeffects left panel, the Elicitation authors plot the effects of Practice Difficulty assignment, andintervals. on the right of Goal treatments. assignment, and on the right the effects of Goal Elicitation treatments. See working SeeDifficulty working paper for more details.

paper for more details.

expectations—lead to revised post-goal-setting expectations; and how different combinations of goals and expectations affect performance in the main task and post-completion sentiments. They find the following: •

Participants set goals they expect to reach on average 78% of the time.

•

When participants are asked to set ambitious or attainable goals, they revise their goals by about +/-25%. These goal revisions have a limited impact on expected performance; instead, participants adjust their expected likelihood of reaching the goal.

•

There is no evidence that eliciting a goal affects expectations; beliefs in the Baseline and No Goal treatments are similar. In other words, goals reflect expectations but largely do not affect them.

•

Changing the practice tables changes performance. Participants in the easy practice solve 1.9 (17%) more tables than those with

READ THE WORKING PAPER NO. 2025-21 · FEBRUARY 2025

Goals, Expectations, and Performance bfi.uchicago.edu/working-papers/goals-expectations-andperformance

Eliciting a goal improves performance. Participants in the Baseline goal treatment complete 0.77 more tables than the No Goal treatment (a 6.5% increase). Further, the average positive effect of having a goal on performance is not driven by the effect of increasing performance to hit a goal, but instead by a broader increase in performance. In other words, goals increase motivation on a task, or the perceived returns to any additional completed task, regardless of distance to the goal. That said, participants do care about the goals and try to meet them, but the effect on performance is local. Some do more to hit the goal, some do less once they did.

Bottom line: Expectations matter. Having a goal is motivating, but conditional on expectations, the difficulty of a goal is not instrumental in changing performance. Let’s make this real, based on an example provided by the authors. Imagine sitting down with your manager to discuss goals. Going into the meeting, you have some expectations about what you can achieve, and you eventually agree to some goal. This research suggests that getting to this goal matters more than the difficulty of the goal. What does this mean in practice? For example, imagine that your manager convinced you to set a goal beyond your expectations. However, this is likely a goal that you will not reach, which is suboptimal for you, your manager, and the company. Instead, imagine a more motivational manager who convinces you of her vision and inspires you to do more, irrespective of some goal. In this instance, rather than putting resources toward choosing goals, efforts to increase expectations result in better outcomes.

ABOUT OUR SCHOLAR

Avner Strulov-Shlain

Assistant Professor of Marketing and Willard Graham Faculty Scholar, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Interest Rate Risk in Banking Based on BFI Working Paper No. 2025-04, “Interest Rate Risk in Banking,” by Peter M. DeMarzo, Stanford University; Arvind Krishnamurthy, Stanford University; and Stefan Nagel, University of Chicago

The present value of banks’ future earnings declines when interest rates rise, contradicting existing models that assume the value of banks’ deposit and lending business rises with interest rates. Recent interest rate hikes therefore exposed banks to losses not only on their long-term securities holdings, but also on the value of their banking franchise. Even so, most U.S. banks retained sufficient franchise value to remain solvent. In 2023, the U.S. banking system experienced significant upheaval, marked by the collapse of Silicon Valley Bank (SVB) and emergency government interventions to stabilize other midsized banks. SVB failed after suffering severe losses on long-term securities as interest rates long-term securities surged. The problem for SVB was that it had invested a large portion of its deposits, most of them uninsured, in long-term securities. When interest rates spiked, those securities plummeted in value and the bank was forced to sell off bonds to cover losses, which triggered a run by panicky depositors who feared losing their money. Within two days the bank collapsed. To contain the crisis and prevent broader financial contagion, regulators took extraordinary measures, including guaranteeing all deposits at SVB and launching new liquidity programs for struggling banks. These events exposed that the effects of interest rate hikes on bank solvency are poorly understood. Existing models and regulatory

Figure 1 · Franchise Value as a Buffer Against Interest Rate Losses Franchise Value as a Buffer Against Interest Rate Losses 40% Share of Banks Total Financial Cushion (Securities + Franchise Value) 30

10 Losses from Securities Holdings 0

-0.25

-0.15

-0.05

0.05

0.15

Note: This figure shows the distribution of security losses in 2023 and how franchise value

Note: This figure shows distribution ofexperienced security losses in 2023 andonhow franchise value helped offset those losses.the While many banks significant losses securities holdings and their franchise due to rising interest rates, mostsignificant remained solvent helped offset those losses.value While many banks experienced lossesbecause on securities their remaining franchise value still provided a sufficiently large cushion. holdings and their franchise value due to rising interest rates, most remained solvent because their remaining franchise value still provided a sufficiently large cushion.

frameworks are built on assumptions that imply that the present value value of banks’ earnings from lending and deposit-taking net of operating costs—their franchise value—rises with interest franchise value

long-term securities: financial assets, such as bonds or mortgage-backed securities, that mature over an extended period (typically more than a year) and are sensitive to interest rate changes present value: the current worth of future cash flows, discounted to reflect the time value of money franchise value: the present value of a bank’s future profits from lending and deposit-taking net of operating costs

bfi.uchicago.edu

rates. For example, one such assumption is that operating costs are insensitive to interest rate changes, while the spread between lending and deposit rates earned by banks rise with interest rates. In this case, the present value of future operating costs falls in response to higher interest rates, while the present value of future cash flows earned from spreads stays stable, leading to an overall increase in value. However, empirical evidence that the components of banks’ earnings conform to these assumptions is lacking. This paper fills this gap by analyzing financial reports from U.S. banks spanning 1984 to 2021 to analyze how banks generate revenue from loans and deposits, while accounting for operating costs. They find the following: •

Banks’ franchise value has positive duration duration, meaning it declines when interest rates rise. This contradicts prior models that predict that the franchise value would rise with interest rates. The long-term securities held by many banks also have positive duration, which means these securities holdings exacerbate rather than hedge the interest rate risk of banks’ lending and deposit-taking business.

•

Banks that do not raise their deposit rates significantly in response to rising interest rates tend to invest more in long-term securities. While this strategy helps insulate banks’ cash flows from exposure to interest rate fluctuations, it increases solvency risk because the value of those securities and the bank’s franchise value falls simultaneously when interest rates rise.

•

Despite significant recent rate hike losses, most U.S. banks still retain sufficient franchise value to remain solvent, justifying forbearance forbearance.

The authors highlight a trade-off between cashflow hedging, which stabilizes earnings, and value hedging, which protects solvency. While holding longer-term securities can help smooth earnings, it also increases duration risk, making banks more vulnerable to interest rate fluctuations. The findings suggest that for regulators’ models to accurately reflect the solvency risk resulting from interest rate hikes, they must reflect the positive duration of bank franchise value.

positive duration: a measure of interest rate sensitivity indicating that an asset’s value declines when interest rates rise forbearance: a temporary regulatory allowance that lets banks delay recognizing losses or meeting certain financial requirements to help them remain solvent during crises

READ THE WORKING PAPER NO. 2025-04 · DECEMBER 2024

Interest Rate Risk in Banking bfi.uchicago.edu/working-papers/interest-rate-risk-in-banking

ABOUT OUR SCHOLAR

Stefan Nagel

Fama Family Distinguished Service Professor of Finance, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Student Loan Forgiveness Based on BFI Working Paper 2025-23, “Student Loan Forgiveness,” by Michael Dinerstein, Duke University; Samuel Earnest, Massachusetts Institute of Technology; Dmitri K. Koustas, University of Chicago; and Constantine Yannelis, University of Cambridge

Student loan forgiveness increases consumption in the short term, with sharp increases in mortgage, auto, and credit card debt following loan forgiveness, and with a negative effect on earnings and the probability of being employed. Since 2010, outstanding student debt and debt per borrower in the United States have increased by 115% and 73%, respectively. Total student loan debt reached $1.77 trillion at the end of 2024, with the average loan balance over $40,000. With increasing concern about the effect of debt on borrowers’ livelihoods, as well as impacts to the broader economy, policymakers have called for broad-based student loan forgiveness, with goals ranging from redistributing toward low earners to providing economic stimulus. Despite this policy momentum and despite assumptions about the effects of such policies, we still do not have consensus on whether forgiveness will increase or decrease outcomes like spending and earnings. To address this gap, this paper analyzes the largest student loan discharge in US history. Beginning in March 2021, the federal government ordered $132 billion in student loans cancelled, or 7.8% of the total $1.7 trillion in outstanding student debt. To assess how forgiveness is targeted and how it affects forgiven borrowers’ consumption, debt, and earnings, the authors study comprehensive national administrative data from TransUnion, one of the largest credit bureaus, complemented by employment records obtained from a second large credit bureau. The TransUnion panel data comprise a 10 percent sample of all individuals who have a credit history

Figure 1 · Borrowers Forgiven and Forgiveness Announcements Borrowers Forgiven and Forgiveness Announcements PSLF $42 Billion

280K Borrowers 240

PSLF $5.2 Billion

200 160 120

Permanent Disability $5.8 Billion

Borrower Defense $5.8 Billion

80 40

IDR Adjustment $39 Billion

0 -40 2021

PSLF $7 Billion 2022

2023

2024

Note: This figure plots the number of borrowers in the sample that have been identified as receiving student

Note: This figure plots theThe number of borrowers thethe sample that have of been identified as loan forgiveness in each month. corresponding dashed linesinmark largest Department Education (DOE) debt student relief announcements along with the reported amountThe of debt to be discharged.dashed These debt receiving loan forgiveness in each month. corresponding lines mark the amounts represent amount of debt the DOE anticipates will be forgiven under each policy adjustment. largest Department of Education (DOE) debt relief announcements along with the reported TransUnion borrower numbers have been scaled by 10 to report a national estimate. Please see working paper for more details. Source: TransUnion and ed.gov amount of debt to be discharged. These debt amounts represent amount of debt the DOE anticipates will be forgiven under each policy adjustment. TransUnion borrower numbers have been scaled by 10 to report a national estimate. Please see working paper for more details. Source: TransUnion and ed.gov

in the United States, and the employment data cover around 1/3 of the US workforce. Six percent of borrowers in the authors’ sample received forgiveness since March 2021, with an average of $32,000 discharged. The authors find the following: •

Predicted monthly earnings of forgiven borrowers were $115 higher than borrowers who did not receive forgiveness and $193 more than the general population.

bfi.uchicago.edu

Figure 2 ·Credit Effects on Credit Outcomes Effects on Outcomes A) Mortgage Balances

B) Earnings - Level

$5K

50 New Mortgage Debt +$2,300 in mortgage borrowing

Monthly Earnings -$44 average decrease 6 months after forgiveness

25 0

3 -25

Month of Forgiveness 2

-50

-75

-100 -125

-1

-6

-4

-2 0 2 Months Before and After Forgiveness

-6

-4

-2 0 2 Months Before and After Forgiveness

Note: This figure plots estimates, 95% confidence intervals (yellow) and 90% confidence intervals (blue) of the average treatment effect on the treated in each of the six months leading up to and after forgiveness.

Note: This figure plots 95%yet confidence intervals (yellow) andforgiveness 90% confidence ofpaper the average treatment effect on the treated in each ofdata the(right). six months leading up to and Comparison borrowers areestimates, those who have to have forgiveness or never will have in sample.intervals Please see(blue) working for more details. Source: TransUnion (left) and employment after forgiveness. Comparison borrowers are those who have yet to have forgiveness or never will have forgiveness in sample. Please see working paper for more details. Source: TransUnion (left) and employment data (right).

•

Student loan forgiveness led to increases in mortgage, auto, and credit card debt by 9 cents for every dollar forgiven. Borrowers experiencing forgiveness increase mortgage borrowing by $2,300, auto loan borrowing by $230, and credit card borrowing by $220 over the six months following forgiveness.

•

There is little to no effect on non-student loan delinquencies.

•

Monthly earnings dropped by $44 (or 2.3%) pooled over the first six months postforgiveness, with the decrease exceeding $75 in the sixth month, suggesting that borrowers may have been in higher-paying jobs or provided increased labor supply in part to pay back debt.

•

There is more job switching across industries and out of public service. This latter effect is likely because forgiven borrowers no longer need to work in public service to qualify for future forgiveness; on that note, the authors estimate larger earnings drops for individuals initially employed in public service jobs.

READ THE WORKING PAPER NO. 2025-23 · FEBRUARY 2025

Student Loan Forgiveness bfi.uchicago.edu/working-papers/student-loan-forgiveness

•

Finally, among hourly workers, the authors estimate a drop in hours worked, where the total earnings drop comes half from an hours reduction and half from a wage reduction. Labor market effects are largest for younger workers with lower earnings, hourly workers, and public service workers. For previously defaulted borrowers, the authors find that earnings increase.

Bottom line: Student loan forgiveness increases consumption in the short term, with sharp increases in mortgage, auto, and credit card debt following loan forgiveness, and with a negative effect on earnings and the probability of being employed. While this work helps policymakers understand the likely outcomes of loan forgiveness plans, future work could study optimal relief for borrowers, and how insurance acts with distributional and macroeconomic consequences of loan forgiveness and other policies to assist student debtors.

ABOUT OUR SCHOLAR

Dmitri Koustas

Assistant Professor, Harris School of Public Policy

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Supply Chain Shocks and Firm Productivity: The Role of Reporting Quality Based on BFI Working Paper No. 2025-11, “Supply Chain Shocks and Firm Productivity: The Role of Reporting Quality,” by Philip G. Berger and Rimmy E. Tomy, University of Chicago

Firms with higher reporting quality prior to a shock experience an 11%–12% increase in total factor productivity (TFP) relative to unaffected firms, with effects concentrated among firms that restructure or make significant cuts in expenditures. A wealth of research has explored how financial and managerial reporting quality influences companies’ day-to-day decisions, from operations to investments. But what happens when firms face a major crisis? Can strong reporting systems—for example, those with rigorous procedures relating to finance and supply chain disclosures, and with few incidents of regulatory and fraud investigations—help firms adapt and recover more effectively? To date, little research has examined whether reporting quality shapes how well firms restructure after an unexpected shock. This paper tackles this question by investigating how firms’ pre-shock reporting quality influences their ability to regain productivity following a disruption. The authors examine this question in the context of the 1999 Taiwan earthquake, which disrupted the supply chains of US high-tech manufacturing firms that relied on Taiwanese suppliers for semiconductors and other electronic

Figure 1 · Total Factor Productivity over Time Total Factor Productivity over Time 0.3 Total Factor Productivity

Affected Firms 0.2 Time of Earthquake 0.1

Unaffected Firms -0.05

1994

1995

1996

1997 1998 1999 2000 2001 Time Before and After Taiwan Earthquake

2002

2003

2004

Note: This figure displays of total factor productivity (TFP) over time, separately for affected and unaffected firms.

Note: displays ofintervals. total factor productivity (TFP) over time, separately for affected ShadedThis areasfigure show 95% confidence and unaffected firms. Shaded areas show 95% confidence intervals.

components. The shock drove up the prices of inputs for these firms, while their competitors sourcing from elsewhere remained unaffected. The authors use a difference-in-differences difference-in-differences research design to compare firms that were and

difference-in-differences: a statistical method that estimates the causal effect of a treatment or shock by comparing changes over time between an affected group and a similar but unaffected control group

bfi.uchicago.edu

were not affected by the earthquake. They find the following: •

Following the earthquake, affected firms were more likely to incur restructuring charges, while also reducing spending on overhead, capital, and research. Affected firms did not significantly change their acquisition activity or workforce size. These changes are consistent with theory: because many competitors were unaffected by the earthquake, affected firms could not pass higher costs to consumers without risking market share loss and were instead prompted to adjust their operations.

•

Firm-level total factor productivity productivity(TFP) (TFP) increased by 10-11% for affected firms compared to unaffected firms following the earthquake, equivalent to a ten-percentile gain in productivity. This finding suggests that operational adjustments made in response to disruptions can drive significant productivity gains.

•

Reporting quality drives these productivity gains. Affected firms with high reporting quality experienced significant increases in TFP, ranging from 11.3% to 12.6%. By contrast, affected firms with low reporting quality either saw no increases in TFP relative to unaffected firms or witnessed a much lower

increase in TFP (5%) compared to firms with high reporting quality. •

The association between TFP and reporting quality only holds among firms that undertake restructuring or reduce investments following the earthquake. This suggests that superior accounting practices and enhanced disclosures are crucial for firms that actively manage their operations in response to shocks, potentially aiding in more effective decision-making and resource allocation.

This research highlights the critical role of reporting quality in shaping firms’ ability to respond productively to disruptive shocks. Firms with higher pre-shock reporting quality achieve greater increases in TFP, particularly when they engage in restructuring or strategic cost reductions. These findings suggest that highquality accounting practices improve decisionmaking and resource allocation under crisis conditions. For policymakers, this underscores the importance of regulatory frameworks that promote robust financial reporting, as higher reporting quality enhances firm resilience and adaptability. For firms, the results indicate that investments in stronger internal information systems can yield long-term productivity benefits, particularly in industries vulnerable to supply chain disruptions and other external shocks.

total factor productivity (TFP): a measure of how efficiently a firm converts inputs (labor, capital, and materials) into output, capturing productivity improvements beyond changes in input usage

READ THE WORKING PAPER NO. 2025-11 · JANUARY 2025

Supply Chain Shocks and Firm Productivity: The Role of Reporting Quality bfi.uchicago.edu/working-papers/supply-chain-shocksand-firm-productivity-the-role-of-reporting-quality

ABOUT OUR SCHOLARS

Philip G. Berger

Wallman Family Professor of Accounting, Chicago Booth

Rimmy E. Tomy

Associate Professor of Accounting and Kathryn and Grant Swick Faculty Scholar, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

Talking about Words Boosts Preschool-Age Children’s Vocabulary: Evidence from a Parent Intervention Based on “Talking about Words Boosts Preschool-Age Children’s Vocabulary: Evidence from a Parent Intervention,” BFI Working Paper No. 2025-24, by Derek Rury, Ariel Kalil, Susan Mayer, and Daniela Bresciani, University of Chicago

Sending conversation prompts to low-income parents encouraging them to talk with their preschool-aged children about vocabulary words leads to growth in children’s vocabulary and strengthens parents’ beliefs that parental input helps children learn. Language development is a fundamental aspect of human capital formation, and researchers continue to explore the most effective ways to support young children in acquiring new vocabulary. Evidence indicates that parental interaction plays a crucial role, yet precisely how much or why conversations with parents influence early vocabulary development remains unclear. These questions are particularly significant for children in low-income families, where parents are less frequently observed to engage their children in conversation. In this paper, the authors run a six-month evaluation of Chat2Learn, a program they designed to increase how much parents talk to their children about vocabulary words using subjects of interest to the child. The authors randomize a sample of 600 lowincome parents of preschool-aged children into three groups, summarized in Figure 1.

Figure 1 · Three Experimental Approaches Three Experimental Approaches

Every parent receives three text messages per week for six months, with content varying by treatment group: Definition Approach Parents receive a text message containing a word and its definition to share with their child.

Conversation Approach Parents receive the same content as the definition approach, and in addition, the text message prompts them to help engage their child in conversation and imaginative thinking. Control Text messages provide the parent only with information unrelated to children’s literacy skills and provided no specific prompt for talking.

Example APPLAUDING is when you clap your hands to show that you like something. What does APPLAUDING mean?

Example APPLAUDING is when you clap your hands to show that you like something. Can you sing a song? I’ll APPLAUD when you’re done!

Example Food isn’t just for nutrition. Food is also a link to culture! [child name] will grow up with fond memories of food from your family’s culture.

human capital: the skills, knowledge, and abilities that individuals acquire through education, training, and experience, which enable them to realize their potential as productive members of society

bfi.uchicago.edu

Different Learning Approaches FigureEffects 2 · Effectsfor for Different Learning Approaches

-2

Larger OLS Regression Estimate

Effect Size

10 8 6 4 2 0

B) Conversation Approach

Smaller

Effect Size

Larger

A) Definition Approach

10 20 30 40 50 60 70 80 90 Vocabulary Score Percentile

10 8 6 4 2 0 -2

10 20 30 40 50 60 70 80 90 Vocabulary Score Percentile

Note: This figure displays results from quantile regressions for the vocabulary and curiosity outcomes. The quantile treatment effects represent the differences in the outcome distribution between This figure displays results from quantile regressions for the vocabulary and curiosity outcomes. the treatment and control groups at each decile. For comparison, the OLS estimates of the treatment effects are included in the respective figures. Both treatments exhibit non-linearities in the The quantile effects represent thewhile differences in the outcome distribution between the in vocabulary. overall vocabulary score, withtreatment lower deciles showing no significant effects, higher deciles—specifically deciles four through nine—experience increases

treatment and control groups at each decile. For comparison, the OLS estimates of the treatment effects are included in the respective figures. Both treatments exhibit non-linearities in the overall

The authors’ two treatment arms reflect vocabulary score, with lower deciles showing no significant effects, while higher deciles—specifically distinct learning modalities. The definition deciles four through nine—experience increases in vocabulary. Related UChicago Work approach emphasizes simply learning the word Making a Song and Dance About It: The Effectiveness of definition, while the conversation approach Teaching Children Vocabulary with Animated Music Videos encourages parents to engage in open-ended This paper uses a randomized controlled trial to evaluate discussions designed to foster curiosity in the effectiveness of Big Word Club (BWC), a classroom children. The authors hypothesized that the program that uses music and dance videos for 3-5 conversation approach would enhance children’s minutes per day to increase vocabulary. The authors find that treated students scored higher on a test of words curiosity, which, in turn, could further support targeted by the program (0.30 SD) after four months of vocabulary development. use and this effect persisted for two months.

At both the beginning and end of the evaluation, the authors administer surveys to participating parents to assess their beliefs about their child’s skills, interests, and other attitudes and perceptions. Additionally, they evaluate children’s vocabulary and curiosity via Zoom before and after the intervention. By comparing results across the three experimental groups, the authors find the following: •

Both the definition and conversation approaches significantly improved children’s vocabulary knowledge compared to the control group, with no statistically significant difference between the two treatment effects. The observed gains were primarily driven by the acquisition of words included in the intervention.

•

Among children who scored highly on the vocabulary assessment, the definition approach also enhanced vocabulary for words not explicitly included in the intervention, suggesting spillover effects on overall language development.

•

The definition approach led to a fivepercentage-point reduction in parents’

It All Starts with Beliefs: Addressing the Roots of Educational Inequities by Shifting Parental Beliefs The authors design two field experiments to explore if changing parental beliefs can be a pathway to improving parental investments in young children. In the first field experiment, over a six-month period starting three days after birth, we use informational nudges informing low-SES parents about skill formation and best practices to foster child development. In the second field experiment, we use a more intensive home visiting program consisting of two visits per month for six months, starting when the child is 24-30 months old. Both field experiments induce parents to revise their beliefs and increase investments in their child. Boosting Parent-Child Math Engagement and Preschool Children’s Math Skills: Evidence from an RCT with LowIncome Families This paper uses a randomized controlled trial to test whether high-quality digital apps and analog math materials could increase parental math engagement and child skills and whether the impact is enhanced with text messages aimed at managing parents’ behavioral biases. Relative to the control group, neither the analog math materials alone nor the analog materials with growth mindset messages increased child math skills during the intervention period. However, the analog math materials combined with messaging to manage present bias and the digital tablet with math apps increased child math skills by about 0.20 standard deviations (p=.10) measured six months after the intervention.

belief that intelligence is fixed, lowering this perception from a baseline average of 18%. •

Neither treatment approach affected parents’ levels of stress, fatigue, or enjoyment of learning activities.

This evaluation demonstrates that parents can effectively teach their children new words using simple, cost-effective tools like Chat2Learn. In principle, parents can teach their children an unlimited number of words. The relatively short duration of the intervention and the young age of participants limited exploration of longerterm effects.

READ THE WORKING PAPER

ABOUT OUR SCHOLARS

NO. 2025-24 · FEBRUARY 2025

Derek Rury

Talking about Words Boosts Preschool-Age Children’s Vocabulary: Evidence from a Parent Intervention bfi.uchicago.edu/working-papers/talking-about-wordsboosts-preschool-age-childrens-vocabulary-evidencefrom-a-parent-intervention

Postdoctoral Researcher, Harris School of Public Policy

Ariel Kalil

Daniel Levin Professor, Director of the Center for Human Potential and Public Policy (CHPPP), Harris Public Policy

Susan Mayer

Professor Emeritus, Harris School of Public Policy

Daniela Bresciani Andaluz

Predoctoral Research Professional, BIP Lab, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • FEBRUARY 2025

The Anatomy of the Great Terror: A Quantitative Analysis of the 1937-38 Purges in the Red Army Based on BFI Working Paper No. 2024-154, “The Anatomy of the Great Terror: A Quantitative Analysis of the 1937-38 Purges in the Red Army,” by Alexei Zakharov, Yale University; and Konstantin Sonin, University of Chicago

Stalin’s military purge sought to preemptively minimize the risk of a possible coup, and he did so by targeting the most competent officers, directly impacting the disastrous Red Army performance in the first years of the German invasion. Historically, purges are one of the main instruments used by authoritarian rulers to keep opposition in check, including both masses and elites. Purges of military elites are commonly used by autocratic rulers to reduce threats—real or perceived—to their rule. In the case of Joseph Stalin and the Great Terror of 1937-38, over half of 1,844 Union of Soviet Socialist Republics (USSR) army officers who held general-grade military ranks were repressed, and at least 780 were executed, including three of the country’s five highest-ranked officers. This episode has particular significance as historians consider the purge a primary reason for the Soviet Army’s failed response to Hitler’s invasion of the USSR in June 1941. In this paper,

Figure 1 · Timeline of Arrests Among the Soviet High Command Timeline of Arrests Among the Soviet High Command 100 Arrests

80 Total Executed or Died Before Sentence 60

0 1936

1937

1938

1939

1940

Note: This figure shows the number of arrests among general-grade Soviet officers each month of Note: This figure shows arrests general-grade Soviet the 1936-1939 period. There the werenumber 11 arrests of in 1936, 13 inamong January-March 1937, 15 in April 1937,officers and 66 each in May 1937. The1936-1939 arrests peaked in June 1937-January and were mostly by the end of 1938. period. There were 111938 arrests in 1936, 13over in January-March 1937, 15 in month of the Arrests made during 1938 were also much more likely to result in prison than in a death sentence.

April 1937, and 66 in May 1937. The arrests peaked in June 1937-January 1938 and were mostly over by the end of 1938. Arrests made during 1938 were also much more likely to result in prison than in a death sentence.

bfi.uchicago.edu

the authors compile a large dataset of Soviet generals that includes their background, positions within the military hierarchy, and service history. In addition, the data allow the authors to study the connections among fellow officers, shedding light on the probability of repression and the timing of arrests. The authors find that the following officers were targeted: 1.

Officers deemed more competent—Conditional on other characteristics such as rank, age, or ethnicity, officers who obtained a high rank had a higher probability of repression.

2. Younger officers—Youth significantly increased the risk of repression. 3. Together, competence and youth were a potent mix—A younger officer of two with the same rank was most often purged, given that such an officer (one who had achieved a given rank at an early age) was considered the most competent. While much of the historical record provides no evidence to explain the Great Terror (Stalin, for example, left no memoirs), this new analysis supports the hypothesis that the purge was preventive in nature and not random. This finding is counter to existing literature which argues that promotion decisions within autocracies prioritize loyalty over competence.

READ THE WORKING PAPER NO. 2024-154 · DECEMBER 2024

The Anatomy of the Great Terror: A Quantitative Analysis of the 1937-38 Purges in the Red Army bfi.uchicago.edu/working-papers/the-anatomy-of-thegreat-terror-a-quantitative-analysis-of-the-1937-38-purgesin-the-red-army

ABOUT OUR SCHOLAR

Konstantin Sonin

John Dewey Distinguished Service Professor, Harris School of Public Policy

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

Boosting Young Children’s Math Skill with Technology in the Home Environment; A Digital Library for ParentChild Shared Reading Improves Literacy Skills for Young Disadvantaged Children; Priming Parental Identity: Evidence from Experimental Data Based on BFI Working Paper No. 2025-28, “Boosting Young Children’s Math Skill with Technology in the Home Environment,” by Daniela Bresciani Andaluz, Ariel Kalil, Haoxuan Liu, Susan E. Mayer, and Rohen Shah, University of Chicago; BFI Working Paper 2025-127, “A Digital Library for Parent-Child Shared Reading Improves Literacy Skills for Young Disadvantaged Children,” by Kalil, Liu, Mayer, Shah, and Derek Rury, University of Chicago; and BFI Working Paper No. 2025-26, “Priming Parental Identity: Evidence from Experimental Data,” by Bresciani, Kalil, Liu, and Mayer

Students with stronger reading and math skills tend to perform better in school and earn higher incomes in adulthood. It is concerning, then, that children from low-income backgrounds enter school with weaker skills, on average, than their higherincome peers. Researchers theorize that a key driver of these early gaps is differences in parental engagement, with lower-income parents less frequently observed to engage with their children.

math skills; in A Digital Library for Parent-Child Shared Reading Improves Literacy Skills for Young Disadvantaged Children, the authors similarly test the effect of providing a digital library for parent-child shared reading on the literacy skills of low-income children; and in Priming Parental Identity: Evidence from Experimental Data, the authors use identity priming to encourage greater involvement in their children’s learning.

Motivated by these disparities, the papers summarized here examine interventions designed to enhance parental engagement. In Boosting Young Children’s Math Skill with Technology in the Home Environment, the authors provide digital apps and analog math materials to preschool aged children from diverse socioeconomic backgrounds and test the effect on children’s

Boosting Young Children’s Math Skill with Technology in the Home Environment In the United States, tailored programs designed to boost child math skills range from light-touch approaches (such as ones that send tips by text to parents about child development) to more complex

bfi.uchicago.edu

and targeted approaches that train practitioners to visit and coach parents in their homes. Few parent interventions designed to boost children’s math skills have been experimentally evaluated in large and diverse samples. In this experiment, the authors recruited families from 35 preschools throughout the City of Chicago to participate in About Time, a six-month randomized trial (RCT) (RCT) aimed at boosting randomized control trial parent-child engagement in math learning. The authors randomly assigned participants to one of three groups: a control group, a treatment group receiving a digital tablet with high-quality math apps, and another treatment group receiving equally high-quality analog math materials. The authors assessed children’s math skills at the outset of the intervention as well as after six months. They also collected a selfreported measure of parental time spent on math engagement, along with survey measures of parents’ attitudes and beliefs. The authors compared these data across their three experimental groups, and found the following: •

•

A Digital Library for Parent-Child Shared Reading Improves Literacy Skills for Young Disadvantaged Children On any given day, 29% of college-educated mothers report reading to their children, compared to only 12% of mothers with a high school diploma and 7% of mothers with less than a high school education. This educationbased gap in reading is the largest among early childhood investment activities linked to cognitive skill development. To examine ways to address this disparity, the authors conducted an RCT with 300 low-income families in Chicago. The study tested whether access to a digital library could increase reading frequency and improve child literacy outcomes. The intervention lasted 11 months, with families randomly assigned to one of four groups, detailed below.

In two-parent households, children who received math apps improved their math skills by 0.23 standard deviations compared to the control group. Providing analog math materials did not improve children’s math skills. The authors theorize that math apps were particularly effective in two-parent households because these environments may provide more support for learning, amplifying the benefits of parental engagement (consistent with Cunha and Heckman, 2007).

Providing math learning apps to families can be an effective way to moderately improve preschool-aged children’s math skills. These findings support efforts to expand at-home use of high-quality math apps, particularly as educational technology remains in the early stages of adoption in home environments. Increasing access to such tools could help enhance early math development and narrow achievement gaps before children enter school.

Control group

Received an activity book with crayons and stickers

Treatment 1

Received a tablet loaded with a digital library called Children and Parents Engaged in Reading (CAPER)

Treatment 2

Received the CAPER tablet along with weekly text messages reminding parents to read to their child

Treatment 3

Received the CAPER tablet along with weekly messages to set a goal for reading from the digital library in the week ahead.

The authors measured children’s literacy skills before and after their intervention. They compare between their four experimental groups, and find the following: •

Access to the digital library alone led to a significant improvement of 0.29 standard deviations in children’s literacy skills compared to families who did not receive it.

•

The behavioral messages offered no additional benefit beyond the library itself.

This experiment demonstrates the potential of technology to improve the literacy skills of low-

randomized controlled trial (RCT): a study design where participants are randomly assigned to either a treatment group or a control group to objectively measure the effects of an intervention

Week for Experiment 1 Figure 1 · Redemption Redemption RateRate Each Each Week for Experiment 1 50% Accumulated Redemption Rate Parental Identity Group 40

Hourly Wage Group

Control Group

20 End of Intervention 10

4 Weeks

This figure shows the accumulated rates byrepresent week for each treatment Note: This figureNote: shows the accumulated redemption rates by week for eachredemption treatment group. The stars the significance level of thegroup. redemption rate difference compared to the control group. The stars represent the significance level of the redemption rate difference compared to the control group.

income children. Digital libraries offer several advantages over large-scale programs that provide physical books to young children. A digital library can provide access to a significantly larger and more diverse collection of books, which can be updated and adjusted as children grow, offering greater flexibility and scalability.

Priming Parental Identity: Evidence from Experimental Data Parents whose parental identities are stronger— i.e., those who place a greater emphasis on their roles as parents as opposed to other roles in their lives—tend to invest more in their children’s development. In the final paper in this trio, the authors examine whether priming parental identity affects parents’ real-world decisions. The authors conduct two experiments using existing samples of Chicago parents who had previously received digital gift cards as compensation for participating in past studies. In each experiment, parents are randomly assigned to either a treatment or control group and receive weekly text messages for four weeks encouraging gift card redemption. In both experiments, the treatment messages appeal to parental identity by encouraging parents to redeem the gift card to buy something for their children. The control group messages

vary between the two experiments: In Experiment 1, the control group receives a generic reminder to redeem their gift card, while in Experiment 2, the control group message is designed to prime self-identity, encouraging parents to use the gift card for themselves. The authors then compare gift card redemption rates across the experimental groups and find the following: •

A text message highlighting what parents could purchase for their children with the gift card significantly increased the redemption rate by 45% compared to a generic reminder to redeem the gift card. The same prompt increased the redemption rate by 39% compared to a message encouraging parents to think about what they could buy for themselves.

•

Parents in the treatment group who received parental identity messages redeemed their gift cards at different stores compared to parents in the control group, suggesting that the parental identity message may have influence parents’ shopping decisions. (The authors caution, however, that since they lack data on the specific items purchased in these stores, they cannot directly show that parents who received parental identity messages were more likely to buy products for their children.)

By demonstrating that parental identity can be effectively primed through low-cost messaging,

Redemption Week for Experiment 2 Figure 2 · Redemption RateRate Each Each Week for Experiment 2 60% Accumulated Redemption Rate

Parental Identity Group

Self Identity Group

20 End of Intervention

4 Weeks

Note: This figure shows the accumulated redemption rates by week for each treatment group. The stars represent the significance level of the redemption rate difference compared to the control group.

this research contributes to a growing body of literature on identity and behavioral economics. These results hold promise for designing scalable interventions to encourage behaviors that promote child well-being, such as parental engagement, healthcare utilization, or savings for education. Future research could explore how identity

priming interacts with other contextual factors, such as socioeconomic status or cultural norms, to better understand its potential applications across diverse settings. Additionally, studying the long-term effects of identity-based interventions would offer insights into whether the observed behavioral changes persist over time.

READ THE WORKING PAPER

ABOUT OUR SCHOLARS

NO. 2025-28 · FEBRUARY 2025

Daniela Bresciani Andaluz

Boosting Young Children’s Math Skill with Technology in the Home Environment bfi.uchicago.edu/working-papers/boosting-young-childrensmath-skill-with-technology-in-the-home-environment NO. 2025-27 · FEBRUARY 2025

A Digital Library for Parent-Child Shared Reading Improves Literacy Skills for Young Disadvantaged Children bfi.uchicago.edu/working-papers/a-digital-library-forparent-child-shared-reading-improves-literacy-skills-foryoung-disadvantaged-children NO. 2025-26 · FEBRUARY 2025

Priming Parental Identity: Evidence from Experimental Data

BFI Research Professional, Harris School of Public Policy

Ariel Kalil

Daniel Levin Professor, Harris School of Public Policy

Haoxuan “Noah” Liu

PhD Student, Harris School of Public Policy

Susan E. Mayer

Professor Emeritus, Harris School of Public Policy

Rohen Shah

PhD Student, Harris School of Public Policy

Derek Rury

Postdoctoral Researcher (Instructor), Harris School of Public Policy

bfi.uchicago.edu/working-papers/priming-parental-identityevidence-from-experimental-data

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

Central Bank Communication with the Polarized Public Based on BFI Working Paper No. 2025-37, “Central Bank Communication with the Polarized Public,” by Pei Kuang, University of Birmingham; Michael Weber, University of Chicago; Shihan Xie, University of Illinois Urbana-Champaign

Individuals who view the Fed as politically aligned report higher independence of and trust in the Fed, leading to lower inflation expectations and uncertainty. Strategic communication on institutional structure and policy objectives mitigates perception biases. The Federal Reserve is designed to operate independently of short-term political pressures to maintain economic and price stability. However, the bank’s perceived neutrality has come under increasing scrutiny in today’s polarized political landscape. With President Trump’s return to office, political pressures on the Fed have intensified,

exemplified by demands for immediate interest rate cuts. These developments raise critical questions about public perceptions of the Fed’s independence, the role of partisanship in shaping trust, and how these perceptions influence macroeconomic expectations.

Figure 1 · Perceived Independence of the Federal Reserve Perceived Independence of the Federal Reserve Quasi-Constitutional

Quasi-Constitutional

No Tolerance

Institutional

Financial and Economic Independent

Personal Democrat

Republican

No Tolerance

Institutional

Financial and Economic

Personal Neutral

In-Group

Out-Group

This figure the perceived independenceofofthe theFed, Fed,measured measured on 1 indicates Strongly Disagree and 5and indicates Strongly Agree. Respondents rated the rated following Note: Note: This figure plotsplots the perceived independence onaa1 1toto55scale, scale,where where 1 indicates Strongly Disagree 5 indicates Strongly Agree. Respondents the following statements: (1) "The Federal Reserve’s legal foundation strongly protects it from political interference" (Quasi-Constitutional Independence); (2) "The Federal Reserve sets key policies, such as statements: (1) “The Federal Reserve’s legal foundation strongly protects it from political interference” (Quasi-Constitutional Independence); (2) “The Federal Reserve sets key policies, such as interest rates, without needing approval from government officials" (Institutional Independence); (3) "Appointments to the Federal Reserve’s leadership positions are made based on expertise interest rates, without rather needing approval from government officials” (Institutional Independence); (3) “Appointments the Federallimiting Reserve’s leadership positions made its based on expertise and qualifications than political loyalty" (Personal Independence); (4) "The Federal Reserve controls its own budget to and resources, the government’s ability toare influence actions" (Financial andrather Economic and(Personal (5) "The Federal Reserve will(4) not“The tolerate higher inflationcontrols rates in order to help reduce real valuelimiting of the U.S. debt" (No to Tolerance). and qualifications thanIndependence); political loyalty” Independence); Federal Reserve its own budget andthe resources, thegovernment’s government’s ability influence its actions” The left panel is categorized by respondents’ political affiliations (Independent, Democrat, Republican); the left panel illustrates the perceived independence across these dimensions based on (Financial and Economic Independence); and (5) “The Federal Reserve will not tolerate higher inflation rates in order to help reduce the real value of the U.S. government’s debt” (No Tolerance). respondents’ alignment with the Fed (in-group, out-group, or neutral). The left panel is categorized by respondents’ political affiliations (Independent, Democrat, Republican); the left panel illustrates the perceived independence across these dimensions based on respondents’ alignment with the Fed (in-group, out-group, or neutral).

bfi.uchicago.edu

In this paper, the authors use a large-scale survey experiment to examine public trust in survey experiment the Federal Reserve. Conducted on the day of President Trump’s 2025 inauguration, the survey includes responses from more than 5,600 U.S. participants, capturing their perceptions of the Fed’s independence across multiple dimensions, including quasi-constitutional, institutional, and financial and economic independence. The study also incorporates a randomized randomized controlled (RCT), where a subset of controlled trial (RCT) participants receives targeted information treatments—framed as Fed communication interventions—designed to inform them about the Fed’s institutional structure, policy objectives, and recent performance. The authors assess whether this messaging can enhance trust in the Fed. The survey and experiment reveal the following: •

Political alignment strongly influences perceptions of the Fed’s independence and credibility. Individuals who perceive the Fed as an “in-group” institution–aligned with their own political stance–consistently attribute higher independence scores, whereas those viewing the Fed as an “out-group” rate it as significantly less independent.

•

Greater perceived independence and trust in the Fed are associated with lower inflation expectations, a reduced perceived inflation target, and lower uncertainty about inflation and unemployment.

•

Providing information about the Fed’s institutional structure, its nonpartisan objectives, and its policy track record significantly increases trust in the institution and reduces perceptions of political bias. These interventions also prompt individuals to place greater weight on official Fed communications when updating their macroeconomic expectations.

This research builds on earlier work documenting partisan variations in trust in the Fed by examining whether targeted Fed communication can reduce those differences. For the Fed and other central banks, actively managing communications is essential to maintain credibility and to ensure effective policy transmission in a politically polarized environment. Future research should examine the influence of different framing techniques, media channels, and messenger credibility on public trust in monetary authorities.

survey experiment: a research method that embeds experimental design within a survey, randomly assigning participants to different conditions to measure causal effects on responses randomized controlled trial (RCT): a study design where participants are randomly assigned to either a treatment group or a control group to objectively measure the effects of an intervention

READ THE WORKING PAPER NO. 2025-37 · FEBRUARY 2025

Central Bank Communication with the Polarized Public bfi.uchicago.edu/working-papers/central-bankcommunication-with-the-polarized-public

ABOUT OUR SCHOLAR

Michael Weber

Associate Professor of Finance, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

Drive Down the Cost: Learning by Doing and Government Policies in the Global EV Battery Industry Based on BFI Working Paper No. 2025-15, “Drive Down the Cost: Learning by Doing and Government Policies in the Global EV Battery Industry,” by Panle Jia Barwick, University of Wisconsin-Madison; Hyuk-soo Kwon, University of Chicago; Shanjun Li, Cornell University; and Nahim Bin Zahur, Queen’s University

The learning rate for EV battery production is 7.5%, meaning costs drop by 7.5% when production experience doubles. Learning by doing enhances EV subsidies’ impact and creates global spillovers. Between 2010 and 2020, the cost of electric vehicle (EV) batteries dropped by nearly 90%, reducing a major obstacle to widespread EV adoption. Industry experts attribute much of this decline to learning-by-doing (LBD), where production experience leads to lower costs through improved efficiency and reduced waste. However, other factors, such as economies of scale and technological advancements, also play a role. In this paper, the authors quantify the impact of LBD on declining EV battery costs and examine how LBD influences the effectiveness of industrial policies like consumer subsidies and local content requirements. The authors use a detailed dataset covering global EV sales, vehicle characteristics, battery suppliers, and financial incentives to build a structural model of the EV market. Their model estimates battery costs by analyzing EV prices, sales trends, and the evolving partnerships between EV manufacturers and battery suppliers. It also accounts for how consumers with different preferences make purchasing decisions and how EV makers and battery suppliers set prices. The

Figure 1 · Effect of Subsidies and LBD on Global EV Sales Effect of Subsidies and LBD on Global EV Sales 3M in EV Sales (Units)

Subsidy & LBD

LBD Only 2M

Subsidy Only 1M

No Subsidy, No LBD

0 2013

2014

2015

2016

2017

2018

2019

2020

Note: This figure illustrates total EV sales across the top 13 EV countries under various scenarios, as labelled.

model reveals the following concerning the role of LBD in battery cost reductions and its broader implications for EV policy effectiveness: •

The learning rate is estimated to be 7.5% after controlling for technological advancements, experience in EV assembly, input costs, and economies of scale. This implies that doubling battery production experience would reduce unit production costs by 7.5%.

bfi.uchicago.edu

•

LBD greatly amplifies the sales impact of EV subsidies through positive feedback loops. In the absence of LBD, subsidies across different countries are estimated to increase cumulative global EV sales by 29.9% during the sample period, consistent with findings in existing studies that focus on the short-term effects of EV purchase subsidies. When both consumer subsidies and LBD were in effect, global EV sales surged by 170% relative to the baseline with neither subsidies nor LBD. This combined effect is 60% greater than the sum of the effects from subsidies and LBD individually, highlighting their complementarity. Consumer subsidies in one country generate global spillovers through LBD in battery production, but the magnitude of spillovers hinges critically on the nature of the supply chain network and trade patterns. For example, the estimated $13.10 billion in U.S. subsidies generated $16.47 billion in global welfare gains, measured as the sum of consumer surplus and firm profit on a global scale, net of subsidy expenditure. The U.S. (and Canada) captured 49% of these welfare gains, as the interaction between subsidies and LBD significantly reduced input costs (batteries) for domestic EV producers and lowered vehicle prices for domestic consumers. U.S. subsidies also benefited battery suppliers in Japan and South Korea, which captured 28% of the global welfare gains. Europe also benefited significantly from

READ THE WORKING PAPER NO. 2025-15 · JANUARY 2025

Drive Down the Cost: Learning by Doing and Government Policies in the Global EV Battery Industry

U.S. subsidies; in contrast, China captured only 3% of the global gains. This modest share reflects China’s limited trade in EVs and EV batteries with foreign countries during the authors’ sample period. •

Upstream LBD creates significant externalities through the supply chain, with upstream firms capturing only a small fraction of the associated economic benefits due to the oligopolistic nature of the supply chain.

•

Lastly, China’s whitelist policy benefited domestic battery suppliers at a cost to other countries. The EU, Japan and South Korea, and the U.S. and Canada collectively incurred $5.88 billion in welfare losses. This was driven by a shift in global battery production from more efficient Japanese and South Korean battery suppliers to (at the time) higher-cost Chinese suppliers. Had the whitelist policy been delayed to 2021-2024, China would have faced net losses, as consumer welfare losses would have outweighed the gains to battery suppliers. The negative impact on other countries would have been smaller.

While prior research has linked industrial policies, such as purchase subsidies, to EV innovation, this study is the first to integrate learning by doing (LBD) into EV market analysis. Prior studies may underestimate the full impact and costeffectiveness of subsidies by overlooking LBD and its reinforcing effect on lower battery costs and increased EV adoption.

ABOUT OUR SCHOLAR

Hyuk-soo Kwon

Assistant Professor, Harris School of Public Policy

bfi.uchicago.edu/working-papers/drive-down-the-costlearning-by-doing-and-government-policies-in-the-globalev-battery-industry

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

Effects of Unemployment Insurance for SelfEmployed and Marginally-Attached Workers Based on BFI Working Paper No. 2025-12, “Effects of Unemployment Insurance for Self-Employed and Marginally-Attached Workers,” by Emilie Jackson, Michigan State University; Dmitri Koustas, University of Chicago; Andrew Garin, Carnegie Mellon University

Expanding unemployment insurance through Pandemic Unemployment Assistance led self-employed individuals, gig workers, and new labor market entrants to reduce their earnings by $0.30, $0.48, and $0.22 for every additional dollar of UI received, respectively. UI benefits reduced mortality among older gig workers. Unemployment insurance (UI) serves as a critical Earnings Figure Response to $1 Increase in UI 1 · Earnings Response to Benefit $1 Increase in UI Benefit social safety net, benefiting both private and social 0.2 Change in Earnings welfare. However, traditional UI programs exclude several vulnerable worker groups, including the Reduction due to PUA self-employed, gig workers, and new labor market entrants. Extending UI to these workers presents 0 challenges, particularly in distinguishing voluntary from involuntary job separations. For instance, a self-employed worker might reduce work hours strategically to qualify for UI benefits, raising -0.2 concerns about moral hazard. hazard This paper examines the largest expansion of UI coverage since the program’s inception in 1935: Pandemic Unemployment Assistance (PUA), which extended UI benefits to self-employed

Placebo Estimates -0.4 2017-18 Platform Gig

2018-19 Other SE

2019-20

High School Graduation Cohort

moral hazard: in this context, a situation where a worker changes their work or job search behavior due to the Note: This figure shows the change in total earnings in response to a $1 increase in UI benefit, Note: This figure shows the change in total earnings in response to a $1 increase in UI benefit, unadjusted for differences in UI take-up, where the outcome in each regression is calculated availability of benefits unadjusted for differences in UI take-up, where the outcome in each regression is calculated over the years indicated on the x-axis. over the years indicated on the x-axis.

bfi.uchicago.edu

individuals, gig workers, and recent high school graduates entering the labor force during the COVID-19 pandemic. The authors use crossstate variation in PUA implementation to assess its effects on labor market earnings, education choices, and mortality using a spatial spatial regression regression discontinuity design. They find the following: •

Every additional dollar of UI benefit for selfemployed workers reduced earnings by $0.21. After accounting for differences in UI take-up rates, self-employed workers reduced their earnings by $0.30, with even larger reductions among gig workers ($0.48). Recent high school graduates reduced their earnings by $0.22 for every additional dollar of UI received.

•

Self-employed workers in industries less impacted by the pandemic exhibited larger earnings reductions, suggesting that UI expansion may have incentivized some individuals to withdraw from the labor force.

•

UI had notable health effects: an additional $5,000 in PUA benefits reduced mortality by 0.36 percentage points among older gig workers. There were no effects on mortality outside the gig economy and no effects on college attainment.

The findings underscore the trade-offs involved in expanding UI to non-traditional workers. While PUA provided an essential safety net during the pandemic, it also altered labor market participation, particularly among workers with flexible employment arrangements. Policymakers should consider targeting future UI expansions based on industry conditions to minimize moral hazard while maintaining essential protections for workers facing economic shocks.

spatial regression discontinuity: a research design that exploits differences across geographic borders to estimate causal effects while holding local economic conditions constant

READ THE WORKING PAPER NO. 2025-12 · JANUARY 2025

Effects of Unemployment Insurance for SelfEmployed and Marginally-Attached Workers bfi.uchicago.edu/working-papers/eﬀects-ofunemployment-insurance-for-self-employed-andmarginally-attached-workers

ABOUT OUR SCHOLAR

Dmitri Koustas

Assistant Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

How Costly Are Business Cycle Volatility and Inflation? A Vox Populi Approach Based on BFI Working Paper No. 2025-34, “How Costly Are Business Cycle Volatility and Inflation? A Vox Populi Approach,” by Dimitris Georgarakos, European Central Bank and CEPR; Kwang Hwan Kim, Yonsei University; Olivier Coibion, University of Texas at Austin; Myungkyu Shim, Yonsei University; Myunghwan Andrew Lee, New York University; Yuriy Gorodnichenko, New York University; Geoff Kenny, European Central Bank; Seowoo Han, Yonsei University; and Michael Weber, University of Chicago

Households are willing to forgo approximately 5–6% of their lifetime consumption to eliminate business cycle fluctuations and around 5% to achieve their desired inflation rate. This amount is higher among consumers whose consumption is more pro-cyclical, those facing greater economic uncertainty, and those living in countries with a history of higher economic volatility. Many economists, early among them UChicago Figure 1Experience · WTP and Historical Experience WTP and Historical Nobel Laureate Robert Lucas, have explored the 8% Reduced in Consumption WTP to Avoid Business Cycles cost of business cycles, advancing theories about South Korea how much consumers would be willing to sacrifice to avoid market fluctuations. The consensus 7 view, based on work by Lucas and others, is that households are willing to sacrifice very little to eliminate business cycles, while inflation appears United States much more costly. In this paper, the authors 6 Ireland Spain provide the first direct estimates of consumers’ Italy Portugal self-reported willingness to pay (WTP) to eliminate business cycle risk as well as their willingness 5 to pay to bring inflation to their ideal level. France Austria Understanding public perceptions of economic Germany conditions is essential for crafting policies that are 4 Belgium Finland not only effective but also widely supported.

Greece

Netherlands

The authors administer a series of large, nationally representative surveys in the United 3 1 2 3 4 5 6 States, Korea, and the eleven largest euro area Standard Deviation of Unemployment Rate, 1991-2023 countries: Austria, Belgium, Germany, Greece, Note: This figure shows the relationship between economic uncertainty, proxied by the standard deviation of the Spain, Finland, France, Ireland, Italy, Netherlands, Note: This figure shows the relationship between economic uncertainty, proxied by the employment rate between 1991 and 2023, and consumers’ WTP to avoid cycles. standard deviation of the employment rate between 1991business and 2023, and consumers’ WTP and Portugal. The survey collects respondents’ to avoid business cycles. willingness to pay (WTP): the maximum price a customer is willing to pay for a product or service (or in the case of this research, for a certain economic outcome)

bfi.uchicago.edu

sociodemographic information and measures their risk aversion, micro-and macroeconomic expectations, exposure to business cycles, and WTP to avoid macroeconomic risks. The authors validate survey responses by showing that self-reported WTP measures are consistent with theoretical predictions. For example, respondents in crisis-scarred countries like Korea and Greece report a higher WTP for macroeconomic stability, and those in countries that experienced high inflation in the past report a higher WTP for reduced inflation. The survey yields the following results: •

Consumers are willing to permanently reduce their consumption by 5-6% to eradicate macroeconomic volatility. Consumers are also willing to reduce their consumption by roughly 5% to bring inflation to their desired level. Both figures are much higher than those implied by traditional theory.

•

Individuals with more cyclical earnings and who are more uncertain about their future consumption are willing to pay more to eliminate business cycles as well as to reduce inflation. In addition, greater uncertainty about the macroeconomic outlook (e.g. uncertainty about GDP growth) raises people’s WTP to reduce macroeconomic volatility, even after controlling for their individual consumption uncertainty. This result indicates that people seem to care about business cycle volatility above and beyond their implications for individuals’ own consumption.

READ THE WORKING PAPER NO. 2025-34 · FEBRUARY 2025

How Costly Are Business Cycle Volatility and Inflation? A Vox Populi Approach bfi.uchicago.edu/working-papers/how-costly-are-businesscycle-volatility-and-inflation-a-vox-populi-approach

While many scholars have theorized on the cost of business cycles, this work provides the first direct test of consumers’ willingness to pay to reduce inflation and avoid business cycle fluctuations. The authors find that people perceive business cycles and inflation as very costly overall—roughly two orders of magnitude more costly than what is proposed by traditional theory. These perceptions, whether they are ultimately correct or incorrect, matter for their peoples’ decisions, for elections, for trust in institutions, and for the success of many policies that rely on confidence. Ignoring them is a recipe for designing policies that may well succeed in theoretical models, and perhaps even in practice, but may yet fail in the polls and ultimately be replaced by populist alternatives.

ABOUT OUR SCHOLAR

Michael Weber

Associate Professor of Finance, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

Income Equality in the Nordic Countries: Myths, Facts, and Lessons Based on BFI Working Paper 2025-25, “Income Equality in the Nordic Countries: Myths, Facts, and Lessons,” by Magne Mogstad, University of Chicago; Kjell G. Salvanes, Norwegian School of Economics; and Gaute Torsvik, University of Oslo

Income equality in the Nordic countries results from a severe compression of hourly wages that reduces the returns to labor market skills; this is achieved through a wage bargaining system with strong coordination within and between industries. Recent calls for economic and social equality within developed countries, including the United States, have brought renewed attention to Nordic countries, where equality is often equated with such policies as subsidized and readily available daycare, generous parental leave policy, universal health care, free college, and strong labor rights and unions. If these countries can have low inequality and economic growth, the reasoning goes, so can we. However, how well do these Nordic admirers understand the model employed by Denmark, Finland, Norway, and Sweden?1 To replicate the Nordic experience, is it simply a matter of implementing certain policies? What explains the Nordic model? This paper combines theory and evidence to examine these and related questions, and to address the challenges for those hoping to emulate the Nordic experience. Before describing the authors’ analysis, let us first broadly review the Nordic economies and what makes them distinctive. First, just 26 million people live in the four countries, with around 10 million in Sweden, roughly twice the size of the other 1 This work focuses on the four largest countries of the Nordic Region, which in total includes Denmark, Norway, Sweden, Finland, and Iceland, as well as the Faroe Islands, Greenland, and Åland. Nordic countries are often conflated with Scandinavia, which typically refers to Denmark, Norway, and Sweden.

Figure 1 · Timeline of Introduction of Social Policies in Nordic Countries (1950-2000)

Timeline of Introduction of Social Policies in Nordic Countries (1950-2000) Norway 1956 1957

1975 1977

Sweden 1955

1960

1974 1975

Finland 1962 1964

1973 1974

Denmark 1956

1961

1964

1967

Universal Maternity Leave

Universal Daycare

Universal Social Security

Universal Health Care

Note: This timeline represents the introduction of key social policies in the Nordic countries. Each point the year a policy was Note: Thisreflects timeline represents theenacted. introduction of key social policies in the Nordic countries.

Each point reflects the year a policy was enacted.

Nordic countries. Though commonly perceived as homogeneous, in 2021 both Norway and Sweden had proportionally larger foreign born populations than the United Kingdom and United States. Nordic residents are also relatively well-educated and healthy, and they enjoy life in countries with high quality-of-life indicators. Finland exports just under 50% of its GDP, while the other three export half or more of the goods and services they produce. Finland’s exports include machinery, electronics, paper products, and chemicals; Norway’s exports are predominantly

bfi.uchicago.edu

Figure 2Trends · Trends Union Density and Barganing Coverage in in Union Density and Barganing Coverage A) Labor Union Density

B) Collective Bargaining Coverage

100% Union Density

100% Bargaining Coverage

Nordic Countries

40 United Kingdom

Continental Europe

United States

0 1980

1990

2000

2010

0 1980

2020

1990

2000

2010

2020

Note: This the figure shows fraction of union members (Panel A) and the fraction of workers covered by collective bargaining agreements (Panel B) 2020 for the United States and Note: This figure shows fraction ofthe union members (Panel A) and the fraction of workers covered by collective bargaining agreements (Panel B) between 1980 and between 1980 and 2020 for the United States and selected European countries. “Continental Europe” includes France, Germany, Spain, and Portugal, and selected European countries. “Continental Europe”Sweden, includes Denmark, France, Germany, Spain,As and Portugal, and A, “Nordic Countries” includes Sweden, Denmark, andwith Finland. “Nordic Countries” includes Norway, and Finland. shown in Panel unionization varies widelyNorway, across advanced economies, the As shown in Panel A, unionization varies widely across economies, with the highest density rates the Nordic countries several times the differences lowest density ratesfrom in the1980 United highest density ratesadvanced in the Nordic countries reaching several times the in lowest density rates inreaching the United States. These expand to States. These differences expand from 2018, 1980 to as the of U.S. U.K.workers union workers has steadily declined over time. as 2018, the share ofshare U.S. and U.K.and union has steadily declined over time.

oil, gas, and fish; Sweden exports a substantial amount of manufactured goods; and Denmark is a leading exporter of chemical products, particularly pharmaceuticals. Average incomes in the Nordic countries are above the Organization for Economic Cooperation and Development (OECD) average and not far below the average income in the United States. Regarding GDP per capita (a key standard-of-living indicator), when this measure is decomposed for labor productivity (as proxied by GDP per work hour, and labor quantity as measured by work hours per capita), the Scandinavian trio (Denmark, Norway, and Sweden) is at least as productive as the United States, and considerably more productive than the United Kingdom and the OECD average. Finally, on average, Nordic citizens work fewer hours per year than their OECD counterparts. Employment rates and labor force participation rates, though, are relatively high. Indeed, the employment rate in the Nordic countries is higher than those in the United Kingdom and United States, and the OECD average, a difference that largely stems from a higher rate of female labor force participation. With that as background, what distinguishes the social and economic model of the Nordic countries? The authors offer the following four pillars: 1.

Significant public investment in family policies, education, and health services;

coordinated wage-setting within and across industries;

substantial expenditure on social insurance to safeguard against income losses due to unemployment, disability, and illness; and

4. high and progressive taxation of labor income, complemented by subsidies for services that support employment. These pillars are fully described in the working paper, and Figure 1 illustrates the development of core policies that address these elements. The range of services provided by these and other programs are meant, in part, to provide equality of opportunity. How successful are these programs? The authors present key facts while also debunking myths and misconceptions about income inequality in the Nordic countries, in comparison with the United Kingdom and United States, which both exhibit high levels of income inequality. They find the following: •

A more equal predistribution of earnings (for example, through minimum wages, access to education, and labor market regulations, among others), rather than income redistribution, mainly explains the lower income inequality in the Nordic countries.

•

Equality in wage rates, not work hours, primarily explains lower inequality in the Nordic countries. This is largely driven by wage compression, meaning there is a small wage gap among employees, regardless of their position and seniority.

•

Conventional wisdom aside, hourly wage compression within gender, not between men and women, is a key component of income equality in the Nordic countries. For example, although the gender gap in hourly wages is about 30 percent lower in the Nordics than in the United States, it explains less than 2 percent of the difference in the dispersion of hourly wages between the Nordic countries and the United States.

What explains these facts? The authors analyze three common hypotheses, giving prominence to the third:

compression explains most of the social equality within Nordic countries.

Governments spend heavily on children and families through subsidized daycare, education and health programs, thus equalizing the distribution of skills and human capital among children of disadvantaged families.

With that mystery solved, another looms. What impact does coordinated wage compression have on productivity and growth? If other countries, including the United States and United Kingdom, adopted wage compression policies, would their economies remain as productive and strong? Here, the evidence is less clear. One view holds that the Nordics benefit because less-equal countries have more incentive to innovate and take chances, which redounds to the Nordics’ benefit. If true, then if all countries followed the Nordic model, innovation would suffer, and economic growth would decline across the globe.

Reality check: Most research evaluating the causal effects of such programs suggests that effects are relatively modest. Indeed, the observed distributions of education and skills are relatively similar between the Nordic countries, the United States and United Kingdom. In contrast, the wage premium for education and skills is twice as large in the United States and United Kingdom. Although the Nordics have relatively high- and progressive-income taxes, which might discourage labor supply, they also subsidize services that are arguably complementary to working, such as daycare and other family friendly policies. These policies may both increase labor force participation and reduce inequality in hourly wages. Reality Check: A growing body of research reveals that the impact of subsidized services is small. For example, subsidized daycare mostly replaces other forms of out-of-home care used by working mothers, resulting in little to no increase in maternal employment or earnings. (The authors are quick to note that this is not an argument against such policies, as their benefits may exceed costs, even with minimal impact on income inequality.) Income equality in the Nordics is primarily attributed to a two-tier collective bargaining structure, starting with sectoral bargaining of wage floors or base wages, followed by local bargaining at the firm level. Importantly, there is strong wage coordination both between and within industries. Reality Check: The authors concur. Both theory and data suggest that this coordination significantly compresses the distribution of wages compared to the distribution of labor productivity, and this wage

READ THE WORKING PAPER NO. 2025-25 · FEBRUARY 2025

Income Equality in the Nordic Countries: Myths, Facts, and Lessons bfi.uchicago.edu/working-papers/income-equality-in-thenordic-countries-myths-facts-and-lessons

In contrast, another view holds that wage compression and social insurance stimulate innovation, productivity, and growth because, for example, they increase the cost of low-skilled labor and lower the price for highskilled workers, thereby affecting the profitability of new technology and driving out inefficient firms. In effect, wage compression and social insurance serve as mechanisms for sharing risk and compensating workers affected by the negative effects of structural change, thus reducing social and political barriers to new technology, international trade, and competition in domestic markets. Research has yet to resolve this puzzle, with most evidence either correlational or circumstantial. To adequately address these and related questions will entail advances in both theory and measurement, including a tighter connection between data and theory. Such a connection is missing in most labor economics, the authors argue, and they offer a critique of current research agendas, many of which focus on micro data to explain a part of the puzzle. Their own research prescription includes the development of models in which skills, labor supply (including international migration), capital investments, wages, and profits are determined simultaneously. Without such models, there is little hope of understanding whether the Nordic model is replicable, or even desirable.

ABOUT OUR SCHOLAR

Magne Mogstad

The Gary S. Becker Distinguished Service Professor, Kenneth C. Griffin Department of Economics

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • MARCH 2025

The Price of Faith: Economic Costs and Religious Adaptation in Sub-Saharan Africa Based on BFI Working Paper No. 2025-33, “The Price of Faith: Economic Costs and Religious Adaptation in Sub-Saharan Africa,” by Eduardo Montero, University of Chicago; Dean Yang, University of Michigan; and Triana Yentzen, University of Michigan

When the opportunity costs of being a Seventh Day Adventist in Sub-Saharan Africa increase, membership growth declines, and existing members report lower satisfaction. Local churches respond by establishing new educational and health institutions, and members reduce adherence to the church’s healthy living tenets. Prior research shows that religious institutions play a major role in shaping economic behaviors, social norms, and development outcomes worldwide. But what happens when religious membership (or belief) entails economic costs? For example, the Seventh Day Adventist (SDA) church prohibits the production of tobacco, coffee, and tea, effectively excluding its members from a major sector of the local agricultural economy in Sub-Saharan Africa. In this paper, the authors examine how opportunity costs of religious membership opportunity costs influence membership in the SDA church, and whether and how local churches adapt to these economic realities. Importantly, the opportunity cost of joining the SDA church due to its production prohibitions varies across different regions—with higher costs in areas that are more suitable for prohibited crops—as well as over time—with higher costs at times when export prices for prohibited crops are higher. Using data on potential crop yields and regional export prices, the authors construct detailed measures of the opportunity costs of SDA membership for sub-national localities across Sub-Saharan

Africa from 1991 to 2022. They then combine these opportunity cost measures with public data on SDA membership and other local statistics, uncovering the following: •

Increases in opportunity costs lead to substantial declines in new church memberships. During periods of non-zero opportunity costs, net membership growth falls by 10.4 percentage points on average, with the decline reaching 19.3 percentage points in periods when opportunity costs are in the top quartile (of the non-zero opportunity cost distribution).

•

In addition, data from SDA member surveys reveal that when the economic costs of membership rise, members report less satisfaction with and less long-run commitment to the church.

•

In terms of local church responses, increases in opportunity costs lead to the establishment of new educational and health institutions. These institutions may be intended to help attract new members, as well as offset opportunity costs for existing members.

opportunity cost: the loss of potential gain from other alternatives when one alternative is chosen

bfi.uchicago.edu

Figure 1 · Opportunity Cost of SDA Crop Prohibitions by SDA Locality Opportunity Cost of SDA Crop Prohibitions by SDA Locality A) 1995

B) 2011

C) 2020

Cost of SDA Prohibitions by SDA Locality $0

$0-1

$1-10

$10-50

$50-700

Note: These maps show the average opportunity cost (in 2010 US dollars per hectare per year) of adhering to SDA prohibitions on production of tobacco, coffee, and tea. Areas colored in darker

Note: These maps show the average (in 2010 US dollars per hectare per year) of adhering to SDA prohibitions on production of tobacco, coffee, and tea. Areas colored in darker shades of blue represent places whereopportunity adherence iscost more costly. shades of blue represent places where adherence is more costly.

•

Local churches also appear to respond to increased opportunity costs by reducing the emphasis placed on the church’s healthyliving prescriptions, which undergirds the prohibition on coffee, tobacco, and tea. When opportunity costs rise, SDA members report hearing fewer messages about the church’s “holistic living” prescriptions from church pastors. In addition, increases in opportunity costs lead to more violations of the church’s prescriptions on healthy living, namely, more consumption of alcohol and tobacco.

Taken together, the results highlight how individuals take opportunity costs of membership into account when deciding whether to join a new religion, while local churches, in return, take measures to balance tradition and adaptation in response to economic conditions. These findings have implications beyond the specific context studied here. First, they demonstrate how economic incentives can drive religious change through multiple channels: directly affecting individual choices about religious participation,

READ THE WORKING PAPER NO. 2025-33 · FEBRUARY 2025

The Price of Faith: Economic Costs and Religious Adaptation in Sub-Saharan Africa bfi.uchicago.edu/working-papers/the-price-of-faith-economiccosts-and-religious-adaptation-in-sub-saharan-africa

and indirectly by inducing institutional adaptation. Second, they reveal religious institutions as dynamic actors that appear to strategically adjust their practices and messaging in response to local conditions, while maintaining their fundamental identity. This adaptability may be particularly important in developing regions where religious prescriptions can significantly impact economic livelihoods. More broadly, this research illuminates mechanisms of institutional and cultural change. While much work emphasizes the persistence of cultural practices, the authors document how religious institutions can facilitate relatively rapid adaptation to economic conditions. This suggests that successful religious movements may act as mediators of cultural change, selectively relaxing certain prescriptions while maintaining their core identity and values. In an era of rapid economic transformation across the developing world, understanding these dynamics of religious and cultural adaptation becomes increasingly important.

ABOUT OUR SCHOLAR

Eduardo Montero

Assistant Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • APRIL 2025

Credit Card Entrepreneurs Based on BFI Working Paper No. 2025-44, “Credit Card Entrepreneurs,” by Ufuk Akcigit, University of Chicago; Raman S. Chhina, University of Chicago; Seyit Cilasun, TED University; Javier Miranda, Halle Institute for Economic Research; and Nicolas Serrano-Velarde, Bocconi University

Credit cards are a critical, yet costly, financing tool for US small businesses. Between 2021 and 2023, usage nearly doubled, interest payments surged, and delinquencies rose. Following the Fed’s 2022 rate hikes, banks most exposed to interest rate risk cut credit card supply, driving a 15.75% drop in balances, a 10% decline in revenue growth, and a 1.5% decline in employment growth among small firms relative to non-exposed banks. Credit cards are one of the most debated sources of small business financing in the United States economy. They have been instrumental in the early stages of some notable entrepreneurial successes, such as Airbnb’s “Visa financing round” and Netflix, whose founder Reed Hastings used credit cards to cover initial expenses and test mailings for the DVD rental service. At the same time, their high interest rates can just as

easily lead to the downfall of many small businesses, turning a potential lifeline into a financial burden. Despite the critical role of credit card financing in sustaining small businesses, its importance and impact on their economic health remain largely unexplored due to a previous lack of comprehensive data. In this paper, the authors address this gap

Figure 1 · Dynamics Dynamics A) Credit Card Balances

B) Credit Card Usage

0.5 Difference in balance for firms with low vs. high income gap providers

0.9 Difference in balance for firms with low vs. high income gap providers

0.3

0.7 Policy Shock

Policy Shock

0.5

0.1 0

0.3

-0.1 0.1 0

-0.3

-0.1 -0.3

-0.5 Jan. 2021

April 2021

July 2021

Oct. 2021

Jan. 2022

April 2022

July 2022

Oct. 2022

Jan. 2023

Jan. 2021

April 2021

July 2021

Oct. 2021

Jan. 2022

April 2022

July 2022

Oct. 2022

Jan. 2023

Note: These figures plot the differential effects of the Federal Reserve’s 2022 rate hikes on small business credit card balances and usage, based on the income gap of their card-issuing banks.

Note: plot themore differential effects of the Federal Reserve’s 2022 rategaps) hikesexperienced on small business credit carddeclines balances and usage, on the income gapshock. of their card-issuing banks. Firms FirmsThese whosefigures providers were exposed to interest rate risk (i.e., with lower income significantly larger in balances andbased usage following the policy whose providers were more exposed to interest rate risk (i.e., with lower income gaps) experienced significantly larger declines in balances and usage following the policy shock.

bfi.uchicago.edu

standard deviation lower income gap gap, they pass on this risk to small businesses. Following the Federal Reserve’s rapid tightening of monetary policy in early 2022, firms borrowing from these more exposed banks saw credit card supply contract, leading to a 15.75% decline in balances, a 10% drop in revenue growth, and a 1.5% decline in employment growth. Higher interest costs also rendered credit card debt unsustainable for many, contributing to roughly half of the observed rise in delinquencies.

by utilizing high-quality and near-real-time data from nearly 1.6 million small businesses using the Intuit QuickBooks online platform, complemented by large-scale surveys of these businesses. They document the following key facts and consequences of credit card financing for small businesses: •

•

Small firms rely heavily on credit card financing. Between 2021 and 2023, monthly credit card repayments were up to three times higher than traditional loan repayments. Credit card interest burden rose by 60% during the post-COVID monetary policy tightening that began in March 2022, leading to elevated levels of credit card delinquencies. Similarly, survey responses reveal that over 55% of businesses report using business credit cards for financing in the past 12 months, far exceeding reliance on alternatives such as credit lines, loans, or internal funding. Entrepreneurs particularly value credit card financing for its accessibility and flexibility in managing cash flow shocks. Younger firms, smaller firms, and those with lower cash reserves consistently allocate a larger share of their payments to credit card financing.

•

Surveys also reveal that credit cards are a key financing source in response to firm-level shocks, such as uncertain cash flows and overdue invoices.

•

Changes in the supply of credit card financing have real effects on firms. When banks that issue credit cards are more financially vulnerable to interest rate hikes, as measured by a one

•

Finally, the authors develop a model of small business behavior to assess how credit card financing affects the transmission of interest rate and loan supply shocks. They find that while credit cards can serve as a short-term buffer by expanding borrowing capacity when other financing is constrained, their high interest costs can strain cash flows over time and slow recovery.

The findings underscore that credit cards are not just a convenience but a critical source of financing for small businesses, especially when other forms of credit dry up. However, this reliance carries real macroeconomic consequences. When monetary policy tightens, and banks vulnerable to interest rate risk pull back, small businesses disproportionately bear the cost through lost revenue, slower hiring, and heightened financial distress. As credit card use continues to rise among small firms, policymakers and lenders should recognize the dual role these products play, as both a flexible lifeline and a transmission channel for systemic shocks.

income gap: measure of a bank’s exposure to interest rate risk, defined by the difference between the share of assets and liabilities that reprice within a given time horizon

READ THE WORKING PAPER NO. 2025-44 · MARCH 2025

Credit Card Entrepreneurs bfi.uchicago.edu/working-papers/credit-card-entrepreneurs

ABOUT OUR SCHOLARS

Ufuk Akcigit

The Arnold C. Harberger Professor in Economics and the College, Kenneth C. Griffin Department of Economics

Raman Singh Chhina

PhD Candidate, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • APRIL 2025

Economic Shocks and Healthcare Capital Investments Based on BFI Working Paper No. 2025, 35, “Economic Shocks and Healthcare Capital Investments,” by Michael R. Richards, Cornell University; Maggie Shi, University of Chicago; and Christopher M. Whaley, Brown University

Recessions restrain IT investments while expansion policy indirectly stimulates them. These effects occur in a symmetrical manner. For example, in the healthcare industry, economic downturns restrain hospitals’ investments while hospitals exposed to a Medicaid expansion purchase more IT.

It is a well-researched and broadly understood Figure 1 · Relative Effect Sizes for Each Shock Compared phenomenon that information technology (IT) has driven to Treated GroupShock Pre-Period Mean to Treated Group Pre-Period Relative Effect Sizes for Each Compared economic growth in recent decades and, especially with 40% Increase Relative to Pre-Period Average the onset of artificial intelligence, it will likely do so into the future. Many companies have directly experienced 30 25 20 IT’s effects on how they conduct business. For example, IT has transformed how managers coordinate, 10 10 3 0 0 0 6 communicate, and guide production to improve output 0 -7 and enhance product quality. And some companies owe -10 -11 -12 -13 -13 -15 their existence to advances in IT. -20

Given the importance of IT, and the competitive pressures that advances bring to the market, you might think -40 that a company would have an unlimited appetite for IT investment. However, capital expenditures (capex) -60 -67 -67 are expensive, and the future is unknown. While large IT Great Recession Effect: Relative Change by 2012 Medicaid Expansion Effect: Relative Change by 2017 investments can positively impact a firm’s growth, a failed -80 investment can have equally large negative effects. Market BackPharmacy Utilization Clinical Supply Info Laboratory Financials Staffing Office Review Service Chain Systems Line uncertainty, it turns out, is key to firms’ decision-making. However, research has offered limited insight into how Note: Leveraging rich data from theand hospital and two substantiveshocks economic shocks in directions Note: Leveraging rich data from the hospital industry twoindustry substantive economic in opposite market fluctuations affect firms’ IT investment(i.e., decisions. opposite directions (i.e., one negative and positive), this figure illustrates authors’managers’ novel one negative and one positive), this figure illustrates theone authors’ novel evidence thatthe hospital IT that hospital IT investment decisions are highly sensitive to fluctuations investment decisions areevidence highly sensitive tomanagers’ fluctuations in market circumstances. Economic downturns restrain

in market circumstances. downturns investment while public insurance while public insurance expansions Economic indirectly promoterestrain it. In other words, hospitals demonstrate consistent This paper addresses that gap by employinginvestment an indirectly promote it. In othermarket words, hospitals and symmetrical actions expansions when facing negative or positive shocks.demonstrate consistent and empirical approach that captures the relationship symmetrical actions when facing negative or positive market shocks. between market events and long-term IT capex budgeting. For example, if firms do respond to changes question helps explain the uneven spread of IT across in market circumstances, it is not clear whether IT and within industries. capex adjustments would simply mirror other shortThe authors focus on health IT, specifically of US run spending decisions or diverge from other business hospitals. The US hospital industry captures over $1 actions. These decisions can be quantified. In other trillion (about 30%) of medical spending annually— words, how market shocks influence firms’ IT adoption the largest share of spending among the provider decisions is an empirical question. Answering this

bfi.uchicago.edu

types. The health IT adoption in decision is ideally suited for the authors’ investigation because hospitals must commit to substantial upfront investment while accepting potentially large adjustment costs and sunk costs tied to health IT procurement. Hospitals also face uncertain payoffs and are regularly exposed to business cycle fluctuations as well as industry-targeted policy and regulatory interventions.

•

The relative effect sizes typically range between 5% to 15% for a given technology purpose; however, some changes are as large as 30% to 67% over baseline levels. These behavior changes are present among for-profit as well as not-for-profit hospitals and are statistically indistinguishable between the two.

•

The impacts on hospitals’ decision-making are not fully explained by pure income shocks shocks. Standalone and financially weaker hospitals are not more responsive to either shock; rather, their responses are more muted at times. And hospitals with relatively greater exposure to the size of the Medicaid expansions do not engage in greater health IT capital investments compared to other hospitals in affected states.

•

Finally, these observed changes in IT investment do not apply to more variable hospital expenditures, like hiring and marketing. This new finding emphasizes that hospitals prioritize the IT capital investment margin differently when facing changes in financial and market circumstances.

The authors study industry-specific microdata that captures nearly all hospitals, including granular information on the timing and type of IT investments made over time, as well as negative shocks (e.g., the Great Recession of 2008-09) and positive shocks (e.g., the 2014 Affordable Care Act Medicaid expansions) to find the following: •

•

Hospitals’ IT investment decisions are sharply influenced by harmful and helpful market shocks in a symmetrical manner; that is, economic downturns restrain hospitals’ investments while hospitals exposed to a Medicaid expansion purchase more health IT. The effects of both shocks are dynamic and become larger with time. Three years after the financial crisis, hospitals operating in the most severely impacted areas decrease their IT capital investments by 10% to 15%. Similarly, hospitals affected by the 2014 Medicaid expansions adopt nearly 10% more technology solutions in comparison to hospitals in non-expansion states by 2017. Hospitals’ adjustments include clinical service-line management, laboratory management, and multiple administrative functions (e.g., back-office management, financials, information systems, and utilization review).

One policy implication of this novel research: Building on work that calculates the implied “savings” for states refusing to expand Medicaid through the ACA, the authors reveal a previously overlooked, indirect consequence of state policymakers’ decisions; namely, they are restraining technological investment within their own hospital industry. Such decisions further challenge ongoing attempts by US healthcare firms to leverage IT advancements to improve performance.

adjustment costs: in economics, these are expenses incurred when economic agents (like firms or individuals) change their decisions or actions, such as adjusting production levels, hiring or firing workers, or investing in new capital sunk costs: an expense or investment that has already been made and cannot be recovered, regardless of future decisions negative shocks: an unexpected event that has a detrimental impact on the economy, potentially leading to decreased output, increased unemployment, and higher prices, for example positive shocks: an unexpected event that leads to a beneficial increase in economic activity, such as increased production, lower prices, or higher employment, often stemming from either increased demand or supply income shocks: one-time, unexpected changes in how much money an individual or firm earns or has

READ THE WORKING PAPER NO. 2025-35 · FEBRUARY 2025

Economic Shocks and Healthcare Capital Investments

ABOUT OUR SCHOLAR

Maggie Shi

Assistant Professor, Harris School of Public Policy

bfi.uchicago.edu/working-papers/economic-shocks-andhealthcare-capital-investments

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • APRIL 2025

The Curious Surge of Productivity in U.S. Restaurants Based on BFI Working Paper No. 2025-39, “The Curious Surge of Productivity in U.S. Restaurants,” by Austan Goolsbee, Federal Reserve Bank of Chicago; Chad Syverson, University of Chicago; Rebecca Goldgof, New York University; and Joe Tatarka, University of Chicago

Real labor productivity at US restaurants surged over 15% during the COVID pandemic. Mobile phone tracking data reveal that this appears driven by take-out customers who spend 10 minutes or less at restaurants. Following a brief dip at the outset of the pandemic, the restaurant industry experienced an unprecedented surge of 15% in real labor productivity. productivity Both sales and visits per employee, measures that had been relatively steady for decades before, rose sharply. In this paper, the authors use micro-level data on mobile phone visits for over 100,000 fast food restaurants across the United States, combined with debit and credit card transaction data, to study the drivers of this surge. They find the following: •

•

Productivity grew most in restaurants where the amount of time that consumers spend when they visit, their “dwell times,” decreased. Decreased dwell times were widespread during the pandemic, with the share of the visits lasting less than ten minutes rising particularly sharply. This pattern is strong enough that nearly the entire productivity increase that the authors observe is attributable to decreased in dwell times. The reduction in dwell times, and the associated productivity gains, appear to stem from increased demand for take-out and delivery. By serving a greater number

Figure 1 · Annualized Sales per Employee

Annualized Sales per Employee 130

120

110

100

70 Jan 1992

Nov Sep 1993 1995

July May March Jan Nov Sep July May March Jan 1997 1999 2001 2003 2004 2006 2008 2010 2012 2014

Nov 2015

Sep July May March Jan 2017 2019 2021 2023 2025

Note: Figure shows an index (1992 = 100) of annualized monthly real sales per employee for the Food Services and Drinking Places industry. seasonally adjusted sales are from monthly the Census real Monthly Retail Trade Survey Note: Figure shows anNominal index (1992 = 100) of annualized sales per employee for the report. Real sales obtained by deflating by CPI series for food away from home. Seasonally adjusted employment Food and Drinking from theServices Bureau of Labor Statistics. Places industry. Nominal seasonally adjusted sales are from the

Census Monthly Retail Trade Survey report. Real sales obtained by deflating by CPI series for food away from home. Seasonally adjusted employment from the Bureau of Labor Statistics.

of quick-turn customers without adding staff, restaurants effectively boosted output per worker, translating into a genuine and measurable rise in productivity. The authors also use their data to rule out three alternative explanations for the productivity surge: •

Falling demand during the pandemic cannot explain the sustained increase in

real labor productivity: the amount of inflation-adjusted output (e.g., sales or visits) produced per worker

bfi.uchicago.edu

Figure 2 · Average Customer Visit Dwell Shares Average Customer VisitTime Dwell Time Shares 70% Share of Visits 0-10 Minutes

60 50 40

Period After Processing Changes

30 20

21-60 Minutes 11-20 Minutes 61-240 Minutes

10 0 Jan 2019

Jan 2020

Jan 2021

Jan 2022

Jan 2023

Note: Figure shows averageNote: share of panel restaurants’ customer visits dwellrestaurants’ time category.customer The shadedvisits area indicates processing SafeGraph’s Monthly Patterns beginning in May 2022. Figure shows average share ofby panel by dwellperiod timeafter category. Thechanges shadedinarea indicates period after processing changes in SafeGraph’s Monthly Patterns beginning in May 2022.

productivity. While there was an initial drop, real consumption in the industry has since rebounded and now exceeds pre-pandemic levels by about 20 percent. •

Economies of scale are not a factor. The average number of employees per restaurant remained flat or declined, indicating that restaurants did not become more efficient simply by getting bigger.

•

Increased market power is also unlikely. While restaurants could have raised prices, the authors account for inflation-adjusted spending and shows that productivity gains were tied instead to behavioral shifts— specifically, shorter customer visits and a rise in takeout—rather than pricing changes.

What makes the restaurant industry’s COVID-era productivity spike especially notable is that it appears to stem not from technological upgrades or industry consolidation, but from a fundamental change in customer behavior. The rise of quickturn visits, take-out, and delivery effectively altered the “technology” of demand, enabling restaurants to serve more customers with the same labor force. This one-time, demanddriven jump in productivity has persisted even as broader conditions returned to normal. The findings raise an important question: if a change in how services are consumed can so dramatically boost productivity in restaurants, might other service-sector industries have experienced similar transformations? Exploring this possibility offers a promising path for future research.

real consumption: the total amount of goods and services consumed, adjusted for inflation to reflect true purchasing power economies of scale: cost advantages that occur when increasing production leads to lower cost per unit

READ THE WORKING PAPER NO. 2025-39 · MARCH 2025

The Curious Surge of Productivity in U.S. Restaurants bfi.uchicago.edu/working-papers/the-curious-surge-ofproductivity-in-u-s-restaurants

ABOUT OUR SCHOLARS

Chad Syverson

George C. Tiao Distinguished Service Professor of Economics, Chicago Booth

Joe Tatarka

Research Professional, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • APRIL 2025

The Social Construction of Race during Reconstruction Based on BFI Working Paper No. 2025-30, “The Social Construction of Race during Reconstruction,” by Anjali Adukia, University of Chicago; Richard Hornbeck, University of Chicago; Daniel Keniston, Louisiana State University; and Benjamin Lualdi, University of Chicago

During the United States’ Reconstruction Era (1865-77), people with the same physical skin tone were more likely racialized as White or Mulatto if they were wealthier or literate. This historical finding underscores the fluidity of socially constructed racial classifications and challenges the notion of fixed racial identities, which continues to influence both academic literature and social structures. The end of the US Civil War in 1865 and the emancipation of enslaved people radically reframed race relations in ways that still resonate today. Indeed, it could be said that the years following the Civil War reconstructed race in the United States, as segregationists—forced to abandon the previous legal distinction of “free” vs. “enslaved”—developed a system of social hierarchy based on skin color. This social construction of race remains deeply entrenched in US institutions and culture, including in social science research. This paper investigates the embryonic phase of this phenomenon, showing how people with the same physical skin tone were racialized differently based on their wealth and other proxies for socioeconomic status in the years following the Civil War. What came to be known as the Reconstruction Era (1865–1877) saw efforts by the federal government to integrate formerly-enslaved people into broader social, economic, and political institutions. However, with Reconstruction’s abandonment in 1877, power was thrown back to the states, giving rise to

Figure 1 · Freedman Bank Depositor Accounts Linked to 1870 Census, by Branch

Freedman Bank Depositor Accounts Linked to 1870 Census, by Branch

New York City Baltimore St. Louis

Louisville

Lexington

Washington, DC Richmond

Lynchburg

Norfolk

Nashville Little Rock

Memphis Columbus

Shreveport

New Bern

Huntsville

Atlanta

Vicksburg Natchez Mobile New Orleans

Augusta

Wilmington

Charleston

Beaufort

Savannah Tallahassee

Number of Depositors 10 50 100 1000

Note: This figure shows the locations of 27 Freedman Bank branches in the authors’ sample, where the size of the Note: This figure shows the locations of 27 Freedman Bank branches in the authors’ sample, circle is proportional to the number of depositors linked to the Census. These branches cover the major Southern the size ofdepositors the circleare is in proportional to the number of depositors linked to the Census. citieswhere at the time. 54% of their branch county and an additional 8% are in a neighboring county in the 1870 Census. When someone opened an account with the Bank, the branch clerk filled out a depositor record, These branches cover the major Southern cities at the time. 54% of depositors are in their which included the depositor’s name, birth state, age, and names of household members along with other information, fieldan for additional “complexion.” branchincluding countyaand 8% are in a neighboring county in the 1870 Census. When

someone opened an account with the Bank, the branch clerk filled out a depositor record, which included the depositor’s name, birth state, age, and names of household members along with other information, including a field for “complexion.”

Reconstruction Era: The Reconstruction era (1865-1877) in the United States followed the Civil War and focused on reintegrating the former Confederate states into the Union while defining the legal status of formerly enslaved people. Reconstruction was abandoned following the disputed election of 1876, which was resolved when the Republican candidate, Rutherford B. Hayes, agreed to end Reconstruction, thus ensuring Congressional approval of his presidency.

bfi.uchicago.edu

state-supported segregation (Jim Jim Crow) Crow that only “ended” with the Civil Rights Act of 1964. To study the construction of post-Civil War race, the authors study the 1870 Census and skin tone data collected by the Freedmen’s Savings Bank (18651874), with a focus on the Census’s distinctions between “Black,” “Mulatto,” and “White.” Census takers who visited households were instructed to take care when counting Mulattos, making sure to include “quadroons, octoroons, and all persons having any perceptible trace of African blood,” on which “important scientific results depend.” Note the use of the term “scientific” in an official government source such as the Census, which provides insight into how people were racialized at this time. Racial perceptions, and the resulting racial categorization, substantially impacted how people were treated by society and the law. In the case of the Freedmen’s Savings Bank Bank, clerks recorded depositor “complexion” and other information to help identify customers when they returned to the bank, including skin tone (e.g., dark, dark brown, brown, light brown, light) and occasionally included other physical descriptors (e.g., distinctive scars or pox marks). The authors’ analysis of the data finds the following evidence for the social construction of race in the post-Civil War United States:

•

Among people with the same skin tone, those with higher wealth or literacy are more likely racialized as Mulatto or White. The authors estimate selected racialization along each of three margins, (White or Mulatto vs. Black; White vs. Mulatto or Black; and Mulatto vs. Black), and find the largest effects among people with light skin tones.

•

The influence of socioeconomic status on racialization is similar across age, gender, and geographical area, suggesting a pervasive social construction of race. The evidence for this finding is revealed in the uniquely detailed data on skin tone from depositor records of the Freedman’s Savings Bank, which reflects the differential racialization by socioeconomic status in the 1870 Census.

The main contribution of this novel paper is not just that race is a social construct (an idea that is increasingly taking hold), but rather to empirically reveal the historical construction of race along socioeconomic lines in the years following the Civil War, which set the stage for racial segregation and continued racial stratification during the Jim Crow Era. Those “scientific” notions linger today. Standard statistical practice still often involves “controlling for race” alongside other variables, which becomes difficult to interpret when race itself is an outcome of socioeconomic status.

Jim Crow: In the wake of Reconstruction’s demise in 1877, state and local governments devised laws to institutionalize racial segregation and discrimination. These laws persisted until federal Civil Rights legislation nearly 100 years later. Civil Rights Act: The Civil Rights Act of 1964 outlawed discrimination based on race, color, religion, sex, or national origin in public places, schools, and employment, and established the Equal Employment Opportunity Commission. Freedmen’s Savings Bank (1865-74): The Freedman’s Saving and Trust Company, known as the Freedman’s Savings Bank, was a private savings bank chartered by the US Congress in 1865 to collect deposits from the newly emancipated communities. The bank opened 37 branches across 17 states and Washington, D.C., within 7 years and collected funds from over 67,000 depositors. The bank failed due to White managers’ mismanagement, fraud, and risky investments that were exposed during financial stress.

READ THE WORKING PAPER NO. 2025-30 · FEBRUARY 2025

The Social Construction of Race during Reconstruction bfi.uchicago.edu/working-papers/the-social-constructionof-race-during-reconstruction

ABOUT OUR SCHOLARS

Anjali Adukia

Assistant Professor, Harris School of Public Policy

Richard Hornbeck

V. Duane Rath Professor of Economics and Neubauer Family Faculty Fellow, Chicago Booth

Benjamin Lualdi

Research Professional, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • APRIL 2025

Why Has Regional Income Convergence in the U.S. Declined? Based on BFI Working Paper 2019-88, “Why Has Regional Income Convergence in the U.S. Declined?” by Peter Ganong, University of Chicago; and Daniel W. Shoag, Harvard University

Income gaps between states have stopped narrowing at the same time that rising housing costs—linked to increased zoning restrictions—have reshaped who can afford to live in high-wage places. Is it worth living in a major labor market, where wages are higher, but so is rent? For much of the 20th century, the answer was yes. Higher wages generally offset the higher cost of living. Today, however, it depends: For high-earning college graduates, the wage premium often justifies the expense, while for low-income workers without a college degree, skyrocketing housing prices wash out the gains. This divergence carries profound implications for our economy, culture, and politics, and in this paper, the authors document these trends and model the forces that may underlie them. Regional Income Convergence Between 1880 and 1980, workers consistently moved from low-income states to high-income

states, a pattern the authors refer to as directed migration. This movement of labor may have migration contributed to regional income convergence by balancing labor supply and human capital between low- and high-wage areas. Over this same period, the authors document that incomes across states converged at an annual rate of 1.8%. Over the past thirty years, both regional income convergence and directed migration have slowed dramatically. From 1990 to 2010, the convergence rate fell to less than half the historical average, and in the years leading up to the Great Recession, it nearly disappeared altogether. At the same time, the steady movement from lowincome to high-income states weakened.

directed migration: as described above, this is the authors’ term to describe the relationship between population growth and income per capita across states. Labor is “directed” to move to regions of relatively higher income per capita, thus increasing income convergence. regional income convergence: regional income convergence, also known as the “catch-up effect,” refers to the economic theory that poorer regions tend to grow faster than richer ones, eventually leading to a narrowing of the income gap

bfi.uchicago.edu

Figure 1Decline · Decline DirectedMigration Migration ofof Directed A) Migration 1940-60

B) Migration 1990-2010 4

Annual Population Growth Rate, 1990-2010

Annual Population Growth Rate, 1940-60

0 0 8

9 Log Income Per Capita, 1940

10.4 Log Income Per Capita, 1990

10.8

Note: These Note: graphsThese plot the relationship between state-level income state-level per capita and growth. Ascapita you canand see,growth. from 1940 1960 states per capita was lower graphs plot the relationship between income per Astoyou can see,where from income 1940 to 1960 states whereexperienced higher growth. income per capita was lower experienced higher growth. This relationship weakened in later years. This relationship weakened in later years.

Divergence by Education The slowdown in convergence has not affected all workers equally. Using census data, the authors show that migration patterns have diverged by education level. In the mid-20th century, workers with and without college degrees tended to move from low-income to high-income states. Today, that pattern holds primarily for workers with degrees. Those without, by contrast, are now more likely to move away from high-income places.

As a result, net migration flows no longer redistribute human capital across regions as they once did. This phenomenon, which the authors term skill sorting, sorting marks a sharp departure from historical trends and may help explain the weakening of regional income convergence. Instead of low- and skill sorting: the pattern in which high-skill workers increasingly move to high-income places while low-skill workers move away from them

Figure 2 · Migration Flows by Education Level, 1995–2000 Migration Flows by Education Level, 1995–2000 B) At Least Bachelor’s Degree 8% Net Migration

Exiting

8% Net Migration

Entering

Exiting

A) Less than Bachelor’s Degree

-4

-4 10.8

11.2 11.4 11.6 Log Nominal Income Low-Income Area High-Income Area

11.8

10.8

11.2 11.4 11.6 Log Nominal Income Low-Income Area High-Income Area

Note: In these plots, the authors separate migration patterns for individuals with less than a bachelor’s degree (left) and with at least a bachelor’s (right).

11.8

Note: In plots, thethose authors separate migrationdegrees patternsare formore individuals lesshigh-income than a bachelor’s degree and with at least a bachelor’s (right). A s you canto see, those without bachelor’s Asthese you can see, without bachelor’s likely with to exit areas, while(left) those with bachelor’s degrees are still migrating such places. degrees are more likely to exit high-income areas, while those with bachelor’s degrees are still migrating to such places.

high-wage workers flocking to the same thriving labor markets, there are now places that attract college-educated workers, and others that retain or receive primarily workers without degrees. Housing Costs and Supply Constraints What explains this shift in migration patterns? The authors point to the rising cost of housing in high-income areas, particularly for low-income workers, as a key part of the story. For much of the 20th century, the returns to moving to a high-income place were similar for workers of all income levels, even after accounting for differences in housing costs. But over the past several decades, those returns have diverged. As housing prices in productive regions have soared, they have absorbed a much larger share of income for low-income workers, effectively erasing the economic benefits of moving. In contrast, highly educated workers still see large net gains from living in those same places.

READ THE WORKING PAPER NO. 2019-88 · JULY 2017

Why Has Regional Income Convergence in the U.S. Declined? bfi.uchicago.edu/working-paper/why-has-regional-incomeconvergence-in-the-u-s-declined

The authors argue that this divergence in returns is linked to changes in housing supply due to the growing role of land use regulations, such as zoning restrictions and permitting limits, which constrain the construction of new housing in high-income areas. They develop a novel panel measure of land use regulation based on the frequency of state court cases mentioning land use. The authors find that housing supply has become significantly less responsive in areas with higher regulatory intensity, leading to higher housing costs, reduced in-migration, and slower income convergence.

ABOUT OUR SCHOLAR

Peter Ganong

Associate Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • APRIL 2025

Evaluating Recent Crackdowns on Disability Benefits: Effects on Income and Health Care Use in Australia Based on BFI Working Paper No. 2025-60, “Evaluating Recent Crackdowns on Disability Benefits: Effects on Income and Health Care Use in Australia,” by Manasi Deshpande, Greg Kaplan, and Tobias Leigh-Wood, University of Chicago

For those removed from disability rolls, family support matters: those living alone experience a drop in income and an increase in antipsychotic medication usage, while those living with family do not. To counter increased enrollment in disability insurance (DI) programs in recent decades, many developed countries, including the Netherlands, Sweden, the UK, Germany, and Australia, implemented reforms to limit access to DI and to remove existing beneficiaries. Similar reforms have been proposed in other countries, including the United States. While research has shed some light on the effects of such reforms on work participation, a full assessment requires analysis of effects on household income and overall well-being. This work addresses that gap by estimating the effects of DI crackdowns on a wide range of recipient outcomes, including household income and health care usage. The authors employ data from the Australian Disability Support Pension (DSP), a means-tested program with no work history requirement, which tightened its medical criteria in 2012 due to rising enrollment. In 2014, these stricter standards were also applied to existing DSP recipients through medical reviews, specifically targeting younger individuals who initially received DSP after January 1, 2008, and who turned 35 after July 1, 2014.

Figure 1 · Responses to Disability Support Pension (DSP) Removal Responses to Disability Support Pension (DSP) Removal DSP Benefits Earnings

All

Non-DSP Government Benefits Family Earnings Household Income

Lives with Family

Lives Alone Parent (Live Apart) Earnings

50,000 Effect of DSP Removal, 2016-2018 Mean (AUD$)

100,000

Note: This figure illustrates how average income responses mask substantial heterogeneity by family structure.

ParentsThis and spouses replace thehow lost DSP income with labor market earnings. In addition to the differences Note: figurefully illustrates average income responses mask substantial heterogeneity driven by having family members versus not, DSP recipients who live with family members (mostly those who live by family structure. Parents inand spouses replace theDSP lostrecipients DSP income labor market with parents) work more themselves response to DSPfully removal. In contrast, who live with alone neither have family support replace the DSP income nor work more themselves to replace the DSP income. Their main earnings. In addition to the differences driven by having family members versus not, DSP channel for income replacement is other government programs. Gray bars indicate 95% confidence intervals around the point estimate. see working for more details. recipients who Please live with familypaper members (mostly those who live with parents) work more themselves in response to DSP removal. In contrast, DSP recipients who live alone neither have family support replace the DSP income nor work more themselves to replace the DSP income. Their main channel for income replacement is other government programs. Gray bars indicate 95% confidence intervals around the point estimate. Please see working paper for more details.

By linking DSP recipients to family members using newly available administrative data, the authors analyze the impacts on household income, which includes the recipients’ own earnings, government benefits, and

bfi.uchicago.edu

those of their family members, as well as health care usage, such as medical visits and prescription drug consumption. They find the following: •

On average, the removal of DSP does not significantly impact household income; in some cases, it even results in a net positive outcome. Recipients manage to recover approximately 55% of their lost benefits by accessing other government programs, and recover about 35% by increasing their work hours, typically in low-skilled positions. Family members also contribute to income recovery, with parents of recipients replacing 35% of lost DSP benefits and spouses replacing 15%.

•

However, family structure is key. For the 43% of individuals living with family, income increases as their family members offset the DSP loss through higher earnings. In contrast, the 57% of individuals living alone do not benefit from family support, do not increase their own earnings, and experience a reduction in household income, relying heavily on other government programs to replace about twothirds of their lost DSP benefits.

•

Additionally, DSP removal is associated with a significant increase in prescriptions for mental health drugs, particularly antipsychotics, especially for men and individuals living alone. Specifically, the likelihood of receiving a mental health

READ THE WORKING PAPER NO. 2025-60 · APRIL 2025

Evaluating Recent Crackdowns on Disability Benefits: Effects on Income and Health Care Use in Australia bfi.uchicago.edu/working-papers/evaluating-recentcrackdowns-on-disability-benefits-effects-on-income-andhealth-care-use-in-australia

prescription rises by 23 percentage points, with a similar increase for antipsychotic prescriptions. The authors suggest three possible mechanisms for the increase in mental health drug prescriptions: an income effect (greater drug use can increase productivity or make work easier), an incentive effect (to the extent that disability benefits discourage treatment, removal could reduce disincentives to seek treatment), and a stress effect (where the stress of losing income prompts increased medication use). The evidence leans toward the stress effect. Bottom line: DI crackdowns may have very different effects depending on what kind of informal safety net recipients have access to. For those with family support, family members may be able to make up the lost income and mitigate the effects on the recipients. But for those without an informal safety net, the effects are potentially quite negative. As for encouraging work among those capable of work, again, those living with family members worked modestly more and mostly in low-skilled occupations, while those living alone did not work more and turned to mental health drugs to cope with the income loss. For policymakers, the message here is that focusing on average effects may lead to a misunderstanding of overall welfare implications, since average effects may mask a wide range of actual effects—from very positive to very negative.

ABOUT OUR SCHOLARS

Manasi Deshpande

Associate Professor, Kenneth C. Griffin Department of Economics

Greg Kaplan

Alvin H. Baum Professor, Kenneth C. Griffin Department of Economics

Tobias Leigh-Wood

BFI Predoctoral Research Professional and Incoming PhD Student

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • MAY 2025

Intuit QuickBooks Small Business Index: A New Employment Series for the US, Canada, and the UK Based on BFI Working Paper No. 2023-84, “Intuit QuickBooks Small Business Index: A New Employment Series for the US, Canada, and the UK,” by Ufuk Akcigit, University of Chicago; Raman Singh Chhina, University of Chicago; Seyit M. Cilasun, TED University; Javier Miranda, Friedrich-Schiller University; Eren Ocakverdi, Independent Researcher; and Nicolas Serrano-Velarde, Bocconi University

In partnership with Intuit, the authors created the Intuit QuickBooks Small Business Index, a new data source covering monthly small business employment and hiring in the United States, Canada, and the U.K. Small businesses play an outsized role in job creation, innovation, and economic growth. According to the most recent data from the US Bureau of Labor Statistics and the US Census Bureau, about 80% of all workplaces in the United States consist of small businesses with fewer than ten employees. Such companies have fewer internal resources, making them more fragile and sensitive to macroeconomic conditions. Despite the importance of small businesses,

real-time data about their performance are hard to produce. In this paper, the authors attempt to fill this gap by providing a new and unique index on the smallest of small businesses in the United States, Canada, and the U.K. The authors collaborated with Intuit, creator of the popular accounting software QuickBooks, to create the Intuit QuickBooks Small Business Index. Each

Figure 1 · Job Creation by Industry in United Job Creation by Industry in States United States 0.3 Net Job Creation

0.2

0.1

Construction

Professional Services

Manufacturing Wholesale

Arts/Entertainment Accomodation and Food

-0.1

-0.2

-0.3

2016

2017

2018

2019

2020

2021

2022

2023

Note: This figure shows job creation patterns byshows industry in the Unitedpatterns States. The area shows out-of-sample projection for the period the second quarter of 2022. As you can see, small Note: This figure job creation bygray industry in thean United States. The gray area shows anafter out-of-sample projection for the period after the second of 2022. As you canthe see, small businesses in the Accommodation businesses in the Accommodation and Food sector experienced the largestquarter employment declines during COVID-19 pandemic, followed by Arts and Entertainment and Wholesale. These sectors also and Food sector experienced largestwere employment declines during theServices COVID-19 followed by Arts andand Manufacturing. experienced the largest recoveries. The sectors in which small the businesses least affected are Professional andpandemic, Finance, followed by Construction Entertainment and Wholesale. These sectors also experienced the largest recoveries. The sectors in which small businesses were least affected are Professional Services and Finance, followed by Construction and Manufacturing.

bfi.uchicago.edu

month, Intuit’s analytics team provides the authors with aggregated and anonymized employment data from their QuickBooks Online Payroll subscribers. The authors use external data to normalize the data so that the Index reflects the broader population of small businesses rather than the QuickBooks customer base. The end product is a dashboard showing monthly stats on employment and employment growth rates for small businesses (for the United States and Canada), and small business job vacancies and growth rates (in the case of the U.K.). The Index offers many insights, which the authors preview in their accompanying paper: •

Small businesses in the Accommodation and Food industry experienced the largest declines in employment during the COVID-19 pandemic, followed by those in the Arts and Entertainment and Wholesale sectors. These industries also experienced the largest recoveries. The least affected sectors were Professional Services and Finance, followed by Construction and Manufacturing.

•

In late 2022, employment decreased in small businesses across all industries, with the smallest declines in Agriculture, Mining, and Utilities, followed by Education and Professional Services. The largest declines were in Information, Transportation, and Leisure.

READ THE WORKING PAPER NO. 2023-84 · JUNE 2023

Intuit QuickBooks Small Business Index: A New Employment Series for the US, Canada, and the U.K.

•

Small business activity experienced sharper declines in the Midwest, New England, and the Great Lakes regions during the COVID-19 pandemic. Least affected were the South, and the Rocky Mountain area. All areas appear to be negatively impacted in late 2022, with the New England region experiencing the smallest declines.

The Intuit QuickBooks Small Business Index provides data about the small business economy in near realtime, making it a valuable resource to both researchers and policymakers alike. The primary purpose of this research release is to provide information on the methodology used to create the Index, and the authors encourage further researchers to utilize the publicly available data for their own work.

ABOUT OUR SCHOLAR

Ufuk Akcigit

Arnold C. Harberger Professor in Economics and the College, Kenneth C. Griffin Department of Economics

bfi.uchicago.edu/working-paper/intuit-quickbooks-smallbusiness-index-a-new-employment-series-for-the-uscanada-and-the-uk

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MAY 2023

Measuring the Characteristics and Employment Dynamics of US Inventors Based on the paper, “Measuring the Characteristics and Employment Dynamics of US Inventors,” by Ufuk Akcigit, University of Chicago; and Nathan Goldschlag, Economic Innovation Group

The authors introduce a new dataset that reveals novel insights about the demographics, employer characteristics, earnings, and employment dynamics of inventors in the United States. Innovation is a key driver of economic growth, and understanding the conditions that lead people to invent new technologies can help reduce inequality between groups as well as help spur growth overall. This paper aims to facilitate such efforts through the introduction of a new dataset that links1 data from the Census Bureau to information on inventors who received patents between 2000 and 2016. The end result is a dataset covering demographic characteristics and employment histories for over 760,000 inventors, from which the authors draw the following insights

concerning inventors’ demographics, employers, earnings, and employment trajectories. Demographics: •

Females are underrepresented among inventors, especially on a citation weighted basis. Females account for less than 12% of inventors, a share that is rising and tends to be higher among young inventors. The share of citations accounted for by female inventors lags behind their share of the inventor population.

•

Over 30% of inventors are foreign born, and China and India account for an increasing share of foreignborn inventors, rising from 25% to 40% between 2000 and 2016.

•

Inventors are getting older. The average age of inventors rose from 43 to 46 between 2000 and 2016. The share of young inventors fell through 2011, but began to rise thereafter.

•

Black Americans are significantly underrepresented among inventors, accounting for less than 2%. Asian inventors are increasingly common, rising from 13% to

1 Available to researchers with approved access via the Federal Statistical Research Data Center (FSRDC) system. Additional information about linkage process available in the paper, “Measuring the Characteristics and Employment Dynamics of US Inventors,” by Ufuk Akcigit and Nathan Goldschlag.

bfi.uchicago.edu

22% between 2000 and 2016. •

Representation of different demographic groups varies significantly by sector. Female inventors have greater representation in the Health Care and Social Assistance and Education sectors, and foreign-born inventors are more common in the Information, Education, and Professional/Scientific Services sectors. The share of young inventors is highest in the Information sector.

Figure • Wage Distribution for Inventors andStar Super Star Inventors Wage1Distribution for Inventors and Super Inventors 100% 98%

•

Inventors tend to work at older, larger firms. Over 68% of inventors work in firms over 20 years old, and almost 63% work in large firms with at least 1000 employees. The share of inventors on patent grants with the largest assignees between 1980 and 2018 rose from about 34% to 47% percent. Inventors, especially the most productive, are much less likely to work at young firms over time. The share of inventors working at firms fewer than five years old fell by almost half between 2000 and 2016, from 15% to under 8%. This share fell the most among super star inventors, defined as those with the most impactful patents. Inventors at young and medium-size firms tend to have the highest impact patents, and inventors at older, smaller firms tend to have the lowest impact patents.

Earnings: •

•

Inventor earnings are highly skewed, especially super star inventors. About 63% of all inventors (and 88% of super star inventors) are among the top 10% of earners. Almost 8% of inventors (and 19% of super star inventors) are in the top 1%. Inventor earnings are closely tied to inventive productivity. Inventors in the top 10% of the inventor earnings distribution tend to receive considerably more citations than inventors in the bottom 10%.

READ THE PAPER MARCH 2025

Measuring the Characteristics and Employment Dynamics of U.S. Inventors link.springer.com/article/10.1007/s10887-025-09251-9

88%

85%

81%

78%

65%

63% 55% 45%

37%

32%

19%

16%

Employers: •

95%

92%

0 60

Population Earnings Percentile Inventors

Super Star Inventors

Note: This figure theof percent ofearnings inventors with earnings more than each percentile of earnings Note: This figure showsshows the percent inventors with more than each percentile of earnings distribution for the population. Earnings percentiles and the share of inventors with earnings less than each percentile distribution for the population. Earnings percentiles and the share of inventors with earnings less than are calculated within each year-quarter from 2000q1 to 2016q3. This figure shows the mean shares across each percentile calculated within year-quarter fromEmployment 2000q1 History. to 2016q3. This figure shows the quarters. For more onare the research please refer to theeach working paper. Source: Inventor mean shares across quarters. For more on the research please refer to the working paper. Source: Inventor Employment History.

Employment: •

Inventors are less likely to switch jobs over time. The hire and separation rates for inventors fell from about 6% and 7% respectively in 2000 to 4% in 2016.

•

Inventors, especially super star inventors, are less likely to start a firm over time. The probability a that super star inventor becomes an entrepreneur fell by 57% between 2000 and 2016.

•

Inventors are increasingly geographically concentrated and less likely to change employment across state lines. The share of inventors working in the 20 largest 4 counties by inventor count rose from 39% to over 47% between 2000 and 2016. The share of inventors switching employment across state lines fell from a peak of 4.6% in 2006 to 2.6% in 2016.

The findings offer a glimpse at the types of insights made possible through the data introduced here. These data will be made available to researchers with approved access via the Federal Statistical Research Data Center (FSRDC) system, drastically expanding the types of analyses possible regarding the role of individuals in the inventive process.

ABOUT OUR SCHOLAR

Ufuk Akcigit

Arnold C. Harberger Professor in Economics, Kenneth C. Griffin Department of Economics

RESEARCH BRIEF • MAY 2025

Non-User Utility and Market Power: The Case of Smartphones Based on BFI Working Paper No. 2025-52, “Non-User Utility and Market Power: The Case of Smartphones,” by Leonardo Bursztyn, University of Chicago; Rafael Jiménez-Durán, Bocconi University; Aaron Leonard, University of Chicago; Filip Milojević, University of Chicago; and Christopher Roth, University of Cologne

Among US college students, most iPhone and Android users would prefer that messages to Android devices no longer appear as green bubbles on iPhones, and iPhone users have a significant willingness to pay to prevent their messages from appearing as green bubbles on other iPhones. Students are also substantially more likely to choose an Android over an iPhone when green bubbles are expected to be removed. These results suggest that firms can enhance their market power by decreasing the utility of competing products’ users, harming both consumers and non-consumers. A traditional way to increase demand for a product Figure 1 · Respondents that Prefer a Software Giving Blue is to make it more useful. But what about making the Respondents Bubbles For Everyone that Prefer a Software Giving Blue Bubbles For Everyone competition less useful? Take smartphones, for example. 100% In addition to adding features that boost the iPhone’s utility, Apple has been accused of increasing the disutility of not owning an iPhone. Specifically, messages 75 exchanged between iPhones and Android devices appear as green bubbles, in contrast to the blue bubbles used for iPhone-to-iPhone communication, prominently distinguishing iPhone users from non-users. Do these 50 tactics harm consumers? Android 78.95%

In this paper, the authors examine this question using the smartphone market as a case study. They begin by surveying college students about green bubble stigma and find the following:

iPhone 66.75%

•

0 Over 90% of respondents believe green bubbles This figure presents the results by phone ownership for the fraction of people that prefer a software update stigmatize Android users, often associating them Note: Note: This figure the to results by phone ownership the95% fraction of people that prefer making all messages appear as presents blue bubbles everyone. The vertical barsfor show confidence intervals. a software update making all messages appear as blue bubbles to everyone. The vertical bars with lower social status and attractiveness.

•

A large majority of Android users (79%) said they would want a hypothetical software change that removes the green bubble distinction, making all messages appear as blue bubbles on iPhones.

•

Strikingly, most iPhone users (66%) also favor such a change.

•

Most respondents believe removing green bubbles would improve the perceived quality of Androids while leaving perceptions of iPhones largely unchanged.

show 95% confidence intervals.

Building on these results, the authors set out to quantify the welfare effects of the green bubble feature through an incentivized deactivation experiment designed to isolate demand for this specific product attribute. The experiment reveals the following: •

US college iPhone users, on average, require a payment of $49 to have their messages appear as green instead of blue bubbles on other iPhones for four weeks.

bfi.uchicago.edu

•

The authors benchmark the effect size against Figure 2 · Respondents Choosing the Google Pixel 9 other iPhone features and find that the median Respondents Option over the Option Choosing theiPhone Google 16 Pixel 9 Option over the iPhone 16 Option valuation of blue bubbles corresponds to 50% 30% and 26% of the median valuation of iMessage and the phone camera, respectively. These estimates show that respondents place a high value on avoiding green bubbles, highlighting the economic significance of the welfare cost. 20

Next, the authors assess how green bubbles influence the relative demand for iPhones versus Androids through an additional incentivized experiment. Respondents are asked to choose between iPhones and Androids in two contingent scenarios: one where green bubbles are banned, and one where they remain. The experiment reveals the following:

10 Green Bubbles Remain 15.8%

Blue Bubbles Mandated 23.1%

0 Respondents are 7.3 percentage points more likely Note: This figure shows the percentage of respondents who chose the Google Pixel 9 (plus a $150 incentive to offset to choose the Android option when green bubbles Note: This figure shows 16, thecontingent percentage respondents who chose thebubble Google Pixel 9 (plus a $150 the price difference) over the iPhone onofwhether the iPhone's green messaging status remains incentive to offset the price difference) over the iPhone 16, contingent on whether the iPhone’s unchanged. Error bars represent 95% confidence intervals. are expected to be removed. This is a sizable green bubble messaging status remains unchanged. Error bars represent 95% confidence intervals. effect, representing a 46% increase in the share of respondents choosing Android from a baseline of These findings reveal that firms can shape consumer 15.8% when green bubbles remain. demand not only by enhancing user utility but also by diminishing non-user utility. This practice raises critical Finally, the authors present a series of case studies competition policy concerns, as it suggests that firms on product features that diminish non-user utility. may deliberately erode non-user utility to strengthen They show that this phenomenon is widespread their dominance, harming both consumers and nonacross industries and illustrate how companies can consumers. Crucially, when there is no outside option strategically strengthen their market power: without the product, it is not possible to distinguish— • Dating apps employ features such as notifications based on choice and price data alone—cases where about missed connections, nudging users to stay user utility increases from cases where non-user utility active to avoid losing potential matches. falls. In light of the growing importance of digital products, developing antitrust measures against the Social media platforms like Instagram foster a • strategic erosion of non-user utility will become an fear of missing out (FOMO) through features like increasingly important task for regulators. ephemeral content (e.g., stories that disappear after

•

24 hours) and push notifications (e.g., “Your friend just posted for the first time in a while”).

READ THE WORKING PAPER NO. 2025-52 · APRIL 2025

Non-User Utility and Market Power: The Case of Smartphones bfi.uchicago.edu/working-papers/non-user-utility-andmarket-power-the-case-of-smartphones

ABOUT OUR SCHOLARS

Leonardo Bursztyn

The Saieh Family Professor of Economics, Kenneth C. Griffin Department of Economics

Aaron Leonard

PhD Student, Kenneth C. Griffin Department of Economics

Filip Milojević

Research Professional, Normal Lab, University of Chicago

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • MAY 2025

Separation of Church and State Curricula? Examining Public and Religious Private School Textbooks Based on BFI Working Paper No. 2025-63, “Separation of Church and State Curricula? Examining Public and Religious Private School Textbooks,” by Anjali Adukia and Emileigh Harrison, UChicago Harris School of Public Policy

Public school textbooks from Texas and California contain similar content, while religious school textbooks have less female representation, feature lighter-skinned individuals, and portray topics like evolution and religion differently. All collections rarely include LGBTQIA+ discussion, portray females in more positive but less active or powerful contexts than males, and depict the US founding era and slavery in similar contexts. Education plays a vital role in shaping our collective memory and understanding of the world. Recognizing this influence, political and religious actors have long sought to shape school curricula to reflect their values and ideologies. In the United States, school choice programs such as charter schools and voucher initiatives have expanded parents’ View the Interactive Research Brief Online

ability to select educational environments that align with their values or religious beliefs. Despite the growing prevalence of these programs and the public investment they attract, we lack comprehensive evidence on how curricular content differs across schools. Motivated by this gap, in this paper, the authors examine popular textbooks used across a variety of educational settings in the United States. The authors examine state-adopted public school textbooks in California and Texas, two states that serve the largest student populations in the in the United States and often represent polar political landscapes, as well as curricular materials often adopted in religious private schools and home school contexts. Using computational social science

bfi.uchicago.edu

methods, including artificial intelligence tools, they analyze the content of these textbooks to assess what topics are covered, which individuals and groups are highlighted, and how those groups and values are portrayed. Their analysis focuses on textbooks in the foundational subjects of Reading, Science, and Social Studies, specifically those intended for third and fifth grade students. They find the following: • There are meaningful parallels between the California and Texas public school collections, while the religious school textbooks differ notably. Compared to public school textbooks, religious school textbooks have less female representation, feature lighter-skinned individuals, and portray topics like evolution and religion differently.

more than other religions, rarely discusses historical or famous individuals who are Asian, Indigenous, or Latine, who were born outside of the North America and Europe, or who identify as lesbian, gay, or bisexual, and depicts the US founding era and slavery in similar contexts. What we teach our children matters. Amid an increase in educational gag orders and statewide changes in the content standards of public school curriculum, this work shines a light on what messages children are receiving both from formal curricula inside the traditional public school system and from informal and non-traditional curricula outside the public school system, messages which could have long-term implications for children’s beliefs and decision-making.

• The authors also uncover important similarities between religious school and public school textbooks. For example, each collection portrays females in contexts that are more positive but less active and powerful than males, discusses Christianity

READ THE WORKING PAPER NO. 2025-63 · MAY 2025

Separation of Church and State Curricula? Examining Public and Religious Private School Textbooks bfi.uchicago.edu/working-papers/separation-of-churchand-state-curricula-examining-public-and-religious-privateschool-textbooks

ABOUT OUR SCHOLARS

Anjali Adukia

Assistant Professor, Harris School of Public Policy

Emileigh Harrison

Postdoctoral Scholar, Inclusive Economy Lab at the Harris School of Public Policy

RESEARCH BRIEF • MAY 2025

The Effect of Medicaid on Crime: Evidence from the Oregon Health Insurance Experiment Based on BFI Working Paper No. 2024-158, “The Effect of Medicaid on Crime: Evidence from the Oregon Health Insurance Experiment,” by Katherine Baicker, University of Chicago; Amy Finkelstein, MIT; and Sarah Miller, University of Michigan

Random access to Medicaid health insurance coverage does not reduce the likelihood of criminal charges or convictions. Those involved with the criminal justice system have disproportionately high rates of mental illness and substance-use disorders, prompting speculation that health insurance, by improving treatment of these conditions, could reduce crime. This study draws on the 2008 Oregon Health Insurance Experiment to provide new evidence on the impact of Medicaid coverage on criminal justice involvement. In 2008, the state of Oregon used a random lottery to allocate 10,000 available enrollment spots in one of its Medicaid programs. Lowincome adults selected by the lottery were 25 percentage points more likely to enroll in Medicaid than those who signed up but were not selected. Researchers linked all study participants to individual-level administrative records from the Oregon Judicial Information Network, which includes data on criminal cases, charges, and convictions from 2007 to 2010, providing, to their knowledge, the first

experimental evidence on the relationship between Medicaid and crime. They find the following: • Medicaid coverage has no statistically significant impact on criminal charges or convictions. • These null effects persist even among high-risk groups, such as individuals with prior criminal records or a history of mental health conditions. This contrasts with earlier quasi-experimental studies, many of which focused on narrower highrisk populations and found reductions in crime. One interpretation is that crimereducing effects of Medicaid for high-risk subgroups may not generalize to the broader Medicaid-eligible population. These results suggest that the impact of Medicaid coverage on criminal justice involvement may be more limited than

bfi.uchicago.edu

previously believed. Importantly, criminal charges and convictions are just one metric for evaluating the benefits of health insurance. Medicaid eligibility and coverage have been linked to a range of positive outcomes, including better access to and increased use of medical care, improved physical and mental health, and enhanced economic security. Nonetheless, for policymakers seeking to reduce criminal behavior specifically, this study suggests that expanding health insurance coverage may not be an effective strategy.

READ THE WORKING PAPER NO. 2024-158 · DECEMBER 2024

The Effect of Medicaid on Crime: Evidence from the Oregon Health Insurance Experiment bfi.uchicago.edu/working-papers/the-effect-of-medicaidon-crime-evidence-from-the-oregon-health-insuranceexperiment

ABOUT OUR SCHOLAR

Katherine Baicker

Provost, University of Chicago and Emmett Dedmon Professor, Harris School of Public Policy

RESEARCH BRIEF • MAY 2025

Toward an Understanding of Discrimination When Multiple Channels Exist Based on BFI Working Paper No. 2025-18, “Toward an Understanding of Discrimination When Multiple Channels Exist,” by Majid Ahmadi, Georgia Institute of Technology; Gwen-Jirō Clochard, University of Osaka; Jeff Lachman, University of Chicago; and John A. List, University of Chicago

This work offers insights into the influence of racial discrimination on drafting (hiring) professional baseball players; importantly, the authors discuss at length how their employed methodology can be generalized to other markets. Gary Becker’s 1957 book-length monograph, The Economics of Discrimination, was one of the first economic treatments of discrimination in the marketplace, giving impetus for a new field of economic research. Among its many contributions, Becker’s work provided a theoretical framework to quantify non-pecuniary motives in discriminatory behavior within labor markets. Until then, such motives were the provenance of sociology, psychology, and anthropology. What could economics possibly have to say about issues not relating to money? Plenty, as Becker revealed not only in his book but throughout his influential career, as he explored the many applications of economic analysis to human behavior. On discrimination in hiring, Becker identified three potential biases: managerial, coworker, and customer. Economists have since investigated these biases, both theoretically and empirically, to determine their influence on hiring. However, five decades of measuring and identifying discriminatory hiring patterns have not resolved all quantitative challenges. For example, limited data mean that researchers must rely on aggregated measures that hide effects related to the biases of managers, employees, and customers. As Becker described, to thoroughly parse the role of

Figure 1 · Index of Fan Bias and Share of African-American Players Drafted Index of Fan Bias and Share of African-American Players Drafted 50% Share of African American Players Drafted TEX 40 MIA SEA

ATL CIN

LAD LAA MIN

AZ WSH

PIT

TOR

CWS

MIL SD CLE

OAK TB

PHI

BOS KC

NYY CHC

COL NYM

-1

BAL

DET HOU

Index of Fans’ Bias Note: This figure reveals a clear relationship between fan bias and the racial composition of drafted players. There is

a negative correlation between the team bias index and thebetween proportion of African players drafted. The Note: This figure reveals a clear relationship fan biasAmerican and the racial composition of X-axis corresponds to the composite index of fan bias computed with reactions to posts related to the Black Lives drafted players. There a negative correlation between team bias index and period the proportion Matter movement. The Y-axis is corresponds to the share of African American the players drafted during the study (2008-20). The gray line is the regression line. β = −0.036, p = 0.03. See working paper for more details. of African American players drafted. The X-axis corresponds to the composite index of fan bias computed with reactions to posts related to the Black Lives Matter movement. The Y-axis corresponds to the share of African American players drafted during the study period (200820). The red line is the regression line. β = −0.036, p = 0.03. See working paper for more details.

discrimination in hiring, we need to consider those distinct discriminatory preferences. This paper addresses that gap by introducing a theoretical framework that shows that to disentangle managerial and customer biases, one’s data must possess certain features: 1.

an objective metric for assessing worker quality,

bfi.uchicago.edu

2. 3.

detailed information regarding the composition of hiring decision-makers, and

In defense of external validity To test their theoretical insights (models) against reality (empirics), economists and other social scientists employ data. These data are often limiting in their explanatory value as they necessarily represent a fraction of a possible dataset. In the present case, for example, to include every piece of data regarding every hiring decision in an economy over time would be impossible, not to mention unwieldy. The trick, rather, is to gather data from a particular subset that delivers strong findings from which we can extrapolate to the rest of the world.

a measure of customer bias magnitude.

It turns out that an ideal setting for this investigation is Major League Baseball’s (MLB) annual draft, which is how MLB assigns amateur players (whether high school, college, or amateur baseball clubs) to its professional teams. This setting adheres to an external validity litmus test for data developed by one of this paper’s co-authors, UChicago’s John A. List (see “In defense of external validity”). Based on List’s rationale from his 2020 paper, the authors conclude that it is nearly impossible to find a more appropriate setting that allows for rigorous testing of the above framework.

This is more than an academic exercise, as the point of most social science is to impact policymaking. And if we are going to affect policymaking—that is, directly impact people’s lives—we want to ensure that our data are valid beyond our subset; that is, our data should be externally valid. For some, such validity is a humbug. For these skeptics of empirical economics, even the slightest doubt about validity renders a study moot.

The authors examine drafting (or hiring) decisions made by MLB teams from 2008 to 2019, when about 12,000 players were drafted, including scouting evaluations and detailed information about each scout, including racial background. (Baseball scouts evaluate players for MLB teams, including on location during games and at training facilities; think of them as a traveling HR department.) Publicly available data on thousands of players—drafted and undrafted—allow the authors to construct the first key metric necessary to distinguish sources of discrimination: an objective measure of player quality. In other words, all players deemed high quality should be drafted, regardless of race or other discriminatory factors.

In “Non Est Disputandum De Generalizability? A Glimpse into the External Validity Trial,” a satirical (and, rare to say for an economics paper, entertaining) defense of external validity, UChicago’s John A. List argues that it is possible to pass an external validity test. Indeed, unique empirical settings—in our case here, MLB hiring practices—are not always a distraction from reality; rather, when that uniqueness allows for relevant testing that no other setting can achieve, then a level of “perfection” is possible whereby we can confidently generalize (and scale) to the rest of the world. Speaking of scaling, this little article only begins to describe List’s longer argument for the validity of empirical research, and the reader is encouraged to visit the full paper via the link above. That said, in sum, here are List’s four tenets of empiricism necessary to address external validity:

To fulfill the second dataset described above—the racial composition of hiring managers—the authors also collect comprehensive data on all MLB scouting directors, including their racial backgrounds. These data allow the authors to assess whether scouting directors exhibit a propensity to recruit players of their own race. Finally, to measure customer (fan) bias, the authors study a naturally occurring event, the Black Lives Matter (BLM) movement in June and July 2020, during which all MLB teams posted messages on social media. The authors then perform a textual analysis of responses to such postings to create an index of fan bias. Further, the authors analyze stadium attendance data from 2008-2019, examining its correlation with the racial composition of the team. Thus, armed with data addressing worker quality, manager discrimination, and customer bias, the authors apply these empirical insights to their models to find the following: •

There is no significant association between race and the likelihood of a player being drafted. When controlling for prospect quality, African American

Theory and empiricism are symbiotic: theory provides a structure for thinking about the world, empirical work tests whether that structure is approximately correct and informs future theories.

One swallow does not make a summer: each study moves priors by an amount corresponding to its quality and the strength of priors.

To explain differences in observed choices across settings, ask if preferences, constraints, or beliefs have changed.

Uniqueness of a setting can be a key strength, not a weakness, if it isolates a particular channel or causal mechanism effectively.

players exhibit a slightly higher draft probability compared to their White counterparts. •

Player compensation is generally consistent across racial groups. However, controlling for player quality, there is some evidence suggesting that

Figure 2 · Attendance Bias and Share of African-American Players Drafted Attendance Bias and Share of African-American Players Drafted

to be drafted later (during rounds 26-40). This suggests that when the stakes are lower and public scrutiny is reduced, scouting directors are more inclined to express their personal preferences, which is supported by the low probability that these players will reach the major leagues.1

50% Share of African American Players Drafted TEX 40 MIA SEA

30 MIL

CIN

WSH

BOS

CHC

-3

-2

LAD LAA

CWS

•

ATL

TOR

STL

NYY

PIT

CLE

MIN

OAK

PHI DET COL BAL NYM -1

0 Index of Fans’ Bias

HOU

Note: This figure shows that fan attendance bias is strongly correlated with the share of African American players

Note: figure shows that fan attendance bias is corresponds strongly correlated the share of drafted.This The X-axis corresponds to fan attendance bias. The Y-axis to the share ofwith African American players drafted during the study period (2008-20). The graycorresponds line is the regression line. βattendance = −0.033, p = 0.04. African American players drafted. The X-axis to fan bias. The Y-axis corresponds to the share of African American players drafted during the study period (200820). The gray line is the regression line. β = −0.033, p = 0.04.

Asian and White players receive lower signing bonuses compared to their peers. •

•

That said, patterns of discrimination loom deep within the data. There is a strong correlation between the drafting of African American players and customer bias during the early rounds of the draft: fan bias is associated with whether a player of a certain race is drafted early. This fan bias correlation, however, is reduced in the later rounds. Collectively, these findings suggest that MLB clubs are likely considering customer preferences when selecting players who will attract significant scrutiny and public attention. Conditional on player quality, scouting directors demonstrate a bias toward players of their own race, with these players 38 percent more likely

Finally, and related to the above finding, these revealed biased preferences carry economic costs: Teams draft lower-valued players when fan bias increases. While such customer bias bears significant opportunity costs (measured as reduced number of wins per season), the financial impact of managerial bias is limited, though, as these players are long shots to reach the majors.

Bottom line: The authors’ novel theoretical and empirical combination provides a framework for analysis of discrimination in economic settings where multiple sources of bias interact simultaneously, including biases hiding within aggregate measures. Likewise, and importantly, the authors’ results plausibly generalize to other markets; that is, this work adheres to List’s external validity test. For scholars, this means caution when examining data for discrimination using establishment level data, as they run a risk when mining findings from aggregate data. When there is tension in biased preferences between management and customers, key aggregates can underestimate, or mask, key biases. For policymakers, understanding the exact channels of bias is key to developing effective and scalable interventions, and this work offers a framework for modeling and estimating relevant sources of discrimination.

1 See Majid Ahmadi, Nathan Durst, Jeff Lachman, John A. List, Mason List, Noah List, and Atom T. Vayalinkal (2023) “Nothing Propinks Like Propinquity: Using Machine Learning to Estimate the Effects of Spatial Proximity in the Major League Baseball Draft,” for a discussion of drafting decisions based on a player’s proximity to a scout’s home.

READ THE WORKING PAPER NO. 2025-18 · JANUARY 2025

Toward an Understanding of Discrimination When Multiple Channels Exist bfi.uchicago.edu/working-papers/toward-an-understandingof-discrimination-when-multiple-channels-exist

ABOUT OUR SCHOLAR

John A. List

Distinguished Service Professor of Economics, Kenneth C. Griffin Department of Economics

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

Authoritarian Propaganda and Social Networks Based on BFI Working Papers 2025-72, “Authoritarian Propaganda and Social Networks,” by Konstantin Sonin, University of Chicago

Authoritarian propaganda is most effective in either highly atomized or fully connected ones, and least effective at intermediate levels of connectivity.

Sonin builds a model to show how propaganda spreads through a society, depending on how people are connected and share information. Citizens can choose to acquire information either directly (by subscribing to media) or indirectly (by receiving messages from others in their network). The dictator chooses how biased the propaganda should be, constrained by whether citizens are still willing to engage with the message. The model reveals the following: •

Network connectivity determines how easily messages spread. When society is atomized (no

Figure 1 · Persuasion on the Simplest Possible Network Persuasion on the Simplest Possible Network Persuasiveness of Propaganda

Propaganda has long been a central tool of authoritarian control, used to shape beliefs, suppress dissent, and legitimize power. Yet the effectiveness of propaganda depends not just on the message itself, but on how information flows through society. In this paper, Konstantin Sonin presents a theoretical model that illuminates the complex interplay between media bias, citizen engagement, and the structure of social networks.

In moderately connected societies, propaganda is least effective

Atomized

Fully Connected

Note:This This figure figure portrays ofof propaganda in ainsimple network. It shows Note: portraysthe thepersuasiveness persuasiveness propaganda a simple network. It shows that that propaganda is most effective when society is either completely disconnected or fully propaganda is most effective when society is either completely disconnected or fully connected. connected. In between, the regime must dilute the message to maintain engagement, Inwhich between, regime must dilute the message to maintain engagement, which limits its impact. limits the its impact.

links) or fully connected, the regime can slant its message heavily and still achieve wide reach. •

At intermediate levels of connectivity, the regime faces a trade-off: the message must be more informative (less biased) to keep citizens engaged, limiting its ability to manipulate beliefs.

bfi.uchicago.edu

•

If the cost of subscribing is high and the network is sparse, few citizens hear the message at all, and propaganda becomes especially ineffective.

•

The structure of the network (not just the number of links) also matters. The same number of connections can produce different levels of persuasion depending on how they are arranged.

•

Counterintuitively, targeting peripheral citizens rather than central influencers can increase overall reach, because central figures can crowd out others’ incentive to subscribe.

•

If a regime can coordinate subscription behavior (e.g., through social pressure or design), it can overcome individual resistance and spread a more biased message more effectively.

•

In some cases, a regime may prefer to suppress social ties entirely, recreating an atomized society that is most susceptible to slanted messaging.

This research highlights how network features, such as connectivity, structure, and access costs, shape the trade-offs regimes face in designing persuasive propaganda. While the analysis is theoretical, it offers a framework for thinking about why certain regimes may tolerate or suppress social ties, or adjust the slant of messaging over time. These insights can inform a better understanding of how information spreads in different political contexts and the subtle ways in which regimes manage belief formation.

READ THE WORKING PAPERS NO. 2025-72 · MAY 2025

Authoritarian Propaganda and Social Networks

ABOUT OUR SCHOLAR

Konstantin Sonin

John Dewey Distinguished Service Professor

bfi.uchicago.edu/working-papers/authoritarianpropaganda-and-social-networks

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

Firm Premia and Match Effects in Pay vs. Amenities Based on BFI Working Paper No. 2025-75, “Firm Premia and Match Effects in Pay vs. Amenities,” by Anders Humlum, Chicago Booth; Mette Rasmussen, University of Copenhagen; and Evan K. Rose, UChicago

Non-wage job amenities vary significantly across firms and play a major role in shaping workers’ overall job utility. High-paying firms often offer worse amenities, and over half of the wage advantage is offset by these tradeoffs, driven largely by worker-firm match effects. In modern labor markets, wages alone fail to capture the full value workers derive from their jobs. Non-wage job attributes, ranging from flexibility and job security to stress levels and workplace culture, play a crucial role in determining overall job quality and utility. Measuring these intangible “amenities” has long posed a challenge for economists. In this paper, the authors address this challenge. The authors administer a survey to over 20,000 workers in Denmark who recently changed jobs. Respondents report their reservation wage wage, the minimum pay they would require to return to their previous job, which serves as a measure of the total non-wage utility derived from their current position over their previous one. The authors combine these self-reported valuations with administrative labor market data to disentangle firm-specific amenities from matchspecific ones (i.e., those arising from unique worker-firm pairings). They find the following: •

Non-pay job amenities vary widely across firms and sectors, with higher amenity values concentrated in public and education sectors and associated with flexibility, security, benefits, and social impact. These values

Figure 1 · Average Firm Amenity Effects by Industry

Average Firm Amenity Effects by Industry Education Other Services Public Administration Real Estate Health and Social Services Arts, Entertainment and Recreation Finance and Insurance Information and Communication Electricity, Gas and Utilities Mining and Quarrying Manufacturing Prof., Scientific and Tech. Services Construction Waste and Water Management Wholesale and Retail Trade Transportation and Storage Administrative and Support Services Accommodation and Food Services -4%

-2%

Note: This figure displays average firm amenity effects—the non-wage value of working at a firm—as a Note: This figure displays average firm amenity effects—the non-wage value of working at a percentage of pay, grouped by industry.

firm—as a percentage of pay, grouped by industry.

correlate with indicators like firm size and PageRank (an algorithm used by Google Search to rank web pages), validating the authors’ empirical approach. •

Higher-paying firms tend to offer worse overall non-wage amenities despite providing better perks and flexibility. On average, a 10% wage increase is accompanied by a 5% decline in amenity value, with only 0.7% of

reservation wage: the lowest wage rate at which a worker would be willing to accept a particular type of job

bfi.uchicago.edu

that attributable to firm-wide characteristics Figure 2 · Predictors of Firm Amenity Effects Predictors of Firm Amenity Effects and the remainder driven by match effects Commute Length between workers and firms. Notably, Weekly Hours Work from Home accounting for amenity effects meaningfully Control of Hours Independence attenuates the gender wage gap in pay premia. Work Pace •

Embedding the empirical findings in a structural job search model shows that amenities significantly offset the advantages of high wages. Over half of the wage advantage is neutralized by worse amenities, and variation in worker wages overstates variation in their job quality by about 50%.

Many aspects of a job beyond pay matter deeply to workers. These results underscore the importance of incorporating non-wage amenities into assessments of job quality and labor market inequality, and imply the need for richer models of labor market sorting and compensation.

Interesting Tasks Physicality Number of Reports Team Work Social Impact Pension Contributions How Family Friendly Quality of Perks Layoff Risk Negotiation Opportunities Continuing Education Quality of Work Environment Support from Colleagues Support from Boss How Respected How Stressed

Bivariate Multivariate

-4

-2 0 2 Effect of Change in Amenity

Note: This figure different firm traitsfirm relate to non-wage job amenities. dots show Note: Thisshows figurehow shows how different traits relate to non-wage jobBlue amenities. Bluethe dots raw correlation (bivariate), while green dots adjust for other firm traits to isolate the independent correlationThe (bivariate), green dots adjust forintervals. other firm traits to isolate the effect ofshow eachthe oneraw (multivariate). whiskerswhile represent 95% confidence

independent effect of each one (multivariate). The whiskers represent 95% confidence intervals.

match effects: the unique, job-specific value that arises from the interaction between a particular worker and a particular firm, beyond what can be attributed to the characteristics of either party alone

READ THE WORKING PAPER NO. 2025-75 · JUNE 2025

Supply Chain Shocks and Firm Productivity: The Role of Reporting Quality bfi.uchicago.edu/working-papers/firm-premia-and-matcheffects-in-pay-vs-amenities

ABOUT OUR SCHOLARS

Anders Humlum

Assistant Professor of Economics and Fujimori/Mou Faculty Scholar, Chicago Booth

Evan Rose

Associate Professor, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

Meaning at Work Based on BFI Working Paper No. 2025-67, “Meaning at Work,” by Nava Ashraf, London School of Economics and Political Science; Oriana Bandiera, London School of Economics and Political Science; Virginia Minni, University of Chicago; and Luigi Zingales, University of Chicago

Randomly implementing an intervention that helps white collar employees at a multinational consumer goods firm find meaning at work leads low-performing employees to exit the firm, and remaining employees to improve their performance. As a result, compensation increases. Many workers today struggle to find meaning in their work. Despite attempts to compensate for this alienation through monetary rewards that connect workers to the firm’s profits, or, more recently, through nonmonetary incentives that aim to connect workers to the firm’s broader purpose, many workers remain unconvinced and disconnected. As the organization of work in large corporations separates workers from the product of their labor, the workplace becomes

a site for producing monetary value rather than personal values, happiness, or fulfillment. In this paper, the authors evaluate an alternative solution. They implement an intervention called “Discover Your Purpose” (DYP), which helps participating workers reflect on their life purpose, and if and how their jobs can aid in achieving that purpose. In a randomized control trial, trial 2,976 white-collar employees participate in a series of readings, essay writing, and a workshop through

randomized control trial: a study design where participants are randomly assigned to either a treatment group or a control group to measure the causal effects of an intervention

Figure 1 · Worker Exit Worker Exit B) Participated in DYP

0.12

0.1

0.08

Worker Exit

A) Received Invitation

0.06 0.04

0.02

-0.02

Months Since Invite

-0.02

Months Since Invite

Note:treatment These plots show treatment effectsemployee on cumulative exits the monthwere invitations werecontinuing sent and continuing for 16 months. leftpresents panel Note: These plots show effects on cumulative exitsemployee beginning thebeginning month invitations sent and for 16 months. The leftThe panel intent-to-treat (ITT) estimates, (ITT) estimates, reflecting the effect of being invited to participate. The right panel displays local average treatment effects (LATE), reflecting the effect presents of beingintent-to-treat invited to participate. The right panel displays local average treatment effects (LATE), capturing the effect on employees who actually participated in the intervention. capturing the effect on employees who actually participated in the intervention.

bfi.uchicago.edu

DYP. The authors measure how the intervention impacts their productivity, performance, and salaries, and find the following: •

•

Workers in the treatment group are 21% more likely to leave their jobs compared to those in the control group. In particular, the treatment inspires workers whose performance is below standards to quit and move to a job with higher meaning. Performance increases following the intervention because low performers either leave the firm or improve in their current jobs. About half of the decrease in below-standard performances is due to employees exiting the firm; the other half is due to improved performances.

•

This improvement is reflected in an increase in workers’ overall compensation. Both the mean bonus earned by employees and the percentage of employees earning performance bonuses increase.

•

The treatment reduces the differences between the job priorities commonly stated by men compared to women, suggesting that the intervention effectively alters traditional gender-based priorities within the workplace. In addition, men in the treatment group are more likely to take parental leave.

•

The firm benefits. If the improved performance lasts for two years, in line with the empirical findings, a cost-benefit analysis yields at least a 72% internal rate of return (IRR) (IRR).

Figure 2 · Gender Gap in Job Priorities

Gender Gap in Job Priorities Flexible Time High Income Work Life Balance Opportunities for Advancement Job Security Useful to Society Personal Contact with People Independent Work

Gender Gap, Control Group

Helping Others

Gender Gap, Treatment Group

Growing, Learning New Skills Interesting Job High Prestige -1

-0.5

0 Gap in Average Ranking

0.5

Note: This figure plots the gender gap in each job priority separately for the treatment and control groups.

In suitable organizations, the approach can be highly beneficial: employees who find purpose in their roles tend to show improved performance, higher earnings, and greater job satisfaction. This is especially true with the growth of generative AI in the workplace. However, there are two limiting factors for firms seeking to implement their own DYP intervention: First, the firm must present a work culture where a worker’s search for meaning can influence their career path, even if that means leaving the firm. Second, firms must commit to keeping all information private. As the authors observe, while more research is needed to establish the efficacy of DYP, this research has the potential to help everyone, from the labor productivity of firms to the well-being of individual workers.

internal rate of return (IRR): a financial metric that calculates the expected annual growth rate of an investment, considering the time value of money. It’s the discount rate at which the net present value (NPV) of all cash flows from an investment equals zero. In simpler terms, it’s the rate at which an investment breaks even.

READ THE WORKING PAPER NO. 2025-67 · MAY 2025

Meaning at Work bfi.uchicago.edu/working-papers/meaning-at-work

ABOUT OUR SCHOLARS

Virginia Minni

Assistant Professor, Chicago Booth

Luigi Zingales

Robert C. McCormack Distinguished Service Professor of Entrepreneurship and Finance, Chicago Booth

Written by Elisa Hsieh • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

Mechanism Design for Personalized Policy: A Field Experiment Incentivizing Exercise Based on BFI Working Paper No. 2025-48, “Mechanism Design for Personalized Policy: A Field Experiment Incentivizing Exercise,” by Rebecca Dizon-Ross, University of Chicago, and Ariel D. Zucker, University of California Santa Cruz

Personalizing policies can substantially improve program performance; in a case involving exercise incentives for individuals with lifestyle-related health conditions, such policies increased the treatment effect of incentives by 80% without increasing program costs. Making policy entails setting rules, regulations, and other guidelines with a desired outcome in mind. However, people respond differently to such efforts, and these heterogeneous responses necessarily complicate a one-size-fits-all approach to policymaking. But what if we could personalize policy? What if we could tailor a policy to individual characteristics and, thus, improve policy outcomes? Such an effort is challenging, in part, because policymakers cannot observe everyone’s behavior or type. Further, when people’s objectives diverge from the policymaker’s, individuals may strategically misreport their types. This information problem makes it harder for the policymaker to personalize policies effectively.

This paper uses a field experiment to test whether policies can be designed to overcome this information problem and, thus, effectively personalize policy. The authors consider a policy that uses financial incentives to influence behavior. Such policies are increasingly common in such sectors as education, the environment, and preventive health. For example, a workplace may incentivize participation in a step count program to improve health outcomes. However, one step plan likely does not fit all participants, with high walkers needing a higher step count and low walkers benefitting more from a lesser goal. If participants were all assigned a personalized goal based on their reported step counts and given $5 to reach their goal, it is easy to imagine certain participants choosing to misrepresent their step

Figure 1 · The Impact of Choice on Steps and Payment The Impact of Choice on Steps and Payment A) Daily Steps: Choice vs. Fixed Medium

B) Daily Payments: Choice vs. Fixed Medium

9,000 Daily Steps

10 Daily Payments in INR

8,500

8,000 6 7,500 4 7,000 2 6,500

6,000

7,192

7,720

8,140

Monitoring

Fixed Medium

Choice

5.9

6.3

Monitoring

Fixed Medium

Choice

Note: Panel (a) of this Figure illustrates that Choice substantially increases average steps relative to Fixed Medium. While the Medium target increases daily steps by 528 steps relative to Monitoring alone, Note: Panel5Aminutes illustrates that Choicethe substantially increases average steps relative to Fixed Medium. Whileatthe increases increase daily steps by 52880%. steps Monitoring or roughly of brisk walking, Choice treatment increases walking by an additional 420 steps (significant theMedium 5% level)target or 4 minutes—an of roughly In relative contrast,to panel (b) showsalone, or roughly Choice significantly increasetreatment payments, increases with the point estimate suggesting a mere change. 5that minutes ofdoes brisknot walking, the Choice walking by an additional 4208% steps (significant at the 5% level) or 4 minutes—an increase of roughly 80%. In contrast, panel (b) shows that

Choice does not significantly increase payments, with the point estimate suggesting a mere 8% change.

bfi.uchicago.edu

count to manipulate their assigned goal. For example, high walkers might falsely report a lower step count to get a lower goal and an easier time earning the incentive payment, and so everyone would end up with the program with the low step count. Mechanism design design, wherein participants are given a menu of contracts designed to incentivize choices that align with a policy’s objectives, approaches a solution to this predicament. From our example above, that could mean offering one contract with higher goals and higher payments, and another with more modest goals and lower payments. In this case, the higher payment could induce the higher walker to choose the contract with the more ambitious goal, while the low walkers would choose the less ambitious contract, thus allowing the policymaker to achieve personalization. Such programs are considered “incentive-compatible,” incentive-compatible in that they ensure that participants have an incentive to choose the contract that aligns with the policymaker’s objective. The authors employ mechanism design to personalize a policy that encourages exercise among 6,800 urban Indian adults to address diabetes, hypertension, and their precursors. Participants were provided pedometers and incentives to meet daily step targets. The authors personalized their program by allowing some participants to choose their incentive contracts from an incentive-compatible menu where contracts with higher step targets featured higher incentive payments. They randomly assigned participants either to a treatment group, or the Choice group; to one of three Fixed groups that each received a uniform (not personalized) step target; or to a Monitoring group that received a pedometer but no incentives. They find the following: •

Choice almost doubles the effectiveness of the incentive policy relative to a uniform, intermediate step target. The Choice treatment increases walking by roughly 4 additional minutes per day, an

80% improvement that both the medical literature and the authors’ experimental data suggest is likely to yield meaningful health impacts. •

The Choice treatment achieves this increase in walking without an increase in payments. Moreover, Choice yields gain across the full distribution of walking; that is, Choice achieves the gains of the low target at the bottom of the distribution and of the high target at the top but avoids the downside of “neglecting” one part of the distribution.

•

Consistent with a standard mechanism design model, the Choice menu is effective because participants sort into contracts in a way that is advantageous to the principal. Specifically, the authors empirically confirm the theoretical prediction that a principal would prefer to assign higher step targets to participants who walk more in the absence of incentives (i.e., who have higher “baseline steps”) and lower targets to those who walk less, as higher step targets generate relatively more steps (but not more payments) from participants with higher baseline steps. Moreover, participants with higher baseline steps choose higher step targets on the menu.

Bottom line: Choice matters. When offered an incentive-compatible menu, many participants prefer the contract that increases their steps most, relative to their payments. This finding has wide implications. Similar incentive-compatible menus could be used to incentivize other beneficial behaviors, such as schooling, R&D by firms, or the adoption of ecofriendly technologies. Incentive-compatible menus could also personalize other types of policies besides incentives, including unemployment insurance, where such menus could enable participants to balance the duration of benefits against the payout levels, sorting based on their underlying employability.

mechanism design: Focuses on designing rules, or “mechanisms,” that incentivize individuals to reveal their private information and achieve desired outcomes, even when those individuals act in their own self-interest. In other words, you start with a desired goal and work backward to create a system that achieves it. incentive compatibility: In game theory and economics, incentive compatibility refers to a situation where individuals have an incentive to reveal their true preferences or act in a way that aligns with the desired outcome of a system or policy. Essentially, it means that the mechanism is designed in such a way that it is beneficial for individuals to be truthful.

READ THE WORKING PAPER NO. 2025-48 · APRIL 2025

Mechanism Design for Personalized Policy: A Field Experiment Incentivizing Exercise bfi.uchicago.edu/working-papers/mechanism-design-forpersonalized-policy-a-field-experiment-incentivizing-exercise

ABOUT OUR SCHOLAR

Rebecca Dizon-Ross

Associate Professor of Economics and Charles E. Merrill Faculty Scholar, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

Saved by Medicaid: New Evidence on Health Insurance and Mortality from the Universe of Low-Income Adults Based on BFI Working Paper No. 2025-66, “Saved by Medicaid: New Evidence on Health Insurance and Mortality from the Universe of Low-Income Adults,” by Angela Wyse, Dartmouth College; and Bruce Meyer, University of Chicago

Recent Medicaid expansions increased enrollment by 12 percentage points and reduced mortality by 2.5% among low-income adults, saving roughly 27,400 lives for only $5.4 million and $179,000 per life and life-year saved, respectively.

The authors focus on the Medicaid expansions implemented under the Affordable Care Act, which states adopted in a staggered manner beginning in 2014. To estimate the causal effect of health insurance on mortality, they use data on 37 million low-income adults, constructed by linking the 2010 Census to administrative tax and mortality records. They compare mortality outcomes in states before and after expansion and find the following:

•

Medicaid expansions increased enrollment by 12 percentage points and reduced mortality by 2.5% in the low-income adult population. These

Figure 1 · Effect of Medicaid Expansion on Mortality Risk

Effect of Medicaid Expansion on Mortality Risk 10 Year of Expansion (2010)

% Change in Mortality Risk

Medicaid is the largest means-tested transfer program in the United States. At an annual cost of over $700 billion, the program insures one in four people in the United States, with enrollment rising by 50% between 2010 and 2021. Despite this substantial investment, credible estimates of the impacts of Medicaid on adult mortality remain limited. Motivated by this, in this paper, the authors examine whether expanding Medicaid eligibility leads to reductions in mortality, particularly among low-income adults.

-5

-10

-2

-1

2 3 4 5 Years Before and After Expansion

Note: This figure shows the impact of Medicaid expansions on mortality risk among non-disabled low-income adults ages 19-59 This in 2010. The vertical bars show 95% confidence intervals. Note: figure shows the impact of Medicaid expansions on mortality risk among non-disabled

low-income adults ages 19-59 in 2010. The vertical bars show 95% confidence intervals.

estimates suggest that people who enrolled in Medicaid experienced a 21% reduction in their mortality risk, on average, assuming no spillovers on those who did not enroll. •

Mortality reductions accrued not only to older age cohorts, but also to younger adults, who accounted for nearly half of life-years saved due

bfi.uchicago.edu

to their longer remaining lifespans and large share of the low-income adult population. •

•

Medicaid expansions saved the lives of 27,400 people between the 2010 passage of the Affordable Care Act and 2022, and a further 12,800 deaths could have been avoided in states that did not expand Medicaid. These expansions appear to be cost effective, with direct budgetary costs of $5.4 million per life saved and $179,000 per life-year saved falling well below valuations of life commonly found in the literature. The authors estimate that 5–20% of the mortality gap between low- and high-income Americans could be attributed to differences in health insurance coverage.

More on this Topic The Effect of Medicaid on Crime: Evidence from the Oregon Health Insurance Experiment The authors study the impact of random access to Medicaid health insurance on the likelihood of criminal charges or convictions. The Effect of Medicaid on Care and Outcomes for Chronic Conditions: Evidence From the Oregon Health Insurance Experiment The authors study the impact of random access to Medicaid health insurance on healthcare utilization and the management of chronic health conditions. Does One Medicare Fit All? The Economics of Uniform Health Insurance Benefits The authors argue for a basic universal insurance with limited coverage for expensive, lower value services.

Medicaid saves lives. This research highlights the significant health improvements caused by Medicaid expansions and avoidable deaths in states that have not yet expanded, while also bringing attention to potential adverse consequences from administrative barriers to Medicaid enrollment, the unwinding of continuous enrollment policies established during the COVID-19 pandemic, and other proposed restrictions to Medicaid currently under debate in Congress.

READ THE WORKING PAPER NO. 2025-66 · MAY 2025

Saved by Medicaid: New Evidence on Health Insurance and Mortality from the Universe of Low-Income Adults

ABOUT OUR SCHOLAR

Bruce Meyer

McCormick Foundation Professor, Harris School of Public Policy

bfi.uchicago.edu/working-papers/saved-by-medicaid-newevidence-on-health-insurance-and-mortality-from-theuniverse-of-low-income-adults

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

Stablecoin Runs and the Centralization of Arbitrage Based on BFI Working Paper No. 2025-76, “Stablecoin Runs and the Centralization of Arbitrage,” by Yiming Ma, Columbia University; Yao Zeng, University of Pennsylvania; and Anthony Lee Zhang, University of Chicago

Stablecoins are vulnerable to panic runs due to their reliance on imperfectly liquid reserve assets. They also face price instability due to a market structure that often concentrates arbitrage among a few participants. Issuers face a fundamental tradeoff: increasing arbitrage efficiency improves price stability in secondary markets but raises the risk of runs by reducing the price impact of investor redemptions. Stablecoins are crypto assets designed to maintain a value of $1. Issuers of stablecoin ensure this stability by holding an equivalent amount of US dollar-denominated assets in reserve, such as bank deposits, Treasuries, corporate bonds, and loans. Because dollars can’t move natively on blockchains, stablecoins enable dollar-based transactions within crypto systems, combining fiat stability with the speed and flexibility of digital assets.

Despite their promise of stability, however, stablecoins often trade above or below $1. In this paper, the authors examine how stablecoins’ unique market structure contributes to their price and financial (in)stability. Stablecoins operate in a two-tiered market. Only select institutional arbitrageurs can mint and redeem tokens directly with the issuer at a fixed $1 price, while most investors trade on secondary markets where prices fluctuate with supply and demand. Arbitrage

Figure 1 STABLECOIN ISSUER

Small set of institutional arbitrageurs can mint/ redeem at $1

PRIMARY MARKET Create new coins when price > $ Redeem coins when price < $1

Arbitrage

SECONDARY MARKET Exchanges like Coinbase and Binance Open to all investors Prices fluctuate with supply/demand

bfi.uchicago.edu

between the two markets allows institutional traders to profit when secondary market prices diverge from $1. This can pull prices back towards $1, but also force issuers to liquidate reserves when arbitrageurs request redemptions.

•

The model reveals a central tradeoff whereby improving price stability through more efficient arbitrage increases run risk, while limiting arbitrage lowers financial fragility but allows larger and more persistent price deviations.

The authors begin by collecting and analyzing data across six major fiat-backed stablecoins, along with market price data and institutional details on redemption mechanics and access. They identify the following stylized facts concerning stablecoin arbitrage:

•

The trading price of stablecoins in the secondary market commonly deviates from $1.

Stablecoin issuers optimally choose the degree of arbitrage access based on their reserve liquidity. Issuers holding illiquid assets find it optimal to restrict arbitrage to limit costly redemptions and protect against fire sales. Investors value both price stability and run risk when choosing which stablecoin to hold.

•

2. The redemption and creation of stablecoins in the primary market is performed by a small set of arbitrageurs, whose concentration varies by stablecoin.

Calibrating the model to September 2021 data, the authors estimate the probability of a run at 2.50% for USDT (Tether) and 2.13% for USDC (USD Coin), two of the largest stablecoins.

•

The model’s policy counterfactuals show that redemption fees can reduce run risk by discouraging mass redemptions, but they also weaken the price peg by dampening arbitrage incentives.

•

Allowing stablecoin issuers to pay dividends from reserve earnings can reduce run incentives while preserving price stability, offering a potentially effective design and regulatory tool.

3. Stablecoins with a more concentrated set of arbitrageurs experience more pronounced price deviations in the secondary market. 4. Stablecoins differ in how much they transform short-term liabilities into long-term or illiquid assets. The mismatch between liabilities and assets means that during periods of stress, coin issuers may need to sell illiquid reserves at a loss to meet redemptions. Building on these patterns, the authors adapt Diamond and Dybvig’s (1983) model of bank runs, which explains how bank runs can occur even when banks hold riskless assets, to the unique structure of stablecoins to investigate how arbitrage impacts the asset’s riskiness. Their model shows the following: •

Stablecoins are exposed to run risk because illiquid reserves must be sold at a discount to meet redemptions during stress. This creates a first-mover advantage, as investors who redeem early are more likely to receive full value.

READ THE WORKING PAPER

In the last several years, the market for stablecoins has grown exponentially — the six largest US dollar-backed coins had a market capitalization of $5.6 billion at the start of 2020, by the beginning of 2022 they were worth over $130 billion. The growth in value has attracted the attention of legislators across jurisdictions, stablecoin regulation has been drafted and is under consideration in the US, UK, and EU. This paper has direct policy relevance for these discussions, as it highlights a key tradeoff between price stability and market stability, and evaluates several policy solutions.

ABOUT OUR SCHOLAR

NO. 2025-76 · JUNE 2025

Anthony Lee Zhang

Stablecoin Runs and the Centralization of Arbitrage

Associate Professor of Finance, Chicago Booth

bfi.uchicago.edu/working-papers/stablecoin-runs-and-thecentralization-of-arbitrage

Written by Daniel Koslovsky with Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

The Persistence of Female Political Power in Africa Based on BFI Working Paper No. 2025-71, “The Persistence of Female Political Power in Africa,” by Siwan Anderson, University of British Columbia; Sophia du Plessis, Stellenbosch University; Sahar Parsa, New York University; and James A. Robinson, University of Chicago

Regions and ethnic groups in Africa with a higher historical prevalence of traditional female political leadership tend to have a higher proportion of elected female representatives in today’s political institutions. Institutional, rather than economic, factors significantly shape the traditional political influence of women, and institutional changes enforced by colonial powers reversed female political power. Global efforts to increase women’s political representation often frame the issue as a universal challenge. Yet Africa presents a notable exception: the continent has produced numerous female heads of state (in 25 of 54 African countries, women have served as either president, vice president, or premier), and boasts some of the highest rates of female legislative representation in the world. Rwanda, for example, leads globally

with 64% of parliamentary seats held by women, more than double the rate in the United States.

Figure 1 · Traditional Female Leadership Traditional Female Leadership

Figure 2 · Contemporary Female Political Representation Contemporary Female Political Representation

Prevalence of Traditional Female Political Leadership Low High

Note:This This map map shows shows the homelands based on the Murdock Note: theboundaries boundariesofofethnic ethnic homelands based on the Murdock Ethnographic Ethnographic Atlas. Traditional ethnic homelands that allowed women to hold Atlas. Traditional ethnic homelands that allowed women to hold political leadership positions political leadership positions are shown in blue, while those that did not are in green. are shown in blue, while those that did not are in green.

What explains this record? In this paper, authors explore whether the roots of contemporary female political power in Africa lie in precolonial institutions, specifically, in places where women traditionally held political leadership roles. To test this, the authors compile two original

Share of Elected Political Seats Held by Females 0-10% 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Note: This map shows ofadministrative administrative units, which are shaded Note: This map showsthe theboundaries boundaries of units, which are shaded to to indicate the indicate the share of elected female political share of elected female political seats withinseats eachwithin unit. each unit.

bfi.uchicago.edu

datasets: one detailing female political leadership in precolonial societies and another on current female representation in local elections. They use these data to assess whether regions with a higher historical prevalence of traditional female political leadership also exhibit greater female representation in local political institutions today. They find the following: •

•

Female political power tends to persist over time. Females hold 13% more elected political seats, on average, in regions where females traditionally held political influence. This pattern holds even between pairs of neighboring ethnic groups where one traditionally permits female political leaders and the other does not, confirming a significant positive influence (of a similar magnitude) from traditional to contemporary female political power. Institutional changes enforced by colonial powers reversed some aspects of female political power. For example, political institutions created by colonial powers, particularly the British, were handed to men to control; and the political representation of women is less likely to persist in ethnicities that were split by colonial borders.

READ THE WORKING PAPER NO. 2025-71 · MAY 2025

The Persistence of Female Political Power in Africa bfi.uchicago.edu/working-papers/the-persistence-offemale-political-power-in-africa

•

Institutional, rather than economic, factors significantly shape the traditional political influence of women. There is a strong positive correlation between political centralization and female representation. In addition, cultural norms shape women’s political power differently across political structures. Matrilineality strengthens women’s political influence in centralized states, likely because dominant matrilineal clans gain prominence as states consolidate power. In contrast, matrilocality matters more in smaller, less centralized societies, where women may leverage kin-based solidarity to achieve representation.

Traditional female political power has a strong, persistent influence on the representation of females in local-level political institutions in sub-Saharan Africa today. Given that in many developing country contexts, particularly in rural areas, ethnic and lineage-based political institutions prevail, and can co-exist alongside more formal administrative units, it is pertinent to understand and acknowledge more broadly the influence of women in these enduring indigenous institutions.

ABOUT OUR SCHOLAR

James Robinson

Professor, Harris School of Public Policy and Department of Political Science; Fellow, Institute of African Studies at the University of Nigeria at Nsukka

Written by Abby Hiller • Designed by Maia Rabenold

RESEARCH BRIEF • JUNE 2025

The Reverse Cargo Cult: Why Authoritarian Governments Lie to Their People Based on BFI Working Paper 2025-73, “The Reverse Cargo Cult: Why Authoritarian Governments Lie to Their People,” by Konstantin Sonin, University of Chicago

Verifiable lies told by politicians change citizens’ perceptions of politicians generally and reduce citizen’s willingness to replace their leaders. Authoritarian regimes invest heavily in propaganda. These lies serve not only to persuade citizens to support the acting regime, but also to shape how citizens interpret external political information, including what they believe is true, trustworthy, or even possible. In this paper, Konstantin Sonin examines why a regime would

lie in ways that citizens can recognize as false, and how might those lies still reinforce political control. Sonin argues that these lies serve a strategic purpose: to prompt citizens to generalize. Rather than focusing only on the failures of their own regime, citizens exposed to transparent

The Kitchen Debate Sonin brings this argument to life with a personal story about his father’s visit to the 1959 American National Exhibition in Moscow. The exhibit famously featured a “typical” American kitchen stocked with modern appliances—dishwashers, electric juicers, and refrigerators—standard features for middle-class American households at the time. For most Soviet visitors, however, these items were unfamiliar or entirely unattainable. The Soviet government set up a parallel display of its own appliances, meant to suggest rough parity. But as Sonin notes, many attendees, accustomed to state lies, assumed that both sides were exaggerating. “Naturally,” he writes, “my father and many others assumed that the U.S. kitchen was as fake as the Soviet display.” Rather than showcasing capitalism’s superiority, the exhibition had a leveling effect: it reinforced the belief that everyone was lying. Khrushchev himself reportedly dismissed the American kitchen as implausible. The episode, famously captured in the Nixon–Khrushchev “kitchen debate” photo, viscerally illustrates the paper’s core insight: when citizens are steeped in domestic propaganda, they often become skeptical of all political messaging.

bfi.uchicago.edu

falsehoods may become disillusioned with political elites more broadly. Sonin’s model yields several key insights about authoritarian propaganda: •

When citizens observe their leader lie about something obviously false, they become more likely to believe that other politicians lie too, including those in other countries. This comparative skepticism reduces the appeal of democratic alternatives and dampens the motivation to replace the incumbent regime.

•

The strategy is particularly effective when the regime can mediate or restrict access to foreign information, such as through state-run media or controlled exhibitions.

•

The model also applies to democratic settings. A candidate might tell an obvious lie not to persuade directly, but to undermine the credibility of an opponent.

Authoritarian propaganda is not just about spreading lies, it is about shaping the cognitive environment in which citizens make sense of the world. Authoritarian regimes deploy messaging not to persuade in the traditional sense, but to reshape comparison, suppress alternatives, and limit hope. This research offers a powerful framework for understanding how propaganda works—even when no one believes it.

READ THE WORKING PAPERS NO. 2025-73 · MAY 2025

The Reverse Cargo Cult: Why Authoritarian Governments Lie to Their People

ABOUT OUR SCHOLAR

Konstantin Sonin

John Dewey Distinguished Service Professor

bfi.uchicago.edu/working-papers/the-reverse-cargo-cultwhy-authoritarian-governments-lie-to-their-people

Written by Abby Hiller • Designed by Maia Rabenold

100

RESEARCH BRIEF • JUNE 2025

The Role of Risk and Ambiguity Preferences on Early-Childhood Investment: Evidence from Rural India Based on BFI Working Paper No. 2025-47, “The Role of Risk and Ambiguity Preferences on Early-Childhood Investment: Evidence from Rural India,” by Michael Cuna, University of Chicago; Lenka Fiala, University of Ottawa, Institute for Replication, and Tilburg University; Min Sok Lee, University of Chicago; John A. List, University of Chicago; and Sutanuka Roy, The Australian National University

Mothers in villages in Rajasthan, India, who exhibit more risk and ambiguity aversion tend to invest more in their young children’s nutrition. These investments improve skill development and even attenuate the negative impacts of certain disadvantages in children’s outcomes. Parents’ investments in their children are a key driver of kids’ long-run outcomes. Despite this, some parents invest heavily in their children, while others do not. In this paper, the authors examine the role of parents’ risk and ambiguity preferences in driving this variation, studying how parents’ tendency towards accepting risk and uncertainty affect their choices regarding earlychildhood education and health investments. The authors study this question using a sample of mothers in 495 villages in the district of Udaipur, in the State of Rajasthan, India. They begin by using methodology established in prior literature to elicit mothers’ ambiguity attitudes and risk aversion. The authors then compare these measures to mothers’ nutrition investments as well as to results from tests of children’s cognitive and noncognitive skills. They find the following:

•

There is substantial variation in measures of mothers’ risk and ambiguity aversion.

•

The more risk and ambiguity averse the mother, the greater her investments in her children’s nutrition, including giving essential milk food products and providing vegetables and fruits, juices, and liquids other than water. For the youngest children in the authors’ sample, the estimates of ambiguity and risk aversion positively predict months of breastfeeding.

•

These investments are positively associated with children’s cognitive and non-cognitive development: mothers who exhibit higher levels of risk and ambiguity aversion tend to have children with stronger early-life skills. Notably, these relationships persist even after controlling for socio-economic differences.

bfi.uchicago.edu

101

In fact, the beneficial effects of maternal risk and ambiguity aversion can partially offset disadvantages linked to factors such as maternal illiteracy, belonging to historically marginalized social groups, and lack of access to media or mobile phones for all measures of cognitive and non-cognitive early-life skills. This study highlights the profound impact that maternal risk and ambiguity aversion can have on early childhood investments and subsequent developmental outcomes. By demonstrating that mothers who exhibit higher levels of these traits tend to invest more in their children’s nutrition, leading to enhanced cognitive and non-cognitive skills, the authors uncover a crucial mechanism for mitigating socio-economic disparities. These findings suggest that targeted support for atrisk populations could leverage these maternal tendencies to foster better outcomes for children, even in the face of socio-economic disadvantages. This result underscores the importance of considering individual preferences and beliefs in the design of public policies aimed at reducing inequality and promoting economic growth.

READ THE WORKING PAPER NO. 2025-47 · MARCH 2025

The Role of Risk and Ambiguity Preferences on Early-Childhood Investment: Evidence from Rural India bfi.uchicago.edu/working-papers/the-role-of-risk-andambiguity-preferences-on-early-childhood-investmentevidence-from-rural-india

ABOUT OUR SCHOLARS

Michael Cuna

PhD Candidate, Kenneth C. Griffin Department of Economics

Min Sok Lee

Assistant Senior Instructional Professor in Economics, Kenneth C. Griffin Department of Economics

John A. List

Kenneth C. Griffin Distinguished Service Professor in Economics and the College, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

102

RESEARCH BRIEF • JULY 2025

Administrative Fragmentation in Health Care Based on BFI Working Paper No. 2025-77, “Administrative Fragmentation in Health Care,” by Riley League, University of Illinois Urbana-Champaign; and Maggie Shi, University of Chicago

A recent Medicare reform that aimed to reduce administrative fragmentation by consolidating billing processes successfully reduced fragmentation, but had only modest effects on administrative efficiency and no discernible impact on patient outcomes or hospital administrative costs. More than half of Americans over age 65 and 13% of all Americans are covered by more than one source of health insurance. This overlapping coverage creates “administrative fragmentation,” or the lack of standardization in billing and administrative processes across payers. The resulting complexity can lead to costly inefficiencies and frustrating delays or denials for patients and providers. But how much do these inefficiencies actually matter? In this paper, the authors study a major Medicare reform that consolidated administrative contracts for hospital and outpatient claims within jurisdictions, eliminating the need for multiple parties to handle processing different parts of the same patient’s care. Importantly, while the reform simplified billing processes, it left other factors— such as payment rates and benefits—unchanged, allowing the researchers to isolate the effects of administrative fragmentation.

Figure 1 · Effect of Reform on Patient Care A) Total Spending 0.01 Total Medicare Payments

0.05

-0.05 Reform -0.1

-24

-18

-12

-6

•

Administrative fragmentation fell sharply. The share of patients who encountered multiple administrative entities during and after hospital stays declined by more than 35 percentage points following the reform. Improvements in administrative efficiency were modest. The reform led to a small (2.3%) reduction in denial rates for physician services delivered after hospital discharge, but had

B) Number of Providers Encountered 0.02 Providers Interacted With

0.01

-0.01

-0.02

Using Medicare administrative data from 20072017 and leveraging the staggered rollout of the reform across jurisdictions, the authors compare hospitals with varying baseline levels of administrative fragmentation to estimate causal effects. They find the following:

Months Before and After Reform

-0.1

Reform

-24

-18

-12

-6

Months Before and After Reform

C) 30-Day Readmissions 0.01 30-Day Readmission Rate

0.005

Reform -0.005 -24

-18

-12

-6

Months Before and After Reform

Note: These figures show the effects of the reform on the care patients received after their initial admission. Error bars give the 95% confidence intervals.

bfi.uchicago.edu

103

no measurable effect on the time it took to process claims. •

Hospital administrative costs and technology adoption remained unchanged. There was no evidence that the reform reduced hospitalreported administrative costs or increased the adoption of billing software and other administrative technologies by hospitals or outpatient providers.

•

There was no effect on patient care or outcomes. The authors find precise null effects on downstream utilization, total Medicare spending, provider access, and 30day hospital readmission rates.

These findings suggest that streamlining billing alone is unlikely to reduce administrative costs or meaningfully affect patient care or outcomes. The authors caution that simply routing all claims through a single administrative entity does not, on its own, improve efficiency or care delivery. It is possible that broader efforts to harmonize other aspects of payer fragmentation, such as inconsistencies in payment rates, coverage rules, and eligibility criteria, may be generate more substantial improvements.

READ THE WORKING PAPER NO. 2025-77 · MAY 2025

Administrative Fragmentation in Health Care bfi.uchicago.edu/working-papers/administrativefragmentation-in-health-care

ABOUT OUR SCHOLAR

Maggie Shi

Assistant Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

104

RESEARCH BRIEF • JULY 2025

Innovator Networks Within the Firm and the Quality of Innovation Based on BFI Working Paper No. 2025-85, “Innovator Networks Within the Firm and the Quality of Innovation,” by Michael Gibbs, University of Chicago; Friederike Mengel, University of Essex; and Christoph Siemroth, University of Essex

Employees with more direct collaborators produce higher quality ideas, while those who bridge different groups face short-term costs but create valuable spillovers for colleagues. Remote work significantly disrupts these innovation networks. A long-standing idea in the social sciences is that networks matter for innovation. They facilitate knowledge transfer and diffusion, and stimulate creativity by providing access to different types of knowledge and perspectives. Ronald Burt’s influential work on “structural holes” suggests that individuals who bridge disconnected groups have advantages in detecting and developing valuable opportunities. Despite the theoretical importance of networks for innovation, empirical evidence remains limited.

In this paper, the authors fill this gap using data from over 28,000 innovators within a major IT services firm. They track employee ideas submitted through the company’s formal suggestion system, and examine how employees’ network positions affect whether their ideas are accepted and implemented. The authors measure three key network characteristics: degree

Remote work significantly disrupts innovation networks.

Figure 1 · Innovation Within Innovation NetworksNetworks Within the Firm the Firm A) Winter 2018-2019

B) Summer 2019

C) Summer 2020 Dense collaboration within teams

Bridging employees connect different clusters

Knowledge flows through bridges

Note: These network graphs provide examples of actual innovation networks IT services company, illustrating how employees collaborate ideasdifferent across different Note: These network graphs provide examples of actual innovation networks from from the ITthe services company, illustrating how employees collaborate on ideason across time periods. Each dot time periods. Each dot and represents an collaboration employee, and lines show collaboration relationships. represents an employee, lines show relationships.

bfi.uchicago.edu

105

(number of direct collaborators), network size (total people in one’s innovation network), and bridge centrality (the extent to which someone connects different clusters of employees). They find the following: •

•

Having more direct collaborators significantly improves idea quality. Each additional collaboration partner increases an employee’s idea acceptance rate by approximately 2.5 percentage points. This effect appears to work through the immediate effort and expertise that collaborators contribute to developing the idea, rather than through lasting knowledge transfer. “Bridging” across different groups within the company creates both costs and benefits. Employees who connect otherwise disconnected clusters of colleagues tend to produce lower-quality ideas in the short term, likely due to the coordination and communication challenges of working across different teams and perspectives. However, past bridging activity shows positive effects on current innovation, suggesting that the knowledge gained from bridging pays dividends over time. Bridging creates powerful positive spillovers for others. While bridgers themselves may struggle with immediate coordination costs, having a high-bridging colleague in one’s network increases idea quality by over 3 percentage points. This suggests that bridgers serve as

READ THE WORKING PAPER NO. 2025-85 · JUNE 2025

Innovator Networks Within the Firm and the Quality of Innovation bfi.uchicago.edu/working-papers/innovator-networkswithin-the-firm-and-the-quality-of-innovation-2

valuable conduits of diverse knowledge and perspectives that benefit their entire network. •

Network size itself doesn’t matter much. After accounting for direct collaborations and bridging, simply being in a larger innovation network shows minimal impact on idea quality.

•

Remote work significantly disrupts innovation networks. During the pandemic’s work-fromhome period, employees collaborated with fewer people. The subsequent hybrid work period showed even more dramatic network degradation, with substantial decreases in collaborations, network size, and bridging activity.

This research provides rare insight into how the social architecture of innovation actually works within organizations. The findings suggest that while collaboration clearly enhances innovation, the most valuable network positions—those that bridge different parts of the organization— come with short-term costs that may discourage employees from pursuing them. Companies might need to explicitly reward bridging behavior to capture its substantial collective benefits. The implications extend beyond individual career strategy to organizational design. Understanding these network dynamics could help companies structure teams and work arrangements to maximize innovation. As firms grapple with remote work policies, these findings highlight how administrative decisions about workplace arrangements can profoundly affect the collaborative networks that drive innovation.

ABOUT OUR SCHOLAR

Michael Gibbs

Konstantin Sokolov Clinical Professorship of Economics, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

106

RESEARCH BRIEF • JULY 2025

Measuring Markets for Network Goods Based on BFI Working Paper 2025-80, “Measuring Markets for Network Goods,” by Leonardo Bursztyn, University of Chicago; Matthew Gentzkow, Stanford University; Rafael Jiménez-Durán, Bocconi University; Aaron Leanord, University of Chicago; Filip Milojević, University of Chicago; and Christopher Roth, University of Cologne

Users value apps like Instagram and YouTube more when TikTok is collectively banned than when TikTok is individually deactivated, suggesting the importance of accounting for network effects when defining markets. Defining a good’s relevant market is a fundamental component to antitrust analysis. Without a properly defined market, regulators cannot fairly assess a company’s market power, or the harms incurred by anti-competitive behavior. The first step to defining a good’s relevant market is to identify its substitutes—goods or services substitutes that can be used in place of one another to satisfy the same need or want.

Figure 1 · Average Difference in Valuations Across Scenarios by Platform Average Difference in Valuations Across Scenarios by Platform $30 Difference in Valuations Individual Deactivation - No Ban Ban - Individual Deactivation Ban - No Ban

In this paper, the authors focus on a critical but often overlooked aspect of substitution analysis, network effects effects. A network effect occurs when the value of a good to one user depends on how many other people are using it. For example, the more users a social media platform has, the more valuable it may become to any individual user. The authors investigate how overlooking such network effects can distort substitution patterns and, by extension, market definitions. The authors conduct an online survey of 900 active TikTok users aged 18 to 27 between January 6-9, 2025.1 Respondents are first 1 The timing of the survey is crucial to the study design because there was tremendous uncertainty regarding TikTok’s future availability in the US due to a pending ban on January 19th. The study leveraged the plausibility that TikTok would no longer be available to US users in order to elicit responses that are more reliable than if TikTok’s ban were purely hypothetical.

$7.50

$13.70

Instagram

$21.10

$10.60

$12.10

YouTube

$22.70

$-0.10

$7.80

$7.70

Snapchat

Note: This figure illustrates the differences in valuations of the alternative app across the three scenarios. Note: figurethe illustrates the differences in valuations the alternative across the three The red This bars depict average difference between valuations underof the individual TikTokapp deactivation scenario and the TikTok scenario, blue bars show the difference average valuation scenarios. Thenored barsban depict thethe average difference betweeninvaluations underbetween the individual the TikTok ban and the individual TikTok deactivation scenario, and the green bars represent the average TikTok deactivation the no banscenario scenario, barsban show the difference in difference in respondents’scenario valuationsand between theTikTok TikTok Ban andthe the blue no TikTok scenario. The verticalvaluation bars indicate 95% confidence intervals. average between the TikTok ban and the individual TikTok deactivation scenario, and

the green bars represent the average difference in respondents’ valuations between the TikTok Ban scenario and the no TikTok ban scenario. The vertical bars indicate 95% confidence intervals.

provided background information on the pending TikTok ban. Then, respondents are randomly assigned to an alternative social media platform— either YouTube, Instagram, or Snapchat—and asked to identify the minimum amount of money that they would accept to deactivate the platform under three different scenarios: 1) TikTok is not banned (baseline); 2) TikTok is not banned but the respondent is forced to deactivate their TikTok account (individual deactivation); and 3) TikTok is banned (collective deactivation).

substitutes: goods or services that can be used in place of one another to fulfill the same need or want network effects: a phenomenon where the value of a good or service increases as more people use it

bfi.uchicago.edu

107

The authors compare the valuations from the three different scenarios to assess the substitutability of the alternative platforms to TikTok. They calculate both the difference in the percentage of respondents who value their alternative platform more in one scenario vs. another, as well as the differences in the average valuation of the alternative platforms across the different scenarios. The authors find the following: •

•

Users value alternative platforms more when TikTok is collectively banned compared to when there is no ban. The net fraction of users who report higher valuations under the collective ban is 48.1 percentage points for Instagram, 41.8 p.p. for YouTube, and 14.8 p.p. for Snapchat, indicating that all three platforms are perceived as substitutes for TikTok in a coordinated exit context. Comparing collective versus individual TikTok deactivations reveals strong network effects. The net fractions of users with relatively higher valuations under collective deactivation (compared to individual deactivation) are 25.0 percentage points for Instagram, 16.0 p.p. for YouTube, and 15.5 p.p. for Snapchat, suggesting that the utility of substitutes depends substantially on whether peers also migrate. When TikTok is deactivated individually (compared to no ban), there is a net increase of 13.9 percentage points in users who value Instagram more, and 24.4 percentage points for YouTube. For Snapchat, the net fraction is negative and close to zero, indicating it is not perceived as a substitute when users leave TikTok

READ THE WORKING PAPER

on their own. This suggests Snapchat’s utility depends heavily on peer coordination, likely due to its core function as a messaging app. In addition to the comparison of platform valuations across the three scenarios, the authors also find: •

A net positive fraction of respondents expect to spend more time on other social apps— namely, Instagram, YouTube, and Snapchat— under the TikTok ban compared to the individual TikTok deactivation. Conversely, intended substitution toward non-social activities, such as playing phone games or meditating, is weaker under the TikTok ban than under the individual TikTok deactivation.

•

Respondents’ expectations about changes in others’ time use on Instagram, YouTube, and Snapchat align with their substitution patterns. Users who expect an above-median increase in the time their friends spend on the assigned platform exhibit a larger gap in valuation between the TikTok ban and individual TikTok deactivation.

Antitrust regulators take note. Accounting for network effects is often critical in defining a good’s relevant market. For TikTok, accounting for network effects reveals that other social apps are closer substitutes than suggested by fixednetwork estimates, making it more likely that they are part of the relevant market. At the same time, our estimates suggest that non-social activities— such as video gaming and meditation—are weaker substitutes for social media, making it less likely that they are part of the relevant market.

ABOUT OUR SCHOLAR

NO. 2025-80 · JUNE 2025

Leonardo Bursztyn

Measuring Markets for Network Goods bfi.uchicago.edu/working-papers/measuring-markets-fornetwork-goods

The Saieh Family Professor of Economics, Kenneth C. Griffin Department of Economics

Written by Daniel Koslovsky with Abby Hiller • Designed by Maia Rabenold

108

RESEARCH BRIEF • JULY 2025

The Benefits of Scholastic Athletics Based on BFI Working Paper No. 2025-94, “The Benefits of Scholastic Athletics,” by James J. Heckman, University of Chicago; Colleen P. Loughlin, Compass Lexecon; and Haihan Tian, University of Chicago

On average, participation in scholastic athletics benefits participants, especially those from disadvantaged backgrounds. More generally, participation in athletics is beneficial (or not harmful) to high school or college athletes and at all levels of participation, varsity or intramural. A wave of recent attention on college athletics, owing in large part to a recent judicial ruling that allows schools to pay athletes, has placed renewed focus on the benefits of athletic programs for individual student athletes. Beneath the headlines lies a fundamental question: Is there any value to participating in high school and college sports, especially for those many thousands of student-athletes who will never get paid or who will never benefit from NIL (name, image, and likeness) endorsements, and do those benefits extend into post-college life? While this question has long interested researchers, this paper presents the first comprehensive longitudinal analysis of the benefits of participation in athletics at both the high school and college level. The authors analyze two large nationally representative datasets: The National Educational Longitudinal Survey (NELS) and the Education

Longitudinal Survey (ELS) to track the careers of student athletes as they progress through schooling and into the labor market. These studies control for a large array of personality, cognitive, and family background measure to control for selection bias. The authors find the following: •

Participation in high school athletics is associated with a higher probability of graduating from high school, a key milestone toward improved life outcomes.

•

Participation in high school athletics is significantly associated with higher probabilities of attending college. High school students, hoping to receive a scholarship, likely invest more in their academic and athletic skills to meet eligibility requirements and gain admission to college; the commitment to academics also brings future rewards.

NIL: Refers to a student-athlete’s ability to profit from their name, image, and likeness—elements of their personal brand. This concept is rooted in the “right of publicity,” which gives individuals control over how their identity is used for commercial purposes. In July 2021, new rules and state laws began allowing college athletes to earn money through sponsorships, endorsements, social media, and other business ventures.

bfi.uchicago.edu

109

Intercollegiate varsity athletes are as likely or more likely to earn at least a bachelor’s degree relative to otherwise comparable non-athletes. “One and done,” used to describe Division I basketball players who move to the NBA after one collegiate season, is an anecdote, not a valid characterization of college athletics, even for the Power Conferences. College varsity athletes are more likely to receive college scholarships that facilitate college completion in addition to health benefits and tutoring.

•

There are no adverse academic or labor market outcomes for students participating in intercollegiate basketball and football in NCAA Division I or FBS schools.

•

Participation in college intramural sports also yields benefits, with both longitudinal studies revealing higher graduation rates for male and female athletes.

READ THE WORKING PAPER NO. 2025-94 · JULY 2025

The Benefits of Scholastic Athletics bfi.uchicago.edu/working-papers/the-benefits-ofscholastic-athletics

•

Participation in college athletics is associated with better starting wages for participants compared to observationally identical nonathletes. This holds for both college varsity athletes and college intramural athletes.

•

Participation in scholastic athletics leads to college education with benefits for minorities and disadvantaged students; thus, scholastic athletics is a vehicle for social mobility, facilitating educational, social and economic opportunities. For example, high school varsity athletes of either sex from single-parent households are more likely to attend 4-year colleges compared to other high school non-athletes from single-parent households; both male and female high school varsity athletes below the poverty line are more likely to attend 4-year college; finally, black male high school varsity athletes are more likely to attend 4-year college compared to their non-athlete peers.

Bottom line: On average, participation in scholastic athletics—intercollegiate or intramural—benefits participants, especially those from disadvantaged backgrounds, with no adverse impacts. The authors stress that data, not anecdotes, should drive the discussion of the benefits of athletic participation and its role in promoting social mobility.

ABOUT OUR SCHOLARS

James J. Heckman

Henry Schultz Distinguished Service Professor in Economics, Kenneth C. Griffin Department of Economics; Director, Center for the Economics of Human Development

Haihan Tian

Predoctoral Fellow, Center for the Economics of Human Development

Written by David Fettig • Designed by Maia Rabenold

110

RESEARCH BRIEF • JULY 2025

The Local Root of Wage Inequality Based on BFI Working Paper No. 2025-78, “The Local Root of Wage Inequality,” by Hugo Lhuillier, University of Chicago

High-paying jobs concentrate in large cities while low-paying jobs exist everywhere, creating higher wages but greater within-city inequality. Over time, all workers in big cities are net winners as they benefit from steeper career ladders. Modern urban hubs present a paradox: they 1 ·Wage WhatInequality Causes Wage in a City? WhatFigure Causes in a Inequality City? offer both the highest wages and the greatest 0.3 inequality. In Paris, the average wage is 65% Wage Premia (Jobs) higher than in mid-sized French cities, yet Worker Characteristics Other Factors low-income workers there earn only 2% more 0.2 nominally and 20% less after accounting for housing costs. This raises fundamental questions: Why are wages higher in larger cities? How come the best and worst opportunities cohabit in these 0.1 places? And why would a worker move there if they earn less after housing costs? To understand these spatial wage patterns, Hugo 0 0.1 0.15 0.2 0.25 0.3 Lhuillier analyzes comprehensive French matched Inequality employer-employee data tracking workers across Note: This figure as that cities more unequal overalloverall (moving right),right), jobs account Note: Thisshows figure that shows asbecome cities become more unequal (moving jobs their careers and locations. By comparing workers for a larger sharefor ofathat inequality blue area), while worker remain account larger share of(growing that inequality (growing blue area),characteristics while worker characteristics relatively stable. remain relatively stable. who move between employers, this approach isolates how jobs shape wages within and across spatial wage inequality: it explains 34% of why cities. The analysis reveals two novel facts: wages are higher in bigger places, and 33% of • High-paying jobs are concentrated in large why there are more dispersed there. cities while low-paying jobs are dispersed • Workers earn high wages in large cities throughout space. In Paris, 17.9% of jobs over time as they climb steeper job ladders. rank in the top 10% nationally compared to Starting wages in Paris are similar to those in only 6.9% in mid-sized cities. Conversely, mid-sized cities after controlling for worker low-paying jobs appear equally across all characteristics. However, workers experience locations. Which jobs are where matters for

bfi.uchicago.edu

111

greater wage gains whenever they switch jobs: each switch delivers 1.5 percentage points higher wages than in mid-sized cities. Over time these gains compound and create substantial divergence, with workers in Paris earning 14% higher wages by age 55. These findings challenge standard explanations for spatial wage inequality. If cities simply had higher productivity, wages would be higher for all workers rather than concentrating at the top. If the explanation were purely stemming from worker heterogeneity, we would not observe systematic wages gains as workers reallocate across jobs. To explain these patterns, Lhuillier develops a theoretical model where companies strategically choose where to produce and hire workers, while workers’ careers are shaped by local job ladders. The key insight is that productive employers concentrate in large cities to access larger worker pools and sidestep hiring frictions. This creates fiercer competition for workers, but the benefits are asymmetric. On the one hand, superstar employers must offer disproportionally high wages to face off the local competition. On the other hand, workers coming out of unemployment cannot reap the productivity gains of large cities as they have little bargaining power. Higher average wages and greater inequality follows. To assess the impact of employers on spatial inequality, Lhuillier builds a quantitative version of the model and estimates it using firm-level data.

READ THE WORKING PAPER NO. 2025-78 · JUNE 2025

The Local Root of Wage Inequality bfi.uchicago.edu/working-papers/the-local-root-ofwage-inequality

The model’s predictions align with the data, and deliver three quantitative takeaways: •

Employers’ productivity, rather than inherent local productivity, is at the root of spatial wage gaps. This matters as the former is not policy invariant.

•

Large cities offer higher lifetime earnings. Despite facing 8.1% lower real wages initially, workers in Paris enjoy 4.2% higher lifetime real earnings than in mid-size cities thanks to the steeper local job ladder.

•

Large cities rely on a dynamic labor market. A counterfactual analysis shows that reducing the job switching rate nationally by one percentage point—as occurred in many countries during the 1990s and 2000s— would reduce Paris’s size by 22% and its productivity by 2 percentage points as its comparative advantage diminishes.

These findings offer a new explanation for the high wages found in cities. Rather than stemming from higher local productivity, wage advantages primarily result from the sorting of productive employers who compete intensely for talent. This competition creates job ladders that benefit workers throughout their careers, making cities attractive despite their costs. Understanding these dynamics is crucial for policies aimed at reducing spatial inequality while preserving the innovation and dynamism that make cities economic engines.

ABOUT OUR SCHOLAR

Hugo Lhuillier

Saieh Family Postdoctoral Fellow, Becker Friedman Institute

Written by Abby Hiller • Designed by Maia Rabenold

112

RESEARCH BRIEF • AUGUST 2025

Engineering Ukraine’s Wirtschaftswunder Based on BFI Working Paper No. 2025-91, “Engineering Ukraine’s Wirtschaftswunder,” by Ufuk Akcigit, University of Chicago; Furkan Kilic, University of Chicago; Somik Lall, World Bank; and Solomiya Shpak, World Bank

Policies targeting entrenched incumbents are essential for Ukraine’s post-war economic transformation, as institutional capture has systematically undermined business dynamism and productivity growth over the past two decades As Ukraine emerges from the devastation of war, it faces a historic opportunity to engineer its own Wirtschaftswunder—a productivity-driven economic transformation akin to post-war West Germany. While investment-led growth may offer quick wins, the country’s long-term trajectory will depend on whether it can foster the forces of creative destruction that drive sustained economic progress. Understanding what has held back Ukraine’s economy over the past quarter-century provides crucial insights for designing effective reconstruction policies. To identify what has constrained Ukraine’s economic dynamism, the authors analyze comprehensive firm-level data spanning nearly the entire universe of Ukrainian enterprises from 2002-2024. They combine financial statements data with detailed information on foreign direct investment flows and the complete registry of state-owned enterprises, enabling unprecedented insight into both private and public sector dynamics. The empirical analysis reveals troubling patterns of declining business dynamism: •

Ukrainian firms shifted from “up or out” to “flat and stay” dynamics. In 2002-2007, new firms exhibited vigorous growth patterns mirroring the United States, with surviving firms expanding rapidly. Between 2008-2013, business dynamics flat lined, resembling stagnation seen

Figure 1 · State Owned Enterprises Crowd Out Private Businesses and are Overrepresented Among the Least State Owned Enterprises Crowd Out Private Businesses Productive BusinessesAmong the Least Productive Businesses and are Overrepresented 16% Share of Sales 2008-2013 2014-2019 12

1 Least Productive

3 Productivity Quintile

5 More Productive

Note: The figure plots the average share of sales accounted for by the state-owned enterprises Note: The figuresector plots the sales accounted for by the in the manufacturing as aaverage function share of theirofrelative labor productivities forstate-owned two time enterprises in the sector as a function of their relative labor productivities for periods; 2015–2018 and manufacturing 2019–2022.

two time periods; 2015–2018 and 2019–2022.

in Mexico. After 2014, the forces of creative destruction appear to have been choked, with young firms barely increasing in size over a decade—a pattern strikingly similar to India’s chronically sluggish business lifecycle. •

Market concentration rose dramatically while productivity stagnated. The four largest businesses in manufacturing sectors increased their market share from 49% in early 2000s to 53% in 2019. This occurred alongside declining productivity growth, from 15.2% annually in 2002-2013 to just 3.7% in 2014-2019.

creative destruction: the process of innovation and technological change where new products, processes, or business models replace older ones, leading to the dismantling of established industries and ways of life

bfi.uchicago.edu

113

•

State-owned enterprises hold a large share of sales, effectively crowding out private enterprise. Though overrepresented among laggard firms, their share in top productivity quintiles substantially declined over time.

10% Entry Rate 9.4%

Foreign investment from tax havens stifles competition. Industries receiving investment from offshore financial centers exhibit 27% lower entry rates, suggesting that recycled domestic capital undermines business dynamism rather than bringing new technology and expertise.

The data reveal a troubling pattern: productive firms struggle to grow while unproductive incumbents not only survive but expand their market share. To explain this reversal of competitive dynamics, the authors develop a theoretical framework embedding institutional capture into a Schumpeterian model of creative destruction. The authors build a quantitative model and estimate its parameters using key patterns in Ukrainian firm data from 2002-2013. The model’s core mechanism centers on “entrenched” incumbents who gain market control through regulatory capture rather than productivity improvements, competing against innovative “transformative” entrepreneurs. The structural estimation reveals that transformative firms risk losing a product line to entrenched incumbents through institutional — rather than technological — displacement roughly every 10 months, creating strong disincentives for innovation and growth. The quantitative results strongly validate the entrenchment mechanism: •

Figure 2 · Industries Receiving FDI from Tax-Havens Have Lower Industries Receiving FDI from Tax-Havens Have Lower Business Dynamism, Measured Here Through Entry Rates Business Dynamism, Measured Here Through Entry Rates

The model shows that institutional capture breaks down resource allocation. Higher entrenchment reduces the correlation between firm productivity and size, explaining the observed deterioration in how efficiently productive firms can expand and attract resources.

READ THE WORKING PAPER NO. 2025-91 · JULY 2025

Engineering Ukraine’s Wirtschaftswunder bfi.uchicago.edu/working-papers/engineering-ukraineswirtschaftswunder

6.9%

0 Industries Without Tax-Haven FDI

Industries WithTax-Haven FDI

This graph shows average entryrate rateininmanufacturing manufacturing industries Note: Note: This graph shows the the average entry industries depending on the depending on the type of foreign direct investment they receive between type of foreign direct investment they receive between 2002 and 2013. 2002 and 2013.

•

Entrenchment flattens firm life-cycle profiles and increases persistence of small firms. The model replicates Ukraine’s transition from dynamic “up or out” patterns to stagnant “flat and stay” equilibrium as institutional capture rises.

•

Policy effectiveness depends critically on institutional quality. R&D subsidies for incumbents can boost growth when entrenchment is low, but become progressively less effective as institutional capture rises. Entry subsidies show minimal impact regardless of institutional environment due to general equilibrium crowding-out effects.

The analysis yields a crucial insight for Ukraine’s post-war reconstruction: targeting firm types alone is insufficient. Even perfectly designed subsidies will have limited aggregate impact unless accompanied by reforms that discipline entrenched incumbents and restore competitive market dynamics. This mirrors the priorities of West Germany’s Wirtschaftswunder, where dismantling industrial cartels and promoting competition formed the bedrock of economic policy. Ukraine’s transformation requires similar emphasis on institutional reform rather than simply directing resources toward particular sectors or firm categories.

ABOUT OUR SCHOLARS

Ufuk Akcigit

The Arnold C. Harberger Professor in Economics, Kenneth C. Griffin Department of Economics

Furkan Kilic

Postdoctoral Scholar, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

114

RESEARCH BRIEF • AUGUST 2025

Post-Roe Planning: The Effect of Dobbs v. Jackson on Contraceptive and Sterilization Choices Based on BFI Working Paper No. 2025-106, “Post-Roe Planning: The Effect of Dobbs v. Jackson on Contraceptive and Sterilization Choices,” by Yana Gallen and Daisy Lu, University of Chicago

The Supreme Court’s 2022 Dobbs ruling, which removed federal protections for abortion, led to an increase of 22% in the monthly rate of female sterilization procedures, an 18% increase in male sterilization procedures, and a 15% increase long-acting reversible contraceptives in states hostile to abortion compared to other states in the months immediately following the decision. When the Supreme Court issued its decision in Dobbs v. Jackson Women’s Health Organization (2022), reversing 1973’s Roe v. Wade that established federal abortion protections on contraceptive and sterilization decisions, one prominent question was how this would affect the behavior of residents living in states that are characterized as hostile (for example, 13 states had enacted trigger laws to ban or restrict abortion immediately if Roe were overturned) or non-hostile to abortion. Would people in various states alter their use of short-acting and longacting reversible contraceptives (SARC and LARC), or opt for sterilization? The authors study a longitudinal dataset covering inpatient and outpatient procedures, and prescription drug claims from May 1, 2021, to December 31, 2023. These data allow for the creation of a balanced panel of individuals

Figure 1 · States That Are Hostile and Nonhostile to Abortion Access, Post Dobbs States That Are Hostile and Nonhostile to Abortion Access, Post Dobbs

Treated States: Hostile to Abortion Access Control States: Protective of Abortion Access Missing Data

Note: The authors define states hostile to and protective of abortion access following data on abortion policies gathered theauthors Center for define Reproductive Rights; based to on and classification of states 2022, two months after Dobbs. Note: by The states hostile protective ofinabortion access following data on

abortion policies gathered by the Center for Reproductive Rights; based on classification of states in 2022, two months after Dobbs.

enrolled in employer-sponsored health insurance. Broadly representative of the privately insured U.S. population, these data also allow the authors to analyze demographic patterns in contraceptive

bfi.uchicago.edu

115

States hostile to abortion diverge from others in July 2022, immediately after the Dobbs decision. Both women and men are more likely to undergo sterilization procedures. and sterilization responses, including pre-Dobbs behavior. The authors find the following: •

States hostile to abortion diverge from others in July 2022, immediately after the Dobbs decision. Both women and men are more likely to undergo sterilization procedures and women are more likely to initiate LARC use.

•

These differences are short-lived. By September 2022, female sterilization procedures and LARC initiation converge to their pre-Dobbs differences across States. The effect on male sterilization fades by the end of the year.

•

However, there is one age subgroup in which effects do not fade: Young women (18-25)

READ THE WORKING PAPER NO. 2025-106 · AUGUST 2025

Post-Roe Planning: The Effect of Dobbs v. Jackson on Contraceptive and Sterilization Choices bfi.uchicago.edu/working-papers/post-roe-planning-theeffect-of-dobbs-v-jackson-on-contraceptive-andsterilization-choices

are more than 80% more likely to undergo sterilization procedures in 2023 as in 2021 in states hostile to abortion relative to other states. Young men are 40% more likely to undergo sterilization even more than one year after the initial ruling. •

For women, the effect on sterilizations is driven completely by women who were previously using SARCs, suggesting that these women were sexually active and may intend to avoid unplanned pregnancy.

•

Finally, the increase in sterilizations is largest among women with no child on their health plan.

This work contributes to the existing literature on this topic through its analysis of a privately insured, nationwide population, its findings regarding differences between short-term and long-term responses, and its discussion of such characteristics as age, past birth control use, and presence of children, tying treatment effects to the likelihood of unplanned pregnancy and future fertility.

ABOUT OUR SCHOLARS

Yana Gallen

Assistant Professor, Harris School of Public Policy

Daisy Lu

PhD Student, Harris School of Public Policy

Written by David Fettig • Designed by Maia Rabenold

116

RESEARCH BRIEF • AUGUST 2025

Reskilling and Resilience Based on BFI Working Paper No. 2025-99, “Reskilling and Resilience,” by Anders Humlum, University of Chicago; and Pernille Plato, University of Copenhagen

Effective reskilling programs prevent one case of depression for every three injured workers who participate, with equally large mental health benefits extending to their partners, while generating $3.20 in additional returns for every dollar invested beyond direct labor market gains. Workplace shocks often trigger a cascade of psychological and relationship effects that extend far beyond the individual worker. While research shows that effective reskilling can restore employment prospects, can it also protect the psychological wellbeing of entire households? In this paper, the authors examine how reskilling impacts workers’ mental health and partner outcomes.

2017. Their analysis focuses on 4,008 male craft workers who worked fulltime for at least three years before suffering a physical accident that reduced their earnings capacity by an average of 35%. The authors compare workers with and without access to higher education based on their pre-injury vocational specializations, and find the following:

The authors analyze comprehensive Danish register data linking work accidents, education records, labor market outcomes, healthcare utilization, and family relationships from 1995 to

•

Reskilling substantially reduces the mental burden of injury. While 10-15% of injured workers end up on antidepressants despite suffering only physical injuries, reskilling

Figure 1 · Prevents Reskilling Preventsfor Depression for Both Workers and Partners Reskilling Depression Both Workers and Partners B) Partner

Percentage Point Change in Workers Taking Antidepressants

A) Injured Worker Accident

15 10 5

Workers With Reskilling Access Workers Without Reskilling Access

0 -5 -4 to -5

-2 to -3

-1 0 1-4 Years Since Workplace Accident

5-7

8-10

20 15 10 5 0 -5 -4 to -5

-2 to -3

-1 0 1-4 Years Since Workplace Accident

5-7

8-10

Note: This figure shows how workplace accidents affect mental health for both injured workers and their partners, comparing those with access to reskilling programs versus those without.

Note: This figure showsthe how workplace accidents affect The mental health for both injured workers and their partners, comparing those with access to reskilling programs versus those without. The The vertical line marks timing of workplace accidents. vertical bars represent 90% confidence intervals. vertical line marks the timing of workplace accidents. The vertical bars represent 90% confidence intervals.

bfi.uchicago.edu

117

prevents one case of depression for every three participants. The mental health benefits are greatest while workers are still in school— before any income gains materialize— suggesting that current engagement or improved prospects play a key role in supporting well-being.

•

Reskilling enables healthier relationship dynamics by reducing partner burdens. When injured workers can reskill, separation rates increase by 3 percentage points while partner employment and mental health remain stable, suggesting reskilling frees both partners from otherwise difficult situations.

•

Partners experience equally large mental health benefits from workers’ reskilling opportunities. About 7% of partners begin taking antidepressants following their partner’s work accident. Remarkably, reskilling access prevents one case of partner depression for every three workers reskilled, with these effects growing stronger over time.

•

Reskilling’s mental health benefits extend to a range of “diseases of despair,” including indicators of alcoholism and opioid abuse. Workers without reskilling access are 5 percentage points more likely to continue using opioids long-term and experience higher rates of alcohol-related diagnoses. However, reskilling does not reduce “deaths of despair,” largely because the work accidents studied do not increase mortality in the first place.

The authors’ cost-benefit analysis reveals that mental health and partner spillovers add $224,000 in benefits per reskilled worker, equivalent to $3.20 in return for every dollar invested. Together, the mental health and partner benefits add 83% to the direct labor earnings gains from reskilling, fundamentally changing the economic case for workforce retraining programs by demonstrating that the full value extends well beyond the worker’s own employment outcomes.

•

Without reskilling access, partners remain loyal but suffer economically and psychologically. Partners of injured workers without reskilling opportunities experience a 5-percentage point decline in labor market attachment yet are 5 percentage points more likely to remain in the relationship, suggesting they become trapped in burdensome caregiving roles.

READ THE WORKING PAPER NO. 2025-99 · AUGUST 2025

Reskilling and Resilience bfi.uchicago.edu/working-papers/reskilling-and-resilience

For policymakers designing workforce development programs, these results highlight the importance of considering family-wide impacts when evaluating interventions. Reskilling emerges as a powerful tool for addressing not just unemployment and disability dependence, but also the broader “diseases of despair” that accompany economic disruption. As many advanced economies grapple with rising social disparities and persistent non-employment, effective reskilling programs offer a comprehensive defense against the psychological and social costs of career setbacks—protecting vulnerable workers while preserving the wellbeing of their families.

ABOUT OUR SCHOLAR

Anders Humlum

Assistant Professor of Economics and Fujimori/Mou Faculty Scholar, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

118

RESEARCH BRIEF • AUGUST 2025

The Personalist Penalty: Varieties of Autocracy and Economic Growth Based on BFI Working Paper No. 2025-95, “The Personalist Penalty: Varieties of Autocracy and Economic Growth,” by Christopher Blattman and Scott Gehlbach, UChicago; and Zeyang Yu, Princeton

To the extent that autocracies underperform economically, this tends to be concentrated in personalist regimes, where power is highly concentrated. In contrast, institutionalized autocracies generally perform as well as democracies. The “personalist penalty,” when it emerges, appears to be driven by some combination of lower private investment, worse public goods provision, and greater conflict. Decades of research on regime type and Figure 1 · Distribution of Growth Rates by Regime Type Distribution of Growth Rates by Regime Type development have debated whether autocracies grow more slowly than democracies. This paper highlights a critical oversight: not all autocracies Democracy are alike. Some are institutionalized, with power constrained by parties, legislatures, or militaries— like Mexico under the Institutional Revolutionary Party or Singapore under the People’s Action Party. Institutionalized Autocracy Others are personalist, where rulers wield unchecked authority—like Mobutu’s Zaire or Saddam’s Iraq. In this paper, the authors study the role of this Personalist Autocracy institutional variation in generating divergent economic outcomes. They extend the dynamic panel -30% -20 -10 0 10 20 30 design of Acemoglu et al. (2019), analyzing GDP GDP Growth Rate and regime-type data for up to 179 countries from Note: The vertical line represents the median andwhite the white mean. blue horizontal Note: The vertical line represents the median and the dot dot the the mean. TheThe blue 1960 to 2010. They estimate how GDP per capita horizontal spans the 25th to percentiles, 75th percentiles, and the horizontal the95th 5th percentiles. to barbar spans the 25th to 75th and the horizontal line theline 5th to 95th percentiles. growth differs across democracies, institutionalized autocracies, and personalist autocracies using constraints, and Geddes, Wright and Frantz’s eight different autocracy classifications (including categorical classification of personalist regimes—as Freedom House, Polity, and V-Dem) paired with six well as broader institutional measures like Polity’s distinct measures of personalism. executive constraints, V-Dem’s presidentialism index, and Henisz’s veto players measure. These personalism measures include direct indicators of power concentration—such as Gandhi and Sumner’s measure of control over political offices and freedom from military/party

To assess causal effects, the authors employ a dynamic two-way fixed effects model controlling for country and year fixed effects, plus four lags

bfi.uchicago.edu

119

of GDP per capita to account for pre-existing income dynamics and the tendency of political transitions to accompany economic downturns. This strategy helps isolate the impact of regime structure from confounding factors such as economic shocks or endogenous transitions. The authors find the following: •

Whether a personalist penalty emerges depends on the specific measure of personalism used, but when it does appear, the effect is typically concentrated in personalist dictatorships rather than being uniform across all autocracies.

•

The growth performance of institutionalized dictatorships shows no consistent pattern of underperforming relative to democracies across most specifications.

•

The mechanisms behind any personalist penalty appear to involve some combination of lower total factor productivity, reduced private investment, worse public goods provision, and greater conflict, though the particular relationships vary depending on the measure of personalism employed.

•

These general patterns hold across different GDP data series, various personalism indicators, and alternative sample restrictions, including excluding extreme growth episodes and planned economies.

READ THE WORKING PAPER NO. 2025-95 · JULY 2025

The Personalist Penalty: Varieties of Autocracy and Economic Growth bfi.uchicago.edu/working-papers/the-personalist-penaltyvarieties-of-autocracy-and-economic-growth

Not all autocracies threaten economic development equally. Democracies and institutionalized autocracies may support growth through credible institutions, policy stability, and checks on executive power. But where power is highly concentrated—particularly in regimes like Putin’s Russia, where institutional constraints have been systematically dismantled—economic outcomes may suffer significantly. Recognizing potential personalist penalties is vital not only for forecasting development trajectories, but also for shaping diplomacy, aid strategies, and institutional reform efforts. The findings suggest that international engagement should focus on strengthening institutional constraints on executive power rather than simply promoting democratic transitions. For global stability and development, the critical distinction may not be between democracy and autocracy, but between institutionalized and personalized rule.

ABOUT OUR SCHOLARS

Christopher Blattman

Ramalee E. Pearson Professor of Global Conflict Studies, Harris School of Public Policy

Scott Gehlbach

Elise and Jack Lipsey Professor, Department of Political Science, Harris School of Public Policy, and the College

Written by Abby Hiller • Designed by Maia Rabenold

120

RESEARCH BRIEF • AUGUST 2025

Trust in Banks and Borrower Behavior: Evidence from Supervisory Actions and Local Information Quality Based on BFI Working Paper No. 2025-102, “Trust in Banks and Borrower Behavior: Evidence from Supervisory Actions and Local Information Quality,” by Samuel Chang, University of Chicago; Rimmy E. Tomy, University of Chicago; and Jizhou Wang, University of Chicago

Trust plays a significant role in borrowers’ decisions to transact with banks, with higherquality borrowers taking the lead in avoiding transactions with banks under enforcement. Americans’ trust in the banking sector has collapsed this century, with the percentage of US adults with high confidence in banks tumbling from 53% in 2004 to 10% in 2023, following the Financial Crisis of 2007-09 and the Silicon Valley Bank failure failure. Further, 56% believe that the government does not do enough to regulate financial institutions. More than just a popularity poll, these numbers have implications for whether and how retail customers choose to work with banks. In this new work, the authors explore one key question that is largely ignored in the literature: Has the decline in trust influenced customers’ decisions to borrow from banks?

To begin, the authors offer the following definition of trust, based on prior literature: “the expectation that another person (or institution) will perform actions that are beneficial, or at least not detrimental, to us regardless of our capacity to monitor those actions.” Given this definition, trust in the banking industry can be compromised through concerns about bankers’ motives or banks’ operational reliability. To the question at hand, given the complexity of a loan agreement and the degree to which customers likewise must rely on banks’ integrity, distrustful borrowers may be disinclined to borrow.

Financial Crisis of 2007-09: Largely attributable to a housing market bubble driven by subprime mortgages, the Financial Crisis of 2007-2009 was a period of severe economic downturn and financial market stress that originated in the United States and quickly spread globally. It was the most significant economic downturn since the Great Depression of the 1930s, earning it the moniker, “Great Recession.” Silicon Valley Bank failure: The Silicon Valley Bank (SVB) crisis was the third-largest US bank failure in history, occurring in March 2023 after a run on deposits led to its seizure by regulators. The crisis was triggered by the bank’s large portfolio of long-term, low-yield securities that lost significant value when the Federal Reserve raised interest rates to control inflation. This, combined with a downturn in the tech sector and widespread panic, caused customers to rapidly withdraw funds, prompting the bank to sell assets at a loss, and ultimately leading to its collapse. retail customers: In banking, retail customers are individual consumers, rather than businesses, who receive financial services from a bank, including access to credit and borrowing, deposit opportunities, and money management advice.

bfi.uchicago.edu

121

Geographic Distribution of Enforcement Decision and Order and Non-EDO Auto Loans Figure (EDO) 1 · Geographic Distribution of Enforcement Decision and Order (EDO) and Non-EDO Auto Loans

EDO Bank Non-EDO Bank Note: This figure shows the geographic distribution of auto loans by the lending bank’s EDO status across the contiguous United States. When overlapping, EDO loans are plotted over non-EDO loans. 7.5% of loans are issued by EDO banks.

Note: This figure shows the geographic distribution of auto loans by the lending bank’s EDO status across the contiguous United States. When overlapping, EDO Measures of are trustplotted in banking generally surveyobscure the actual cost of financing and inflate loans overare non-EDO loans. 7.5% of loans are issued by EDO banks. based and tend to be regional or economy-wide expenses, making it difficult for borrowers to rather than bank-specific and so describe notions identify the full cost of the loan. Further, loan of generalized rather than personalized trust providers may overcharge for add-on products, (i.e., trust in a particular institution). The authors wrongfully repossess vehicles, and mislead address this gap through a novel approach that customers about loan payments, all stressing the allows them to study enforcement actions (also importance of borrowers’ trust in banks in the auto referred to as enforcement decisions and orders, loan market. The authors find the following: or EDOs) EDOs issued by US bank supervisors against • New loans originated following an financial institutions. Importantly, EDO banks suffer enforcement action are of relatively lower significant reputational damage, as enforcement quality and become delinquent 28% sooner actions are publicized in the local news media and than those originated by control banks. The on regulators’ websites. This allows the authors results are concentrated in the first year of to exploit the coverage and quality of local the enforcement action, suggesting that information to study the effect of declining trust on consumers react to news of the EDO. retail borrowers’ financial decisions. • This downward shift in the quality of Specifically, the authors use granular data on auto originated loans and new borrowers while loans from a credit reporting agency, which links an enforcement action is open suggests that borrowers and lenders, including on a geographic higher-quality borrowers avoid transacting basis, allowing the authors to examine geographic with enforced banks. Higher-quality variation in trust for the same EDO. Auto loan borrowers are more likely to react to changes pricing can be complex as rates vary based on the in trust because they have more options, borrower’s credit, loan structure, and vehicle type, including access to multiple sources of forcing borrowers to rely on banks to interpret credit, and are likely to be more sensitive to loan terms. Promotional offers and add-ons often perceived changes in service quality. Enforcement decisions and orders, or EDOs: Directives issued by regulatory agencies to force a financial institution or individual to take corrective actions to comply with laws and regulations, resolve violations, or prevent unsafe practices.

Geographic Distribution Figure 2 · Geographic Distribution of News Desertsof News Deserts

News Desert Not News Desert Note: This figure shows the geographic distribution of news deserts across the contiguous United States.

Note: This figure shows the geographic distribution of news deserts across the

United Thecontiguous authors provide theStates. following evidence that trust is at the heart of these findings: •

•

Survey data reveal that enforcement actions are associated with significant declines in trust in banks and bankers. The shift toward lowerquality borrowers during enforcement actions only occurs in regions that witness a decline in trust based on the survey-based measure. The authors quantify negative and positive sentiments, including trust, from local news articles discussing bank enforcement actions.

READ THE WORKING PAPER NO. 2025-102 · AUGUST 2025

Trust in Banks and Borrower Behavior: Evidence from Supervisory Actions and Local Information Quality bfi.uchicago.edu/working-papers/trust-in-banks-andborrower-behavior-evidence-from-supervisory-actionsand-local-information-quality

•

Finally, the authors use variations in the quality of local information access. Their results do not hold in regions with no local newspaper coverage. In such regions, consumers are unlikely to learn about enforcement actions, so their trust in banks should not be affected.

By extending the literature on how confidence in banks influences consumers’ decisions, including the influence of local media, this paper raises important questions for further research, especially regarding potential effects on local GDP, credit markets, or employment.

ABOUT OUR SCHOLARS

Samuel Chang

PhD Student, Chicago Booth

Rimmy E. Tomy

Associate Professor, Chicago Booth

Jizhou Wang

PhD Student, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

122

123

RESEARCH BRIEF • AUGUST 2025

Violent Backlash to Political Reform: Evidence from Anti-Jewish Pogroms in the 1905 Russian Revolution Based on BFI Working Paper No. 2025-96, “Violent Backlash to Political Reform: Evidence from Anti-Jewish Pogroms in the 1905 Russian Revolution,” by Paul Castañeda Dower, University of Wisconsin-Madison; Scott Gehlbach, University of Chicago; Dmitrii Kofanov, University of Pittsburgh; Steven Nafziger, Williams College; and Vladimir Novikov, Stanford University

Communities with larger Jewish populations experienced significantly less violent backlash following Russia’s 1905 October Manifesto, demonstrating that strength in numbers can deter violence when political reforms systematically alter power dynamics between ethnic groups. Local violence often accompanies momentous political change, as feelings of political threat intersect with preexisting prejudices to endanger groups popularly associated with reform. This dynamic played out dramatically in Imperial Russia during the 1905 Revolution. Following military defeats in the Russo-Japanese War and a crippling general strike, Tsar Nicholas II issued the October Manifesto in October 1905, ending centuries of absolute autocracy and promising civil and political rights. Many Russians celebrated in the streets, but conservative supporters of the monarchy blamed Jews—who were prominent in liberal and radical movements and comprised over 30% of political arrestees from 1903-1905— for forcing the Tsar’s hand. Within weeks, over 250 anti-Jewish pogroms erupted across the empire’s western territories, with pogromists shouting slogans like “There’s your freedom, there’s your constitution and revolution.” Within this general context of reform-induced backlash, what sort of communities proved most vulnerable to violent attacks? Existing scholarship offers competing predictions. Some theories emphasize that conflict should be more likely when ethnic groups are roughly equal in size—

Figure 1 · Population Share of Jews and Pogrom Incidence

Population Share of Jews and Pogrom Incidence A) Below-Median Share of Jews

B) Above-Median Share of Jews

60 Settlements with Pogroms October Manifesto Published 40

0 July 1904

February 1905

October 1905

May 1906

July 1904

February 1905

October 1905

May 1906

Note: Note:These Thesefigures figuresshow showthe theincidence incidenceofofpogroms pogromsbefore beforeand andafter afterthe thepublication publicationofof the October the October Manifesto, comparing settlements with above-median and below-median Manifesto, comparing settlements with above-median and below-median Jewish populations. Jewish populations.

that is, when society is polarized—because this maximizes the perceived threat groups pose to one another. Other perspectives suggest simpler relationships, with historians of this period arguing that violence was either more likely in settlements with large Jewish populations (due to their visibility and perceived threat) or less likely (because Jews could mount stronger resistance or had greater political influence).

bfi.uchicago.edu

124

In this paper, the authors study how community demographics shape violent backlash. They analyze comprehensive data on over 1,370 Jewish settlements in the Russian Empire’s Pale of Settlement from 1904-1906. The authors exploit the October Manifesto as a natural experiment, comparing pogrom incidence before and after the reform across communities with vastly different religious compositions, from settlements where Jews comprised less than 1% of residents to those that were nearly 100% Jewish. •

The sharp increase in pogroms following the October Manifesto was much smaller in settlements with larger Jewish communities. A one standard deviation increase in Jewish population share (26 percentage points) reduced the differential probability of a pogrom by more than one-third of the mean pogrom incidence rate.

•

In contrast, religious polarization showed no systematic relationship with violence. Unlike Jewish population share, measures of ethnic polarization, which capture the probability that two randomly selected residents belong to different religious groups, had no robust relationship with pogrom incidence after the October Manifesto.

•

The pattern held across numerous robustness checks. The core finding persisted when controlling for economic shocks, spatial diffusion, industrial composition, prior pogrom history, proximity to transportation networks, police presence, and military mobilization patterns.

To explain these results, the authors develop a theoretical model building on established conflict theory. The key insight is that political reforms like the October Manifesto create systematic winners

READ THE WORKING PAPER NO. 2025-96 · MAY 2025

Violent Backlash to Political Reform: Evidence from Anti-Jewish Pogroms in the 1905 Russian Revolution

and losers, unlike typical models that focus on random local factors. In this context, only the group that loses from reform, conservative supporters of the old regime, decides whether to use violence to reverse the changes. Their willingness to fight depends on whether they expect to win against organized resistance. The model reveals the following: •

Violent backlash decreases as the beneficiary group grows larger because the disadvantaged group faces higher expected costs when confronting a relatively larger opponent capable of effective resistance. This dynamic distinguishes reform-induced violence from symmetric ethnic conflicts, where polarization typically matters most for determining conflict likelihood.

This research provides new insights into a fundamental question about political transitions: under what conditions does reform trigger violent backlash rather than peaceful adaptation? The findings suggest that while expanding political rights is essential for justice and development, the safety of reform beneficiaries depends critically on their local demographic strength. Minority communities with insufficient numbers to organize effective resistance face the greatest vulnerability during periods of political change. The implications extend to contemporary debates about democratization, constitutional reform, and minority protection. Understanding how demographic composition shapes responses to political change could help policymakers design transitions and security arrangements that minimize risks of violent backlash while preserving the essential goals of expanding political inclusion and civil rights.

ABOUT OUR SCHOLAR

Scott Gehlbach

Elise and Jack Lipsey Professor, Department of Political Science, Harris School of Public Policy, and the College

bfi.uchicago.edu/working-papers/violent-backlash-topolitical-reform-evidence-from-anti-jewish-pogroms-in-the1905-russian-revolution

Written by Abby Hiller • Designed by Maia Rabenold

125

RESEARCH BRIEF • AUGUST 2025

What Drives Educational Technology Adoption in Classrooms Serving Young Children? Evidence from Two Experiments Based on BFI Working Paper No. 2025-103, “What Drives Educational Technology Adoption in Classrooms Serving Young Children? Evidence from Two Experiments,” by Daniela Bresciani, University of Chicago; Ariel Kalil, University of Chicago; Michelle Michelini, University of Chicago; and Rohen Shah, Yale University

This work reveals that educators can be influenced by research evidence when considering adoption of educational technology, but the results are mixed and suggest differential effects of an information intervention highlighting the effectiveness of math apps on pre-k teachers, elementary teachers, and pre-k leaders respectively. For teachers, the desire to adopt educational apps in their classrooms to improve student learning is challenged by the increasing number of available apps and by the difficulty of judging their effectiveness. For school leaders, questions surround weighing the cost-benefit of the financial and administrative burden required to adopt new apps vs. the return on investment to student learning. In a world of many choices and limited review time, what influences teacher and principal decisions? To address these questions, the authors use two survey experiments to examine whether an app’s adoption is affected by two factors: popularity and efficacy. The first experiment was conducted online from April to June 2024 with teachers from 11 school districts serving about 161,000 preschool and elementary students. The final sample included 1,104 teachers, including 289 pre-k teachers, 206 kindergarten teachers, and about 150 teachers from each of grades 1-4. The experiment was part of a 10-minute survey completed by teachers, which also asked about their familiarity with, and barriers to, recommending digital math games.

Figure 1 · School Leader Willingness to Pay for Digital and Analog Materials School Leader Willingness to Pay for Digital and Analog Materials $4.5 Per Student $4.10 $3.68 3

Video Treatment Control

$3.79

$3.16 $2.29

$2.43

$2.41

$2.20

1.5

Digital Game

Physical Game

Digital Worksheet

Physical Worksheet

Note: Regarding the authors’ second experiment with school leaders, this figure reveals that

Note: Regarding the authors’ secondthe experiment school leaders, thisworksheets figure reveals that the treatment video seems to decrease likelihood ofwith recommending physical and treatment increase thevideo willingness toto pay for digitalthe games. the seems decrease likelihood of recommending physical worksheets and increase the willingness to pay for digital games.

The authors use a within-subjects design where teachers view short descriptions of fictional math learning apps and then rate how likely they are to recommend the app to students for home use. The four app descriptions are followed by either 1) a statement about research evidence showing the app’s effectiveness, 2) a statement that the app is popular among teachers, 3) both statements

within-subjects design: an experimental design wherein each participant is exposed to all levels of the independent variable (also called treatments or conditions). Essentially, every participant serves as their own control, allowing for direct comparison of their responses across different conditions.

bfi.uchicago.edu

126

on efficacy and popularity, or 4) no additional information. The authors find the following: •

The likelihood of a teacher recommending a math app increases by 0.24 standard deviations (SD) when they are informed that there is research evidence that the app improves math skills, and by 0.21 SD when teachers are told that the app is popular among educators.

•

Telling teachers about both the app’s popularity and its efficacy results in a 0.30 SD increase.

•

The treatment effect on pre-k teachers is half as strong as k-4 teachers, which could indicate that preschool teachers’ opinions are less easily swayed than those of elementary school teachers. There may also be evidence of the “bandwagon” effect, as teachers are equally influenced by peer popularity and research evidence.

In the second experiment, also conducted online between April and June 2024, pre-k school leaders were randomly assigned to either watch an informational video on research evidence supporting a digital game or view a control video describing analog learning materials without discussing research evidence. After the video,

pre-K leaders were asked about 1) their likelihood of endorsing a digital math game among other learning resources for pre-k students to use at home, from “Very Unlikely” to “Very Likely,” and 2) their willingness to pay for a digital math game for each pre-K student using school funds, on a scale of $0-$10. The authors find the following: •

There is no significant effect on the likelihood of recommending digital apps, and a marginally significant increase in their willingness to spend more on digital apps from their school budgets.

•

While these results might suggest that school leaders are not easily influenceable, the authors caution that this may reflect the experiment’s methodology, and that a larger sample size may reveal effects resembling those in the teacher survey experiment.

This paper is part of an extensive literature on what influences decision making in various contexts, and more research is necessary to confidently suggest policy outcomes. That said, this study reveals that teachers are influenced by research evidence, which highlights the importance of rigorously evaluating ed-tech products.

standard deviations: a statistical measure of the dispersion or variability of a set of data points from the mean, representing the average distance of each data point from the mean.

READ THE WORKING PAPER NO. 2025-103 · AUGUST 2025

What Drives Educational Technology Adoption in Classrooms Serving Young Children? Evidence from Two Experiments bfi.uchicago.edu/working-papers/what-drives-educationaltechnology-adoption-in-classrooms-serving-youngchildren-evidence-from-two-experiments

ABOUT OUR SCHOLARS

Daniela Bresciani

PhD Student, Harris School of Public Policy

Ariel Kalil

Daniel Levin Professor, Harris School of Public Policy; Director, Center for Human Potential and Public Policy; Co-Director, Behavioral Insights and Parenting Lab

Michelle Michelini

Executive Director, Behavioral Insights and Parenting Lab

Written by David Fettig • Designed by Maia Rabenold

127

RESEARCH BRIEF • SEPTEMBER 2025

A Tale of Two Transitions: Mobility Dynamics in China and Russia after Central Planning Based on BFI Working Paper No. 2025-110, “A Tale of Two Transitions: Mobility Dynamics in China and Russia after Central Planning,” by Kristina Butaeva, University of Chicago; Lian Chen, University of California Los Angeles; Steven Durlauf, University of Chicago; and Albert Park, Hong Kong University of Science and Technology (HKUST) and Asian Development Bank

Analysis of intergenerational mobility during China and Russia’s transitions from central planning reveals that China’s higher educational mobility was largely driven by structural changes. Russia demonstrates greater steady-state educational mobility once transitional dynamics are accounted for, while both countries exhibit similar occupational mobility. The end of central planning and emergence of market economies in China and Russia represents one of the major economic transformations of the last century. While the economic effects of these transitions have been extensively studied, comparatively less attention has been paid to how they affected intergenerational mobility. In this paper, the authors add to this literature by studying intergenerational mobility—the ability of children to achieve different socioeconomic outcomes than their parents. Specifically, the parents in this study were born roughly 1950-70 and grew up under central planning, while their children were born 19781997 during the market transitions. There are important distinctions between these transitions: the Russian transition (beginning 1991) involved regime collapse, while the Chinese transition (beginning 1978) was implemented by the same government. Second, Russia pursued shock therapy while China implemented gradual reforms. Finally, the countries started from different development stages—Russia was already industrialized and urban, while China was mostly poor and agrarian.

Figure 1 · Dynamics of Overall, Structural, and Exchange Educational Mobility Dynamics of Overall, Structural, and Exchange Educational Mobility 60% Probability of Changing Educational Class 50 40 30

Structural Mobility

China, Father-to-Child China, Mother-to-Child Russia, Father-to-Child Russia, Mother-to-Child Overall Mobility Exchange Mobility

20 10 0

3 4 5 Generations into the Future

Note: This This figure plotsplots overall mobility (total (total probability childrenchildren have different education Note: figure overall mobility probability have different education than parents) with solid lines, exchange mobility (mobility after economic transitions stabilize) than parents) with solid lines, exchange mobility (mobility after economic transitions with dashed lines, and structural mobility (mobility due to temporary economic changes like stabilize) with dashed lines, and structural mobility due to temporary economic education expansion) with vertical distance between lines,(mobility across generations. Steady-state mobility is represented by the rightmost points where overall equals exchange changes like education expansion) with vertical distance between lines,mobility. across generations. Shaded areas represent 95% intervals. Steady-state mobility is confidence represented by the rightmost points where overall equals exchange

mobility. Shaded areas represent 95% confidence intervals.

The authors use data from the China Family Panel Studies (8,788 child-parent pairs) and the Russian Longitudinal Monitoring Survey (3,718 child-parent pairs) to analyze how educational and occupational status transmits from parents to children. They develop new measures that distinguish between different types of mobility: structural mobility (driven by changes like educational expansion) and exchange mobility (reflecting genuine movement

bfi.uchicago.edu

128

between social classes). The authors also measure “steady state mobility” to capture what mobility would look like once the transition period ends. They find the following: •

•

Both China and Russia exhibit very high levels of overall mobility during three decades of transition. For education, the probability of changing educational class for children is very high: 52-53% in China and 45-46% in Russia. These differences occur at different educational levels—mobility in China is driven by children who, unlike their parents, complete either high school or college, while Russian mobility is entirely due to increased college attendance. Similar results hold for occupational mobility, with overall rates closely aligned: 57-58% in China and 54-57% in Russia. However, the underlying sources differ significantly. Chinese occupational mobility is driven by movements out of agriculture, while Russian mobility is driven by shifts away from the manufacturing sector. These dramatic changes are largely due to structural rather than exchange mobility. Approximately 68-81% of individuals in China and 57-68% in Russia who transitioned out of their parental educational class did so because of the gap between parental and child educational distributions.

READ THE WORKING PAPER NO. 2025-110 · AUGUST 2025

A Tale of Two Transitions: Mobility Dynamics in China and Russia after Central Planning bfi.uchicago.edu/working-papers/a-tale-of-twotransitions-mobility-dynamics-in-china-and-russiaafter-central-planning

•

For occupational mobility, the structural component accounts for 60% of class changes in China. In Russia, structural mobility is responsible for 50% of class shifts in fatherto-child samples, but only 13% in mother-tochild samples—reflecting that women in the Soviet Union already held higher positions within the occupational structure.

•

After accounting for structural changes, individuals in Russia have a higher probability of moving out of their parental educational class at steady state (42%) compared to China (19% for father-to-child and 27% for mother-to-child samples). In contrast, occupational steady-state mobility is similar in both countries, with 50-55% probability of exiting the parental occupational class.

•

Comparing these results to the US, we find that steady state mobility in education is substantially higher in the US and Russia compared to China, but occupational steady state mobility is comparable in all three countries.. This suggests that China’s impressive current educational mobility may be temporary, declining as structural transitions complete.

These results demonstrate the importance of distinguishing between temporary structural changes and permanent mobility patterns when evaluating social mobility in transitioning economies.

ABOUT OUR SCHOLARS

Kristina Butaeva

Postdoctoral Scholar, Stone Center for Research on Wealth Inequality and Mobility, Harris School of Public Policy

Steven Durlauf

Frank P. Hixon Distinguished Service Professor; Director, Stone Center for Research on Wealth Inequality and Mobility, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

129

RESEARCH BRIEF • SEPTEMBER 2025

Chat2Learn: A Proof-of-Concept Evaluation of a Technology-Based Tool to Enhance Parent-Child Language Interaction Based on BFI Working Paper No. 2025-119, “Chat2Learn: A Proof-of-Concept Evaluation of a Technology-Based Tool to Enhance Parent-Child Language Interaction,” by Linxi Lu and Ariel Kalil, University of Chicago

Parent-child dyads were randomly assigned to use Chat2Learn, a low-cost, text-messaging tool that delivers open-ended conversation prompts, in a brief laboratory experiment. Parents spend less time inattentive or withdrawn during interaction, and produce higherquality language input and more open-ended questions. The language-rich conversations that parents have with their young children are crucial building blocks for development. Children who hear more diverse vocabulary, complex sentences, and open-ended questions from their parents typically develop stronger language skills that carry into school and beyond. However, families from different socioeconomic backgrounds show significant differences in the amount and quality of parentchild talk, contributing to persistent achievement gaps. Most interventions designed to address these disparities require intensive parent training or coaching, making them difficult to scale and sustain. This study tests a different approach. Rather than teaching parents why and how to talk with their children, Chat2Learn provides parents with ready-to use conversation starters delivered via text message. Grounded in Cognitive Load Theory, the tool aims to reduce the mental effort parents need to generate engaging topics, especially when managing competing demands of daily life. Each prompt includes open-ended questions about imaginary

Ask Jamie what he would like to have as a pet if he could choose anything at all?

scenarios paired with simple illustrations to capture children’s attention and spark decontextualized conversations that are particularly beneficial for language development. Using a randomized laboratory experiment with 63 parent-child dyads from low-income Chicago preschool programs, the authors compare families who had access to Chat2Learn during a 10-minute waiting period with those who did not. Parents were not told the study’s purpose and were simply asked to spend time with their child as they usually

bfi.uchicago.edu

130

would. The researchers analyze video recordings to measure parent engagement patterns and coded all conversations for quantity and quality of language use by both parents and children. They find the following: •

Parents using Chat2Learn spent significantly more time jointly engaged with their children, averaging 92% of the interaction in joint engagement compared to 78% for control families. This improvement was primarily driven by parents spending less time disengaged or withdrawn.

•

Chat2Learn enhanced both the quantity and quality of parental language input. Parents in the intervention produced more total words, used more diverse vocabulary, constructed longer and more complex utterances, and asked substantially more open-ended questions compared to control parents.

•

The tool also successfully shifted conversations from routine “here and now” topics toward more sophisticated decontextualized discussions about imaginary scenarios, hypothetical situations, and abstract concepts. Parents using Chat2Learn asked an average of 37 openended questions during the 10-minute session, far exceeding the 12 questions provided in the prompts themselves.

READ THE WORKING PAPER NO. 2025-119 · SEPTEMBER 2025

Chat2Learn: A Proof-of-Concept Evaluation of a Technology-Based Tool to Enhance Parent-Child Language Interaction bfi.uchicago.edu/working-papers/chat2learn-a-proof-ofconcept-evaluation-of-a-technology-based-tool-toenhance-parent-child-language-interaction

•

These improvements did not translate into immediate increases in children’s language output—children in both the control and intervention groups produced comparable amounts of talk during the brief interaction task. This may be because decontextualized conversations place higher cognitive demands on preschoolers and may require sustained parental support to elicit extended responses from young children.

This proof-of-concept study demonstrates that a simple, scalable intervention can meaningfully improve parent-child language interactions without requiring intensive training. By reducing the cognitive burden of generating conversation topics, Chat2Learn helps parents engage in exactly the types of rich, decontextualized talk that research shows benefits children’s language development. While longer-term studies are needed to assess sustained effects and impacts on children’s language skills, the findings suggest promise for technology-based tools that support parents with concrete conversational resources rather than abstract training in communication strategies.

ABOUT OUR SCHOLARS

Ariel Kalil

Daniel Levin Professor, Harris School of Public Policy; Director, Center for Human Potential and Public Policy; Co-Director, Behavioral Insights and Parenting Lab

Linxi Lu

Postdoctoral Fellow, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

131

RESEARCH BRIEF • SEPTEMBER 2025

Laboratories of Autocracy: Landscape of Central–Local Dynamics in China’s Policy Universe Based on BFI Working Paper No. 2025-121, “Laboratories of Autocracy: Landscape of Central– Local Dynamics in China’s Policy Universe,” by Kaicheng Luo, Massachusetts Institute of Technology; Shaoda Wang, University of Chicago; and David Y. Yang, Harvard University

China’s policymaking has historically been highly decentralized, with 82% of local policies originating as local initiatives, but since 2013 has become substantially more centralized as political incentives shifted from rewarding bottom-up innovation to strict enforcement of central policies. Top-down industrial policies are 18-22% less suitable for local conditions and significantly less effective, with the costs of centralization outweighing the benefits. A fundamental question in political economy concerns the appropriate level for making policy decisions. While top-down policy promotion may streamline adoption, internalize spillovers, and enhance efficiency, it often sacrifices the local suitability that bottom-up policy initiatives provide. This tension is especially relevant in governing large polities with high levels of regional heterogeneity. Despite its theoretical importance, studying the centralization of policymaking remains empirically challenging because it requires systematically tracing the origin and diffusion patterns of all policies across layers of government hierarchy. In this paper, the authors study the centralization of policymaking in China over the past two decades, and how it affects the local suitability of policies. They investigate two key questions: what share of local governments’ policy portfolios is shaped by the central government’s direct involvement, and whether the central government’s direct involvement undermines policy suitability and effectiveness at the local level.

To study these questions, the authors compile a dataset of 422 thousand central government policy documents and 3.3 million local government policy documents and work reports. From this corpus, they identify 115,679 distinct policies implemented from 2004 to 2020 and trace their origins and vertical and horizontal diffusion patterns. For the subset of industrial policies aimed at promoting industrial growth, they also measure policy suitability and effectiveness using data on industrial output, exports, and patent filings. They construct measures of policy-locality suitability based on pre-existing regional supply chains and private firms’ ex ante investment preferences to evaluate how well policies align with local conditions. The authors find the following: • China’s policymaking has historically been highly decentralized, with local bureaucrats playing crucial roles in both creating new policies and spreading them. Over the past two decades, 82% of the policies appearing

bfi.uchicago.edu

132

in local governments’ portfolios originated as local initiatives; of these, 74% diffused solely horizontally among local governments and never involved explicit central government action.

rate of top-down initiatives has nearly tripled, and local replication of central policy details has more than doubled.

• This shift likely resulted from changes in local bureaucrats’ career incentives. Qualitative accounts suggest that, after Xi Jinping assumed power in late 2012, authority was rapidly consolidated within the central government, suppressing local Figure 1: Centralization Trend in Policymaking policy initiatives and experimentation. At Figure 1 · Centralization Trend in Policymaking this time, political promotion was granted to bureaucrats who most actively implemented # of Prefectures A New Policy Reaches in First 3 Years top-down policies rather than those who pioneered new policies as observed before. 50

• Since 2013, policymaking has become substantially more centralized. The share of top-down policies in local governments’ portfolios has increased by 40%, the adoption

40 30

Central Policies

Local Policies

2004

2008

2012

2016

2020

% of Local Attention on Assigned Tasks 66% 64 62 60 58 56

2004

2008

2012

2016

2020

Similarity Between Local and Central Documents .22 .20 .18 .16 .14 .12 .10 2004

2008

2012

2016

2020

Note: This figure shows the growing centralization of policymaking after 2013. Panel A plots how widely central-government policies spread, by tracking the number of prefectures adopting each within three years. Panel B shows how much local governments focused on centrally assigned tasks, measured as the share of policies promoted by the center. Panel C reports the average similarity between the first central document on a topic and local follow-ups.

• Locally initiated and horizontally diffused policies tend to be associated with higher ex ante local suitability and better ex post economic outcomes. Focusing on industrial policies aimed at promoting sector-specific industrial growth and innovation, the authors find that those initiated or adopted by local governments without central involvement are better aligned with local conditions, as measured by pre-existing regional supply chains and private firms’ ex ante investment preferences. In contrast, top-down industrial policies initiated or explicitly adopted by the central government show systematically weaker alignment with local conditions. Moreover, industrial policies that are better matched to local conditions prove significantly more effective, on average, in achieving policy objectives — including increased industrial output, patenting, and exports — underscoring the costs of centralization. • Centralizing policymaking curbs strategic competition among local bureaucrats that otherwise impedes learning from peer jurisdictions. When policies are initiated and diffused locally without central government’s explicit involvement, local bureaucrats competing for the same political promotion opportunities may be reluctant to adopt policies from one another for fear of boosting their competitors’ credentials. Since local bureaucrats with similar promotion prospects are often posted to economically comparable localities, the authors find that such rivalry stifles the diffusion of local policy innovations, undermines local suitability among diffused policies, and dampens economic performance. Centralization alleviates these distortions, since local bureaucrats no longer worry about political competitors receiving credit when they implement top-down policies.

Note: This figure shows the growing centralization of policymaking after 2013. Panel A plots how widely central-government p tracking the number of prefectures adopting each within three years. Panel B shows how much local governments focused on tasks, measured as the share of policies promoted by the center. Panel C reports the average similarity between the first centr topic and local follow-ups.

133

Figure 2: Career Incentives and Policyvs. Innovation Figure 2 · Career Incentives and Policy Innovation Compliance vs. Compliance 2 Impact of Policy Innovation on Promotion Likelihood 2013 - Policy Centralization Begins

1.5

1 Impact of Policy Compliance on Promotion Likelihood

0.5

1 0 0.5 -0.5

0 -0.5 2005

2012

2018

-1

2005

2012

2018

Note: point represents the statistical relationship between a (left) bureaucrat's innovation compliance (right) behavior and position their likelihood Note: EachEach point represents the statistical relationship between a bureaucrat’s innovation compliance (right) behavior(left) and their likelihood of being promoted to a higher for a given year. ofblue being promoted to a effect, higher forbands a given The blue dots show the estimated effect, and the light blue bands show standard errors. The dots show the estimated andposition the light blue showyear. standard errors.

• The gains from centralization in promoting policy diffusion are outweighed by the losses from reduced policy suitability due to central intervention. Compared to the 2012 level of decentralization, the post-2013 centralization trend converted 2,562 prefecture-level industrial policies from bottom-up to topdown, with each top-down policy being 22% less suitable for local conditions than a bottom-up one. Back-of-the-envelope calculations suggest that the yearly cost of lowered policy suitability attributed to post2013 centralization is 580 billion RMB in industrial output, 437 billion in exports and 10,486 patent filings. Meanwhile, centralization mitigates competition-induced suitability loss by 0.007 for each political rival among economic neighbors, yielding yearly benefits of 121 billion RMB in industrial output, 91 billion RMB in exports, and 2,194 additional patent filings. The costs consistently exceed benefits by more than 400%.

READ THE WORKING PAPER NO. 2025-121 · SEPTEMBER 2025

Laboratories of Autocracy: Landscape of Central–Local Dynamics in China’s Policy Universe

This research demonstrates that under wellstructured political incentives, an autocracy can function as a vibrant laboratory for policy innovation—generating new, locally suited ideas—much like Justice Brandeis’s famous characterization of federalist systems as “laboratories of democracy.” However, China’s shift from decentralized to centralized policymaking since 2013 has come at significant economic cost. The findings provide empirical support for theoretical literature emphasizing the importance of decentralized information while also illustrating the distortions that can arise from decentralized regional competition. As governments around the world grapple with challenges that demand both coordination and customization—from climate mitigation to education policy to industrial strategy—understanding the optimal hierarchical level for decision-making and the associated trade-offs becomes increasingly imperative.

ABOUT OUR SCHOLAR

Shaoda Wang

Assistant Professor, Harris School of Public Policy

bfi.uchicago.edu/working-papers/laboratories-of-autocracylandscape-of-central-local-dynamics-in-chinas-policyuniverse

Written by Abby Hiller • Designed by Maia Rabenold

134

RESEARCH BRIEF • SEPTEMBER 2025

Navigating the College Affordability Crisis: Insights from College Savings Accounts Based on BFI Working Paper, “Navigating the College Affordability Crisis: Insights from College Savings Accounts,” by Guglielmo Briscese, John A. List, and Sabrina Liu, University of Chicago

While 61% of parents could save enough to cover half of their child’s college costs, they still believe their savings would be meaningless, suggesting behavioral barriers may be as important as financial ones for college affordability. With higher education costs consistently Figure 1 · Differences in Financial Literacy Between 529 Owners and IL Parents outpacing inflation and with public funding Differences in Financial Literacy Between 529 Owners and IL Parents declining, college affordability has become 100% of Respondents 8% a critical barrier to economic mobility for middle- and low-income families. While College 24 43% Savings Accounts (CSAs), or 529 plans, offer 75 tax-advantaged vehicles for college savings, Financial Literacy their adoption patterns and educational impacts Score (Out of 3) 3 remain poorly understood. 50 2

1 36 In this paper, the authors conduct the first 0 large-scale analysis of College Savings 25 Account participation and effectiveness using 7 27 comprehensive administrative data. They draw 14 on administrative data from over 900,000 Illinois 0 529 accounts (2000–2023) linked to educational 529 Account Owners IL Parents outcomes, plus complementary surveys of Note: This Note: figureThis shows share of 529 owners vsvsthe ofIllinois Illinois parents figurethe shows the share of 529 owners theshare share of parents who who scored between scored between 3 survey on three survey on financial literacy as measured account owners and parents. Their research 0 and 30onand three questions onquestions financial literacy as measured by (Lusardi and Mitchell, 2023). by (Lusardi and Mitchell, 2023). reveals the following patterns: • Financial literacy emerges as a key barrier: • CSA participation remains concentrated 61% of parents who could save enough to among higher-income, more educated cover half of future college costs still perceive families although adoption has expanded to their potential savings as meaningless. every ZIP code in Illinois.

bfi.uchicago.edu

135

of Savings Own Savings to Cover Future Costs of College FigureParents’ 2 · Parents’Underestimation Underestimation of Own to Cover Child’s Future Child’s Costs of College 80% of Respondents Believe Their Savings Would Not Cover a Meaningful Amount 76% 66

0-10%

11-20

21-30

31-50

More than 50

Expected Percentage Covered (Average Between Cheapest and Most Expensive Schools) Note: This figure shows the share of Illinois parents who believe their savings would not cover a meaningful amount, broken down by the projected share that their self-reported possible savings Note: This figure share of Illinois parents who believe their savings would not cover a amount may be able to cover given shows the future the cost of their child’s college, after accounting for inflation.

•

meaningful amount, broken down by the projected share that their self-reported possible savings amount may be able to cover given the future cost of their child’s college, after accounting for inflation.

Among participants, higher savings are strongly correlated with better educational outcomes, including four-year college enrollment, attendance at selective institutions, and pursuit of post-graduate degrees.

These findings suggest that targeted interventions addressing financial literacy gaps

READ THE WORKING PAPER NO. 2025-107 · MAY 2025

Navigating the College Affordability Crisis: Insights from College Savings Accounts bfi.uchicago.edu/working-papers/navigating-the-collegeaffordability-crisis-insights-from-college-savings-accounts

and misperceptions about modest savings could significantly expand CSA effectiveness as a tool for educational equity. Beyond state-level 529 program optimization, the findings also suggest promising avenues for federal policy coordination and institutional innovation.

ABOUT OUR SCHOLARS

Guglielmo Briscese

Postdoctoral Scholar, Harris School of Public Policy

John A. List

Kenneth C. Griffin Distinguished Service Professor in Economics and the College, Kenneth C. Griffin Department of Economics; Director, Becker Friedman Institute for Economics

Sabrina Liu

Senior Research Analyst, Inclusive Economy Lab

Written by Abby Hiller • Designed by Maia Rabenold

136

RESEARCH BRIEF • SEPTEMBER 2025

Partial Language Acquisition: The Impact of Conformity Based on BFI Working Paper No. 2025-109, “Partial Language Acquisition: The Impact of Conformity,” by William A. Brock, University of Wisconsin–Madison; Bo Chen, Southern Methodist University; Steven N. Durlauf, University of Chicago; and Shlomo Weber, Southern Methodist University

Peer pressure within minority communities creates complex, non-linear effects on language learning that can make small policy changes produce dramatically unpredictable shifts in community-wide learning patterns. When individuals can choose between full fluency, basic skills, or no learning, conformity pressures generate U-shaped and bell-shaped relationships that traditional cost-benefit models cannot capture. For immigrants and minority communities navigating dominant-language societies, the decision to learn a majority language involves weighing the costs and benefits of language acquisition against social pressures within their own communities. While previous research has focused on binary choices—either learning a language fully or not at all—many immigrants achieve intermediate proficiency levels. In this paper, the authors examine how peer pressure and conformity within minority groups influence patterns of majority language acquisition when individuals can choose among three options: full learning, partial learning, or no learning of the majority language.

Figure 1 · Language Acquisition Patterns Under Conformity Pressure Language Acquisition Patterns Under Conformity Pressure

The authors develop a game-theoretical model analyzing language acquisition decisions in an economy with one majority group and multiple minority groups. The key innovation is incorporating a conformity factor, or, peer pressure costs that increase as an individual’s language choice deviates from their community’s average choice. Minority agents face different learning costs based on their linguistic aptitude (distributed uniformly from low to high cost) and receive communicative benefits

from interacting with majority group members and other minorities who know the language. The authors analyze how varying levels of conformity pressure affects language learning preferences, and find the following: •

Share Choosing Full Fluency

Share Choosing Partial Learning

Share Choosing No Learning

Small Cost for Partial Learning

Decreases then increases as peer pressure rises

Always increases as peer pressure rises

Always decreases as peer pressure rises

Intermediate Cost for Partial Learning

Always decreases as peer pressure rises (least popular choice)

Increases then decreases as peer pressure rises

Decreases then increases as peer pressure rises

Large Cost for Partial Learning

Always decreases as peer pressure rises

Either always increases or always decreases (depends on how many started learning fully)

Always increases as peer pressure rises

In a scenario where individuals can only choose to learn fully or not at all, the relationship between conformity and language learning is consistently monotonic:

bfi.uchicago.edu

137

if learning is more popular without peer pressure, greater social pressure drives more minority agents to learn the majority language, and vice versa. •

In contrast, in a scenario where individuals can choose partial learning, outcomes show non-monotonic relationships: the number of full learners, partial learners, and non-learners can follow U-shaped or bell-shaped patterns as conformity pressure increases.

•

When partial learning costs are small, the number of people choosing full fluency follows a U-shaped pattern as peer pressure increases— high when conformity is weak, low with moderate pressure, then high again with intense pressure, while the number choosing partial learning consistently increases with conformity.

•

When partial learning costs are intermediate, the numbers of partial learners and non-learners can follow curved patterns (bell-shaped or U-shaped) depending on which choice is most popular in the community, while the least popular choice consistently decreases.

•

When partial learning costs are large, the number choosing full fluency always decreases with conformity pressure and non-

READ THE WORKING PAPER NO. 2025-109 · AUGUST 2025

Partial Language Acquisition: The Impact of Conformity bfi.uchicago.edu/working-papers/partial-languageacquisition-the-impact-of-conformity

learners always increase, but partial learners can either increase or decrease depending on how many people were initially learning fully. •

This non-monotonicity can lead to complex dynamics, as even minor changes in language acquisition costs or communicative benefits can result in significant shifts in language acquisition patterns among minority agents.

These findings reveal that policy implications derived from traditional binary language acquisition models can be misleading when applied to real-world settings where multiple levels of proficiency are possible. Small adjustments in language program costs may unpredictably influence language acquisition patterns across minority groups due to conformity effects. Policymakers should recognize that social pressure within minority communities can either amplify or undermine the effectiveness of language programs in ways that simple cost-benefit analyses cannot predict. Effective language policies must account for peer dynamics and community norms, as interventions that ignore these social factors may produce counterintuitive results where increased support paradoxically reduces participation.

ABOUT OUR SCHOLAR

Steven Durlauf

Frank P. Hixon Distinguished Service Professor; Director, Stone Center for Research on Wealth Inequality and Mobility, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

138

RESEARCH BRIEF • SEPTEMBER 2025

Religion in Emerging and Developing Regions Based on BFI Working Paper No. 2025-111, “Religion in Emerging and Developing Regions,” by Sara Lowes, University of California, San Diego; Benjamin Marx, Boston University; and Eduardo Montero, UChicago Harris School of Public Policy

While economic development has coincided with religious decline in wealthy countries, emerging and developing nations show persistent and often increasing religiosity, with traditional beliefs coexisting alongside major world religions. This religious divergence challenges standard secularization theories and reveals that religious institutions are adapting to modernization rather than disappearing, creating competitive marketplaces where denominations adjust their services and doctrines to meet changing social and economic needs. A growing literature in economics studies the multifaceted relationship between religion and development. One of the most enduring debates in this field concerns whether economic development inevitably leads to religious decline. The traditional secularization hypothesis predicts that economic growth and modernization lead to declining religious belief and practice—a pattern observed in many Western societies.

In this paper, the authors challenge the secularization hypothesis. Rather than waning, they argue that religious beliefs, practices, and institutions may be adapting to the profound social and economic changes sweeping across the globe. In addition, the authors posit that non-Western societies may be leading some of the trends observed in terms of religious innovation worldwide, including the rise of highly decentralized religious movements which

Figure ImportanceofofReligion Religion& &Income Income (OECD Non-OECD) Figure11:· Importance (OECD vs.vs. Non-OECD) Importance of Religion in Life

Relationship Between Income and Religiosity

3.4 / 4.0 Scale

0 Coefficients on Income for Importance of Religion

3.2

-.01

Non-OECD Countries IMPORTANT

-.02

NOT IMPORTANT

-.03

3 2.8

OECD Countries

WEAK RELATIONSHIP

OECD Countries

STRONG NEGATIVE RELATIONSHIP

-.04

2.6

-.05

2.4 ‘89–’93

Non-OECD Countries

‘94–’98

‘99–’04

‘05–’09

Survey Years

‘10–’14

‘17–’22

‘89–’93

‘94–’98

‘99–’04

‘05–’09

‘10–’14

‘17–’22

Survey Years

Note: The left chartshows showsaverage averageresponses responsesto to“How "Howimportant important is is religion religion in comparing Note: The left chart in your your life?" life?”on onaascale scalefrom from1 1(not (notatatall allimportant) important)toto4 4(very (veryimportant), important), comparing wealthy countries (OECD) with wealthy countries (OECD) with developing countries (non-OECD). The right chart shows how the relationship between personal income and religiosity has developing countries (non-OECD). The right chart shows how the relationship between personal income and religiosity has changed over time within countries. The analysis controls for country-level changed over time within countries. The analysis controls for country-level differences to isolate the income effect. Higher values on the right chart indicate a differences to isolaterelationship the incomebetween effect. Higher on the right chart indicate a stronger negativewhile relationship between income and little religiosity (wealthier people being less religious), while values stronger negative incomevalues and religiosity (wealthier people being less religious), values closer to zero indicate relationship between income andlittle religiosity. closer to zero indicate relationship between income and religiosity.

bfi.uchicago.edu

139

coexist and compete with well-established major denominations and secular institutions. The authors use empirical evidence to support these hypotheses while reviewing how recent scholarship explains the persistence and adaptation of religion in developing contexts. The authors analyze six waves of World Values Survey data covering 433,181 observations from 108 countries between 1989 and 2022, complemented by Pew Research Center data on religious beliefs in sub-Saharan Africa. They examine trends in religious importance, service attendance, and beliefs across different regions and income levels, distinguishing between OECD (high-income) and non-OECD (developing) countries to document the religious divergence. The analysis incorporates both major world religions like Christianity and Islam, as well as traditional belief systems that often coexist with formal religions in developing regions. They begin by documenting the following: • Respondents in the MENA region, sub-Saharan Africa, and South/Southeast Asia are most likely to report that religion plays an important role in their life and to frequently attend religious services. The self-reported importance of religion and frequency of religious service attendance have not declined in these regions since the 1990s, while Latin America displays stable trends and former communist countries are experiencing a resurgence of religion. • Unlike survey respondents in OECD countries, respondents from non-OECD countries report a sharp increase in the self-reported importance of religion in life: In 2017-2022, 88% of respondents in non-OECD countries reported believing in God compared to 67% in OECD countries, 76% believed in heaven compared to 53%, 56% prayed daily compared to 30%, and 36% attended religious services weekly compared to 20%. • The magnitude of correlation between income and religiosity has been declining over time, with emerging and developing countries (EDCs) leading this trend. Since the 2010s, the correlation between religiosity and income (conditional on country fixed effects) is close to zero in EDCs, while OECD countries are converging toward this pattern.

• Traditional belief systems coexist with more formal religions in many EDCs. Using data from sub-Saharan Africa, 43% of individuals report believing in witchcraft on average, 42% believe in the evil eye, 48% believe in evil spirits, 19% own traditional sacred objects, and 15% participate in ceremonies to celebrate ancestors. • Even where over 90% of respondents report affiliation to a major transcendental religion (e.g., Islam or Christianity), a large majority also believe in witchcraft as a causal logic for misfortune. This demonstrates widespread religious syncretism where individuals hold multiple religious beliefs rather than transcendental religions displacing traditional practices.

What explains these results? The authors review recent literature to understand what drives continued religious adherence in emerging and developing countries. They identify two broad sets of mechanisms that explain why modernization has not uniformly reduced religiosity in these settings: economic insecurity and mutual aid, and identity and adaptation during cultural transitions. They draw the following conclusions: • The literature identifies religion as a source of security. Research shows that higher GDP has not eliminated income volatility and financial insecurity in many EDCs, which are characterized by exposure to shocks, conflict, and weak institutions. Studies demonstrate that this drives people to seek religion for both risk management and psychological comfort. Research has shown that negative economic shocks, climatic disasters, and conflict exposure

The authors identify two broad sets of mechanisms that explain why modernization has not uniformly reduced religiosity in these settings: economic insecurity and mutual aid, and identity and adaptation during cultural transitions.

140

increase religious adherence. The literature reveals that religious networks provide both spiritual comfort and tangible mutual aid where formal welfare systems are weak. Experimental evidence from Ghana demonstrates that formal insurance can crowd out religious giving, revealing systematic substitution between formal coverage and what researchers term “God insurance.” • The literature also shows that religion serves as a source of cultural identity during transitions. Research demonstrates that religion serves as an anchor for identity during periods of rapid social and economic change, including urbanization, migration, and political upheaval. Studies document that religious communities provide not only spiritual comfort but also ready-made social networks, job information, and informal insurance for urban migrants. The literature from Latin America, Egypt, and South Asia demonstrates how religious movements grow during periods of economic dislocation and cultural transition, offering migrants supportive identities and helping individuals cope with unfulfilled aspirations. • Finally, the literature documents religious institutions as competing service providers. Research shows that in many EDCs, religious

READ THE WORKING PAPER NO. 2025-111 · SEPTEMBER 2025

Religion in Emerging and Developing Regions bfi.uchicago.edu/working-papers/religion-in-emerging-anddeveloping-regions

institutions operate as powerful competitors to secular state institutions, delivering education, healthcare, and financial support where government capacity is lacking. Studies document that this creates enduring “dual” institutional systems where religious organizations often command superior legitimacy and trust compared to state agencies. The literature demonstrates that religious institutions also compete like firms in a marketplace, adapting their doctrines, pricing, and service provision in response to economic conditions and policy changes.

These findings reveal that development policies must account for religion’s enduring and adaptive role rather than assuming secularization. Religious institutions often operate in parallel with or in place of state services, meaning their responses can determine whether development programs succeed or fail. Policymakers should recognize that religious actors command superior legitimacy and trust in many contexts, making partnerships rather than competition more effective for service delivery. Understanding religious competition and adaptation is crucial for predicting how communities will respond to economic opportunities, social programs, and institutional reforms.

ABOUT OUR SCHOLAR

Eduardo Montero

Assistant Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

141

RESEARCH BRIEF • OCTOBER 2025

Artificial Writing and Automated Detection Based on BFI Working Paper No. 2025-116, “Artificial Writing and Automated Detection,” by Brian Jabarian and Alex Imas, University of Chicago

Commercial AI detection tools significantly outperform open-source alternatives, with Pangram achieving near-zero error rates while open source options misclassify up to 78% of human text as AI-generated. A new policy framework allows institutions to systematically compare detectors based on their tolerance for false accusations versus missed AI content. Generative Artificial Intelligence tools have been adopted faster than any other technology on record, giving rise to writing that is either assisted or entirely completed by Large Language Models (LLMs). The ubiquity of AI-generated writing across domains such as school assignments and consumer reviews presents a new challenge to stakeholders aiming to detect whether content was written by humans. While automated detection tools hold promise, their accuracy claims are difficult to verify since they rely on proprietary data and methods. In this paper, the authors audit the set of leading AI detection tools and offer a framework to evaluate how they should be incorporated into potential policies. They evaluate four detectors—three commercial tools (Pangram, OriginalityAI, GPTZero) and one open-source baseline (RoBERTa)—on their ability to minimize two critical errors: false positives (wrongly flagging human text as AI) and false negatives (missing actual AI content). Understanding detector performance requires grasping how these tools work. A detector evaluates text and assigns it a score, such that higher scores imply a greater likelihood that the text is AI-generated. A detector’s performance

Figure 1 · Detector False Positive Rates by Genre Detector False Positive Rates by Genre 80%

GPTZero Originality Pangram RoBERTa

Amazon Review

Blog

News

Novel

Restaurant Review

Resume

depends critically on the score threshold at which it classifies content as AI-generated. A higher threshold implies that the detector requires a higher score to classify a passage as AI-generated; this will naturally decrease the false positive rate while at the same time increasing the false negative rate. Given this threshold trade-off, the authors evaluate performance in two ways: First using detector-optimized thresholds calculated to maximize the difference between true positive rate and false positive rate, and second by manipulating thresholds exogenously to demonstrate how policy designers can adjust detector settings based on their tolerance for different types of errors.

bfi.uchicago.edu

142

For their evaluation, the authors use a 1,992-passage text corpus that spans six everyday genres (news, blogs, consumer reviews, novels, restaurant reviews, and résumés) as input for their evaluation. Verified human-generated text is matched with AI-generated text using four frontier LLMs (GPT-4.1, Claude Opus 4, Claude Sonnet 4, Gemini 2.0 Flash). They also examine the effectiveness of AI “humanizers” (StealthGPT) in potentially bypassing detectors.

•

Pangram’s false negative rate is robust to the use of current “humanizers” and remains low even when AI-generated passages are modified using tools such as StealthGPT. The other detectors are less robust to humanizers, with GPTZero largely losing its capacity to detect AI-generated text, showing false negative rate scores around 50% and above across most genres and LLM models.

•

After converting vendor fees into cost per correctly flagged AI passage, Pangram is two times cheaper than OriginalityAI and is almost three times cheaper than GPTZero both overall and on shorter passages. Cost-per-truepositive analysis sharpens the price gap, making Pangram the most cost-efficient detector.

•

The policy caps framework, which sets exogenous thresholds to test detector robustness, reveals that Pangram is the only detector that meets stringent policy requirements without compromising the ability to detect AI text. When policy caps are set at very low levels (0.5% false positive rate), Pangram continues detecting AI content effectively while other detectors see sharp degradation in their detection capabilities.

They find the following: •

•

Commercial detectors significantly outperform open-source alternatives across all metrics and AI models. Among commercial options, Pangram achieves essentially zero false positive rates and false negative rates on medium-length to long passages, both when using detector-optimized thresholds and exogenously-set thresholds. The false positive rate and false negative rate increase slightly on short passages, but remain well below reasonable policy thresholds. The performance gap between commercial and open-source tools is substantial. OriginalityAI and GPTZero constitute a secondary tier among commercial detectors with partial strengths, making the choice between the two dependent on the user’s priority: minimizing false positive rate favors GPTZero, while maximizing ability to distinguish AI from human text favors OriginalityAI. In contrast, the open-source RoBERTa base is deemed unsuitable for high-stakes applications, misclassifying most human text with false positive rates of approximately 30-78% across scenarios.

READ THE WORKING PAPER NO. 2025-116 · AUGUST 2025

Artificial Writing and Automated Detection bfi.uchicago.edu/working-papers/artificial-writing-andautomated-detection

It is important to note that the implications of AI detection for writing and text-based work more generally are not obvious. LLMs are incredibly valuable tools that can facilitate idea generation and help tighten writing. At the same time, the use of LLMs to off-load a task where the receiver explicitly desires human input creates a host of agency problems. The use of AI text detectors in practice must thus strike a delicate balance to avoid discouraging the former while mitigating the issues posed by the latter.

ABOUT OUR SCHOLARS

Brian Jabarian

Howard and Nancy Marks Fellow and Roman Family Center for Decision Research Principal Researcher, Chicago Booth

Alex Imas

Roger L. and Rachel M. Goetz Professor of Behavioral Science, Economics and Applied AI and Vasilou Faculty Scholar, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

143

RESEARCH BRIEF • OCTOBER 2025

Firms Have Partial Knowledge: Evidence from a Reform Based on BFI Working Paper No. 2025-120, “Firms Have Partial Knowledge: Evidence From a Reform,” by Avner Strulov-Shlain, University of Chicago

A pricing reform in Israeli supermarkets that forced firms to adjust prices revealed that they operated with only partial knowledge of optimal pricing. When forced to explore new pricing strategies, firms gradually learned and improved performance, though meaningful deviations from optimal pricing persisted. Economists typically assume that firms are sophisticated and make optimal decisions. Yet, in practice, firms often make mistakes. This study asks why: Do firms know the right action but fail to implement, or do they lack knowledge of the optimal action?

Figure 1 · Shares of Price Endings Over Time

Shares of Price Endings Over Time 60% of Prices

Strulov-Shlain uses data covering supermarket prices and transactions from 2013 to 2015 to examine prices before and after the reform. He also measures how “crossing the left-digit threshold,” e.g., changing a price from $4.99 to $5.00, affects demand. He finds the following: •

Before the reform, roughly 45% of supermarket prices ended in 99, the optimal ending under

INTERIM

POST

Price Ending 0 90 99

To explore this question, Strulov-Shlain examines a 2014 Israeli reform that banned prices ending in denominations of coins that no longer existed, in this case the equivalents of the U.S. penny and nickel. The change forced supermarkets to round prices to the nearest 10 Agorot (e.g., 2.90 or 3.00). Importantly, the reform eliminated the possibility of “99” endings, which are optimal due to left-digit bias (the tendency for consumers to overweight the leftmost digit in a price). In this way, the reform created a natural experiment: Could firms predict the new optimal price endings?

PRE

August 2013

February 2014

August 2014

February 2015

Note: TheThe figure plots thethe shares of of 99-, 90-, and 00-ending prices Note: figure plots shares 99-, 90-, and 00-ending pricesacross acrossproducts productsand and stores stores by sampling period. shaded represents the time between announcement by sampling period. TheThe shaded areaarea represents the time between announcement of the of the reform on October 17, 2013, and its enactment on January 1, 2014. Squares reform on October 17, 2013, and its enactment on January 1, 2014. The blue line represents represent 99-ending prices, triangles 90-ending prices and circles are 00-ending prices.

99-ending prices, the green line 90-ending prices and the red line 00-ending prices.

left-digit bias, while nearly none ended in 00. This suggests firms were partially knowledgeable about optimal pricing. •

When prices ending in 99 were banned, about 40% of prices moved to 90-endings (the new optimal format), but 20% shifted to 00-endings, crossing the left-digit threshold. This suggests that firms lacked full knowledge of how leftdigit bias would translate under new pricing.

•

Crossing the left-digit threshold was costly. Changing a product’s price from 4.99 to 5.00 reduced demand by 5-9% compared

bfi.uchicago.edu

144

to changing it to 4.90. Despite this sizable effect, many firms initially adopted suboptimal 00-endings, suggesting that mispricing was due to lack of knowledge, not inability to implement. •

The share of 00-ending prices fell from roughly 20% immediately after the reform to 5-8% within a year, and around 3% by 2018–2019. Some retail chains exhibited abrupt, “light-bulb” shifts rather than gradual adjustment. At the product level, once prices moved to 90-endings, they tended to stay there, while those initially set to 00 rapidly reverted to 90. Products that transitioned from 00 to 90 were 15 times more likely to remain at 90 than revert, suggesting incremental learning through experience.

•

To quantify the implications of these behavioral patterns, Strulov-Shlain builds a model that translates the estimated demand responses into optimal pricing prediction. He finds that firms lost about 0.75 percentage points in profit immediately after the reform due to mispricing but recovered within a year as they adapted. By the end of the adjustment period, average pricing efficiency exceeded pre-reform levels, suggesting genuine learning rather than mechanical adaptation.

READ THE WORKING PAPER NO. 2025-120 · SEPTEMBER 2025

Firms Have Partial Knowledge: Evidence From a Reform bfi.uchicago.edu/working-papers/firms-have-partialknowledge-evidence-from-a-reform

These findings challenge the standard economic assumption that firms effectively optimize, suggesting instead that partial knowledge should be treated as a benchmark for firm behavior. The mechanism sustaining partial knowledge appears to be insufficient exploration. Had firms experimented more extensively with different price endings before the reform, they would have learned the full structure of demand. For researchers, these findings encourage taking partial optimization seriously when modeling firm behavior. Assuming full optimization may lead to incorrect predictions about how firms respond to regulatory changes or market shocks.

ABOUT OUR SCHOLAR

Avner Strulov-Shlain

Associate Professor of Marketing, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

145

RESEARCH BRIEF • OCTOBER 2025

Human Capital Accumulation Across Space Based on BFI Working Paper No. 2025-134, “Human Capital Accumulation Across Space,” by Klaus Desmet, Southern Methodist University; Dávid Krisztián Nagy, Centre de Recerca en Economia Internacional; and Esteban Rossi-Hansberg, University of Chicago

Regions with lower education costs maintain persistently higher levels of human capital and development over centuries. Improving educational access in poor regions generates local benefits but may reduce global welfare when population shifts away from more productive areas. Policies that equalize education costs within regions can produce unintended consequences as people relocate toward less productive locations, offsetting the benefits of cheaper schooling. Human capital, capital as measured by levels of schoolbased education, is unevenly distributed across space. In 2000, people in the Netherlands had an average of 10.8 years of schooling, compared to 2.5 years in the Central African Republic. Comparing inhabitants of the most and least educated corners of the globe—specifically, 1° × 1° grid cells at the 90th and the 10th percentiles of educational attainment—this range goes from 11.8 to 3.4 years.

What drives these large differences in human capital across space? The authors examine how two factors shape the geography of development, both today and in the future. First, the cost of acquiring human capital varies widely across locations—in some places, access to education is relatively expensive or difficult, limiting the supply of human capital. Second, the productivity of human capital differs across locations—where human capital generates higher returns, demand for

human capital: the collective skills, knowledge, and abilities of individuals that can be used to create economic value

Figure 1 · of Years of Schooling andof Cost of Education Years Schooling and Cost Education A) Average Years of Schooling 2000

B) Education Cost

80° N

60° N

40° N

20° N

0°

20° S

40° S

180°

120° W 0

60° W 2

0° 6

60° E

120° E

180°

120° W 0.1

60° W 0.2

0° 0.3

0.4

60° E 0.5

0.6

120° E

180°

0.7

Note: Note: Panel A shows the average years ofyears schooling in year 2000. Panel B shows education cost, as recovered the model. Panel A shows the average of schooling in year 2000. Panelthe B shows the education cost, as from recovered from the model.

bfi.uchicago.edu

146

Population-Weighted Cost of Human Capital

Figure 2 · Cost ofand Education and GDP per Capita Cost of Education GDP per Capita Central and South Asia EU and North America East and Southeast Asia Latin America

Middle East Oceania Sub-Saharan Africa WLS Slope

They find the following: •

The model predicts strong persistence in the geography of development. Over the span of two centuries, today’s developed regions, such as coastal Australia, Western Europe, Japan, and the United States, remain the most developed 200 years from now. The same persistence holds for population density, as locations that are dense today continue to be dense two centuries from now.

•

Even after 200 years, the world economy remains far from reaching a steady state where all regions grow at the same rate. This finding stands in sharp contrast to spatial models that ignore human capital, which instead predict that poor but densely populated areas will eventually catch up to wealthier regions through agglomeration effects alone.

•

Low education costs drive persistent advantages in developed regions. Because highly developed regions tend to have low education costs, their human capital levels tend to be high, both in the short and the long run. This advantage is magnified by dynamic feedback loops over the transition path. Current levels of human capital improve future productivity, because human capital is an input in the growth of humancapital-augmenting productivity.

•

The model reveals a strong negative correlation between education costs and local economic fundamentals, as places with better amenities, higher productivity, and more favorable conditions for development also tend to have cheaper access to education. As a result, the low cost of acquiring human capital in the developed world keeps these regions ahead, generating the persistence the model projects.

•

Reducing education costs by the same percentage across poor regions generates local gains but may cause global losses. The authors examine counterfactual policies that lower the cost of human capital acquisition while maintaining the relative differences between locations within a region. Whether implemented in sub-Saharan Africa, Latin America, or Central and South Asia, the local economy benefits: human capital levels rise, both in the short

Household Spending per Pupil (as % of income per capita) Note: This figure displays the relationship between the population-weighted cost of

Note: This figure thespending relationship between the population-weighted human capital anddisplays household per pupil as a share of income per capita. cost of human capital and household spending per pupil as a share of income per capita.

it will be greater. These forces interact with migration, trade, and innovation to determine how human capital evolves across the globe over time. To address this question, the authors develop a dynamic spatial model of the world economy at a 1° × 1° resolution. In the model, individuals choose where to live and how much human capital to acquire, taking into account that both moving between locations and upgrading human capital are costly, and these costs vary across space. Firms in each location produce differentiated goods using labor, human capital, and land, with trade between locations subject to transport costs. The model incorporates two key productivity forces. First, a location’s overall productivity benefits from economies—as population density agglomeration economies increases, so does productivity. Second, locations accumulate human-capital-augmenting technology over time through two channels: local innovation (which depends on the local stock of human capital) and diffusion from other locations. This creates dynamic feedback loops where human capital today boosts innovation tomorrow, attracting more people and generating further innovation. The authors quantify the model using data on population, income, and schooling from 2000. They identify location-specific costs of acquiring human capital by matching the model to observed changes in schooling levels between 2000 and subsequent years. They then simulate the model forward for 200 years to project how human capital and economic development evolve across space.

agglomeration economies: the productivity benefits that arise when economic activity concentrates in a particular location, as firms and workers gain advantages from being near one another

147

and the long run. This enhances innovation in human-capital-augmenting productivity, generating positive dynamic effects. •

•

Higher local welfare retains a larger share of the global population in the target region, further reinforcing productivity through agglomeration economies. However, the effects on global welfare differ markedly by region. When the policy is implemented in a low-income region, like subSaharan Africa and Central and South Asia, the local increase in population comes at the expense of regions with better economic fundamentals, reducing global agglomeration and innovation. In contrast, when implemented in a middleincome region like Latin America, the population reallocation comes partly from regions with worse fundamentals, improving outcomes globally. Equalizing educational costs across space may lead to unintended consequences. As an alternative policy, the authors consider setting the cost of education to the same level across all grid cells within a region, such as sub-Saharan Africa—eliminating the variation in education costs between locations. Because schooling costs are higher in less developed areas within the region, equalizing costs lowers them more dramatically in those areas.

READ THE WORKING PAPER NO. 2025-134 · OCTOBER 2025

Human Capital Accumulation Across Space bfi.uchicago.edu/working-papers/human-capitalaccumulation-across-space

•

This creates incentives for population to migrate toward locations with weaker economic fundamentals. As a result, an increasingly larger share of the region’s population resides in less productive locations, which hurts overall innovation and weakens agglomeration economies through greater geographic dispersion. These negative effects may partly or even fully offset the positive impact of cheaper access to human capital.

These findings suggest that effective development policies must account for spatial frictions, agglomeration effects, and the dynamic relationship between human capital and productivity across space. When evaluating education policies, policymakers must consider not only local benefits but also how these policies reshape the global distribution of population and economic activity. Policies that retain population in regions with weak economic fundamentals may generate local gains while producing global losses. Moreover, withinregion heterogeneity matters: equalizing access to education across locations with varying economic potential can trigger population movements that undermine the policy’s intended benefits.

ABOUT OUR SCHOLAR

Esteban Rossi-Hansberg

Glen A. Lloyd Distinguished Service Professor, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

148

RESEARCH BRIEF • OCTOBER 2025

Jealousy of Trade: Exclusionary Preferences and Economic Nationalism Based on BFI Working Paper No. 2025-138, “Jealousy of Trade: Exclusionary Preferences and Economic Nationalism,” by Alex Imas, University of Chicago; Kristóf Madarász, London School of Economics; and Heather Sarsons, University of Chicago

Many voters support tariffs and protectionist policies that materially hurt them because they derive value from consuming or possessing goods that others want but do not have. Individuals with such “exclusionary preferences” are significantly more willing to accept higher prices from tariffs than from other policies, like stimulus spending. Why do voters support protectionist policies that materially harm them? Recent evidence shows that tariffs raise consumer prices and generate retaliatory trade measures that reduce employment. Such policies remain politically popular, however, particularly among those most economically affected. In this paper, the authors offer a new explanation: support for nationalist economic policies stems from a fundamental desire for dominance, which generates preferences for excluding others from consumption opportunities. The authors build on prior research documenting that a substantial portion of the population derives utility not just from consuming goods, but from consuming goods that others desire but cannot obtain. They incorporate this desire for dominance into a model of international trade, showing that such exclusionary preferences reduce the value of trade and generate support for restrictive policies. The model predicts that people with exclusionary preferences will support tariffs that harm both their own consumption and their trading partner’s consumption, but will show no such preference for policies that affect only domestic consumption.

To test these predictions, the researchers conduct two surveys. They begin by measuring respondents’ exclusionary preferences using an incentivized experimental method in which participants bid on a unique good under three scenarios with varying degrees of exclusion of other potential buyers. Those whose willingness to pay increased with the level of exclusion are classified as having “preferences for exclusion,” a pattern observed in roughly 40% of respondents (consistent with prior research). Respondents are then randomly assigned to evaluate tariff policies under different conditions and asked about their support for various economic policies. The authors find the following: •

Exclusionary preferences strongly predict tariff support, but only when tariffs harm trading partners. Those with exclusionary preferences are 12.3 percentage points more likely to support a 15% tariff that would raise prices domestically. When respondents are told the tariff would not harm the foreign country, support between those with and without exclusionary preferences is statistically indistinguishable.

bfi.uchicago.edu

149

•

Those with exclusionary preferences are more accepting of inflation caused by tariffs than by other policies. When comparing support for tariffs versus stimulus policies that would generate identical 15% price increases, respondents with exclusionary preferences show significantly higher support for tariffs.

•

Exclusionary preferences predict support for a broad range of protectionist policies that harm domestic consumers. Beyond tariffs, those with exclusionary preferences are significantly more likely to support policies explicitly designed to maintain consumption gaps between nations, even when informed these policies would raise prices for Americans. They also show higher support for restricting foreign investment, emphasizing that the US should “come out on top” in trade relations, and limiting purchases from foreign countries. These patterns held across different trading partners (China, Mexico, and Canada), suggesting the effects are not driven by hostility toward specific nations.

•

The relationship between exclusionary preferences and policy support is not explained by political ideology or cognitive biases. While political preferences partially mediate the relationship (Democrats are less likely to hold exclusionary preferences), the core association remains strong and statistically significant after controlling for party affiliation and zero-sum thinking (a cognitive bias where people believe gains for some come at others’ expense).

These findings have important implications for understanding the political economy of trade policy. The results suggest that voter support for protectionist measures may be driven less by misunderstanding of economic costs or by narrow self-interest than by a fundamental preference for policies that exclude foreign consumers from consumption opportunities, even at personal economic cost. This helps explain why tariffs remain politically popular despite clear evidence that they raise prices and harm employment. The findings also suggest that inflation stemming from protectionist policies may generate less political backlash than equivalent price increases from other sources, as voters with exclusionary preferences view such costs as more acceptable when they serve to limit foreign consumption.

READ THE WORKING PAPER

ABOUT OUR SCHOLARS

NO. 2025-138 · OCTOBER 2025

Alex Imas

Jealousy of Trade: Exclusionary Preferences and Economic Nationalism bfi.uchicago.edu/working-papers/jealousy-of-tradeexclusionary-preferences-and-economic-nationalism

Roger L. and Rachel M. Goetz Professor of Behavioral Science, Economics and Applied AI and Vasilou Faculty Scholar, Chicago Booth

Heather Sarsons

Associate Professor of Economics and William Ladany Faculty Scholar, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

150

RESEARCH BRIEF • OCTOBER 2025

Superstar Firms Through the Generations Based on BFI Working Paper No. 2025-118, “Superstar Firms Through the Generations,” by Yueran Ma, University of Chicago; Benjamin Pugsley, University of Notre Dame; Haomin Qin, London Business School; and Kaspar Zimmermann, University of Hamburg

New technologies that exhibit economies of scale, that confer low adoption costs for new entrants, and that require organizational learning, give rise to superstar firms for a long period of time. These firms enjoy systematic advantages relative to both firms that came before and potential entrants thereafter. In recent years, researchers and the business media have focused attention on superstar firms, or that small set of top companies that account for a large share of output. However, though we understand a good deal about the current makeup of these firms, questions persist about how these firms are born, and whether/how these firms attain their superstar status over time.

In this work, the authors collect new data to conduct an extensive analysis of the largest US companies over the past century. These data include the 2018 Fortune list (a recent example year before COVID), which covers the largest 1,000 companies by sales across all sectors; the first Fortune list in 1955, which covers the largest 500 industrial (i.e., manufacturing and mining)

Figure 1 · Birth Years of Years 2018 Fortune 1,000 Companies Birth of 2018 Fortune 1,000 Companies A) Construction (26 Firms)

B) Finance (168 Firms)

C) Industrials (388 Firms)

40 Firms

1800 1850 1900 1950 2000

D) Services (139 Firms)

E) Trade (159 Firms)

F) Utilities (120 Firms)

40 Firms

1800 1850 1900 1950 2000

Note: of This figure plots the number that of 2018 Fortune 1,000 companies originated decade, Note: This figure plots the number 2018 Fortune 1,000 companies originated in each decade, based on that the firm’s industryin in each 2018 (the main sectors correspond to Standard Industrial on the firm’s industry in 2018 (the main sectors correspond to Standard Industrial Classification Classification codes; please based see working paper for more details). For industrials, there is substantial clustering of birth years around the turn of the 20th century: the cohort from 1880 to 1920 is the codes; please see working paper for more details). For industrials, there is substantial clustering of birth most represented among today’s largest industrials. years around the turn of the 20th century: the cohort from 1880 to 1920 is the most represented among today’s largest industrials.

bfi.uchicago.edu

151

Figure 2 · Age Distribution of the Largest Industrials Age Distribution of the Largest Industrials

the 388 industrials in the 2018 Fortune list, 137 were born between 1880 and 1920, but only 50 were among the top 388 industrials in 1955; the rest (and the majority) is represented by “late bloomers.”

25% of Firms 2018 1955 1917

20 15

•

10 5 0

100 150 Age Decade

200

250

Note:The The blue shows thedecade age decade distribution of the388 largest 388 (by sales) in Note: blue lineline shows the age distribution of the largest industrials industrials (by sales) in the 2018 388 Fortune list (which contains 388shows industrial the 2018 Fortune list (which contains industrial firms). The green line the age decade firms). Theofgreen line shows the age(by decade of thelist. largest 388 distribution the largest 388 industrials sales) distribution in the 1955 Fortune The red line shows industrials (bydistribution sales) in the 1955 Fortune list. The red shows the age the age decade of the largest 388 industrials (byline assets) in 1917. decade distribution of the largest 388 industrials (by assets) in 1917.

companies by sales; Fortune lists of top retailers and wholesalers; and a list of the largest 500 industrials by assets in 1917. The authors also research the origin story for each firm using extant resources.

By comparison, the authors find a similar pattern in the industrial history of Germany. In the UK, though, superstar firms are relatively young today. In part, this evidence suggests that the special generation among top industrials is not just a result of country-specific regulations (given the similarities in the US and Germany), or military buildup in the world wars (which would also be relevant in the UK).

Retail and wholesale •

For reference, the authors use “superstar firms” to denote large companies that have achieved an extraordinary size relative to other businesses (for example, total sales of the top 1,000 firms in the 2018 Fortune list exceed 40% of U.S. private sector gross output). A review of the data reveals the following:

In contrast to large industrial firms, superstar retailers and wholesalers are relatively stable in their age distribution. Today, the birth years of large retailers and wholesalers cluster around 1960 to 1980. In the 1950s, though, top companies primarily date back to around 1900.

•

The authors show that the largest retailers and wholesalers have stayed 60 to 70 years old on average, without a special generation.

Industrials

Service

•

The emergence of superstar industrial firms in the United States is highly uneven over time—with a special generation from 1880 to Second Industrial Industrial 1920 (around the time of the Second Revolution) Revolution remaining dominant among the largest manufacturers in the 1910s, 1950s, and 2010s—but top firms still experience substantial turnover. This “special generation” has a lasting influence; the median age of top firms was around 30 in 1917, 60 in 1955, and 100 by 2018. However, the persistence of this special generation does not extend to individual firms. Twenty-one percent of the top industrials on the 1955 Fortune list remain on the 2018 Fortune list. Correspondingly, among

•

Today, birth years for superstar service firms cluster around 1960 to 1980, and few large ones existed before then; also, few services companies would qualify for the largest businesses in the economy until recently.

•

In 2018, the large services in Fortune 1,000 are young, with a median age of 43, which resembles the youth of top industrials in 1917.

•

A possible explanation for the relative youth of service superstars is that the cohort of services firms born around the Third Industrial Revolution forms a special generation, like the cohort of industrial firms born around the Second Industrial Revolution, but several more decades of data are necessary to confirm this account.

industrial firms: firms in manufacturing and mining industries according to the Standard Industrial Classification codes, which are four-digit codes that classify a company by its economic activity Second Industrial Revolution: a period from around 1860 to 1900, marked by advancements in steel, electricity, and petroleum, and giving rise to large-scale industrialization and major corporations Third Industrial Revolution: begun in the late 20th century, this era is characterized by the shift from mechanical and analogue technologies to digital electronics; key drivers include computers, the internet, and telecommunications

152

The authors take these data to a model of firm dynamics, which offers the following explanations for the birth and persistence of superstar firms: •

•

Declining adoption costs of the scalable technology generate the advantage of the special cohort relative to firms that came before; The accumulation of productivity over time through learning gives superstars first-mover advantages relative to potential entrants afterwards; Idiosyncratic firm-level shocks keep the top firms changing despite the persistence of the special generation; and The persistence of the special generation does not necessarily imply staleness or lack of dynamism among top firms.

Figure 3 · Age Distribution of the Largest Retailers andDistribution Wholesalers of the Largest Retailers and Wholesalers Age 20% of Firms

2018 1995 1970 1956

100 Age Decade

150

Note: The blueline lineshows shows the distribution of the 50 retailers Note: The blue theage agedecade decade distribution oflargest the largest 50 and wholesalers in the 2018 Fortune The green line the ageline decade distribution retailers and wholesalers in the list. 2018 Fortune list.shows The green shows the age of the decade distribution of Fortune the largest 50 red in the list. Thedistribution red line of the largest largest 50 in the 1995 list. The line1995 showsFortune the age decade shows age distribution the largest 50 in 1970. The yellow 50 inthe 1970. Thedecade yellow line shows theof age decade distribution of the largest 50 line in 1956. shows the age decade distribution of the largest 50 in 1956.

Bottom line: Certain historical settings produce special generations of entrants that give rise to superstar firms for decades to come. These settings occur occasionally, and include new technologies that exhibit economies of scale, that confer low adoption costs for new entrants, and that require organizational learning. The combination of these forces produces special cohorts that have a strong edge relative to both firms that came before and potential entrants thereafter. That said, individual superstars keep churning due to idiosyncratic shocks.

READ THE WORKING PAPER NO. 2025-118 · AUGUST 2025

Superstar Firms Through the Generations bfi.uchicago.edu/working-papers/superstar-firms-throughthe-generations

200

ABOUT OUR SCHOLAR

Yueran Ma

Carhart Family Professor of Finance, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

153

RESEARCH BRIEF • OCTOBER 2025

The Breakdown of the English Society of Orders: The Role of the Industrial Revolution Based on BFI Working Paper No. 2025-108, “The Breakdown of the English Society of Orders: The Role of the Industrial Revolution,” by Cara Ebert, RWI Leibniz Institute for Economic Research; Leander Heldring, Northwestern University; James A. Robinson, University of Chicago; and Sebastian Vollmer, University of Goettingen

The Industrial Revolution fractured the strictures from a centuries-old “society of orders” in England, introducing the phenomenon of social mobility that created new opportunities for workers, and initiating a “dawn of liberty” for those previously entrenched in rigid social hierarchies. Few areas of economic research have Figure 1 · Relationship of 17th Century Probate Inventory Wealth andCentury NumberProbate of Hearths attracted as much attention in recent decades Relationship of 17th Inventory Wealth and Number of Hearths as issues surrounding social mobility, both 300 Mean Probate Wealth among researchers and policymakers. Even as economists argued about causation (does moving 250 to a new neighborhood, for example, reduce intergenerational poverty, or does selection 200 bias explain positive outcomes?), policymakers 150 initiated programs to help Americans move. However, social mobility has long been a feature 100 of western economies, though the drivers of that mobility have changed over time. This is 50 0 1 2 3 4 5 ≥6 especially true of the dramatic impact that the Number of Hearths Industrial Revolution had on the traditional Note: This figure shows a binned scatter plot (mean inventory wealth for each categorical number of Note: This figure shows a binned scatterprobate plot (mean probate inventory wealth for each categorical hearths) and a linear fitted line of the probate inventory wealth on the number of hearths. The sample consists “society orders,” which divinely ordered society of orders number of tax hearths) and a linear fitted linewealth of the from probate inventory wealth onthat thecover number of 269 matched hearth observations to probated a dataset of probates theofhearth hearths. sample consists of 269 matched hearth tax observations to probated wealth from a tax period. Please seeThe working paper for more details. allocations of people into separate parts of a dataset of probates that cover the hearth tax period. Please see working paper for more details.

Industrial Revolution: a period of rapid technological advancement beginning in Great Britain in the mid-18th century, that transformed economies from agricultural and handicraft-based systems to those dominated by machine manufacturing and large-scale factories society of orders: the system of social organization in pre-industrial Europe that divided society into a rigid hierarchy of status groups, where an individual’s position was determined primarily by birth and fixed by law and custom

bfi.uchicago.edu

154

hierarchy, and which existed in English society in the centuries leading to the transformative 18th century and beyond. The authors of this new work are the first to empirically conceptualize the society of orders, which allows them to examine its consequences for social mobility over time. Prior to the Industrial Revolution, English society was shaped by a rigid social hierarchy where people’s wealth and status were largely predetermined by birth, noble titles, and family surnames. Medieval society was composed of three orders, those who work, pray, and fight, with surnames often indicating specific occupations that sons would inherit from their fathers (Smiths were blacksmiths, for example, Millers ground wheat, Bakers baked bread, and Coopers forged barrels). Also, surnames largely indicated, if not determined, societal status (children born into such surnames had no chance to rise above their station). That this “great chain of being” was ordained by God ensured that people kept their place, at least until the onset of an unexpected economic “miracle” broke those chains. To investigate the impact of the Industrial Revolution on English society, the authors assembled two unique historical datasets: their pre-Industrial Revolution data consists of hearth tax returns from 1662-1674, which recorded household wealth based on the number of fireplaces, covering 343,022 heads of households across 26 counties, while their post-Industrial Revolution data includes digitized Principal Probate Registry records from 1862-1899, documenting the wealth of deceased persons who held at least 5 pounds at death. Both datasets include individuals’ names, titles, and locations, allowing the researchers to analyze how well traditional social markers predicted wealth distribution across these two periods. They find the following: •

Before the Industrial Revolution, being a member of the nobility or gentry explains 17% of the variation in the share of wealth

individuals owned, afterwards it explains only 6%, a decline of two-thirds. •

Family surnames, which historically carried strong occupational and status associations, explained 10% of wealth variation in the preindustrial period but only 6% afterward.

•

This breakdown was not uniform across England but was most pronounced in regions that experienced greater industrialization. In heavily industrialized areas, the explanatory power of surnames for wealth distribution fell to essentially zero, while in less industrialized regions, surname-based wealth prediction remained constant at around 10%.

The authors conducted benchmarking exercises to confirm that these data do, indeed, represent a breakdown of traditional society and not just statistical noise. They reveal that before the Industrial Revolution, the existing social hierarchy captured about 41% of its theoretical maximum explanatory power for wealth distribution, suggesting a society where traditional orders still significantly influenced economic outcomes. After the Industrial Revolution, this figure plummets to less than 10% of the theoretical maximum. The authors also investigate the extent to which different attributes predict whether a person will be wealthy, or in the top decile of the wealth distribution. Essentially, the Industrial Revolution has little impact on the probability that a noble is rich, while the probability that a member of the gentry is wealthy increases significantly. The authors interpret this effect on the gentry as reflecting the flexibility of this tier of the society of orders. As the economy grew, new entrants were naturally wealthier than incumbents who were less able to take advantage of new economic opportunities. This research also addresses two questions relating to geographic mobility and social mobility: Where did people with socially mobile surnames tend to end up geographically and within what economic

noble: member of the highest-ranking aristocrats, whose titles were often inherited and whose ranks, in descending order, were duke, marquis, earl, viscount, and baron gentry: a class below noble that included landowners, knights, baronets, and esquires, often living on rental income from their estates and often held local positions like justices of the peace

155

sectors; and what were the characteristics of socially mobile parishes parishes? They find that: •

People with socially mobile surnames, or those showing increased wealth dispersion within family lines, were significantly more likely to migrate to northern England and find employment in manufacturing industries.

•

For parishes, those that experienced more mobility were likely urban, had an institutionalized market prior to the Industrial Revolution, were less agrarian, and had lower income levels.

•

Initial social structure was key, with mobility associated with having more gentry and fewer yeomen yeomen, which were a class of wealthier landowning peasants.

•

Political characteristics are also significant: parishes including a member of Parliament are more mobile.

The implications of this research extend beyond historical curiosity to inform our understanding of social mobility more broadly. By focusing on the general population, the authors show that the Industrial Revolution fundamentally altered the relationship between family background and economic outcomes. Likewise, this work challenges narratives that focus solely on the Industrial Revolution’s negative social impacts, suggesting instead that this period may represent a “dawn of liberty” for those previously entrenched in rigid social hierarchies. In doing so, this research contributes to ongoing debates about the relationship between economic development and social mobility, showing how a major economic transformation can weaken traditional status-based systems and create new pathways for advancement.

parish: a fundamental administrative unit for both religious and secular purposes, centered around a church and its priest, that served as the primary unit for managing local taxes, poor relief, and maintaining public infrastructure like roads and bridges yeoman: a member of a rural middle class, ranking below the gentry, who typically owned and cultivated their own land, but also included skilled employees such as manor bailiffs, constables, and household servants

READ THE WORKING PAPER NO. 2025-108 · AUGUST 2025

The Breakdown of the English Society of Orders: The Role of the Industrial Revolution bfi.uchicago.edu/working-papers/the-breakdown-of-theenglish-society-of-orders-the-role-of-the-industrialrevolution

ABOUT OUR SCHOLAR

James Robinson

Professor, Harris School of Public Policy and Department of Political Science; Fellow, Institute of African Studies at the University of Nigeria at Nsukka

Written by David Fettig • Designed by Maia Rabenold

156

RESEARCH BRIEF • OCTOBER 2025

The Effects of Parental Income and Family Structure on Intergenerational Mobility: A Trajectories-Based Approach Based on BFI Working Paper No. 2025-115, “The Effects of Parental Income and Family Structure on Intergenerational Mobility: A Trajectories-Based Approach,” by Yoosoon Chang, Indiana University; Steven N. Durlauf, University of Chicago; Bo Hu, Indiana University; and Joon Park, Indiana University

Parental income and family structure during childhood and adolescence affect adult income, with these familial influences strongest in middle childhood and adolescence. The effects of income and family structure trajectories exhibit a complementary relationship during key developmental periods. A rich literature within economics focuses on mobility, or, the relationship intergenerational mobility between parents’ socioeconomic status and their children’s outcomes later in life. While much research focuses on how parents’ average income shapes that of their child, research also suggests that timing matters across childhood. Moreover, joint consideration of family income and family structure is less common, despite

evidence that family structure impacts a child’s stability, emotional well-being, and mental health in ways that may interact with financial resources. Motivated by this, this paper tracks how parental income and family structure trajectories throughout childhood affect adult income. Using data from over 1,000 families in the Panel Study of Income Dynamics, the authors develop

intergenerational mobility: the extent to which children’s economic outcomes differ from those of their parents, measuring whether advantage or disadvantage passes from one generation to the next

Figure 1 · Dynamics of Overall, Structural, and Exchange Educational Mobility Main and Interaction Effects of Parental Income and Family Structure A) When Parental Income Matters Most for Children's Future Earnings

B) When Family Structure Has the Strongest Effect on Children's Future Earnings 0.006

Smaller Impact Greater Impact Impact on Child's Adult Income

0.005 0.004 0.003 0.002 0.001 0

10 Child's Age

0.005 0.004 0.003 0.002 0.001 0

10 Child's Age

Note:This These figures thatmobility family conditions during the teenage years have the strongest influence on with whether end up in mobility high, middle, or low income brackets as adults.stabilize) with Note: figure plotsshow overall (total probability children have different education than parents) solidchildren lines, exchange (mobility after economic transitions The shaded areas represent confidence intervals. Parental income (left) and living in a two-parent household (right) both show their peak effects around ages 15-17, with relatively dashed andduring structural (mobility due to temporary economic changes like education expansion) with vertical distance between lines, across generations. Steady-state mobility is smallerlines, impacts earlymobility childhood. represented by the rightmost points where overall equals exchange mobility. Shaded areas represent 95% confidence intervals.

bfi.uchicago.edu

157

a novel statistical framework that captures both the timing of family circumstances and their interactive effects across different developmental stages. Rather than examining income levels directly, the authors focus on income categories— classifying children into low-, middle-, and highincome brackets as adults. They define low income as below two-thirds of the median, high income as exceeding twice the median, and middle income as falling between these thresholds. The authors analyze how parental income and family structure trajectories from birth to age 20 predict the probability that a child will end up in each of these three income classes, allowing them to identify when during childhood these family conditions have the strongest effects and whether they work as complements or substitutes. substitutes They find the following: •

While parental income benefits children at all ages, the effects are strongest during middle childhood and adolescence (ages 10-17) rather than early childhood. In other words, a child whose family has higher income during their teenage years is significantly more likely to achieve high adult income than one whose family had the same total income concentrated in early childhood.

•

Family structure effects peak in adolescence. Children from two-parent households show modest advantages in early childhood, but these benefits become pronounced during middle childhood and peak around age 17. Living with both parents during the high school years provides the greatest boost to adult economic outcomes.

•

Financial resources and family stability are complements, not substitutes. Rather than compensating for each other, parental income and two-parent family structure work synergistically. The benefits of higher family income are amplified when combined with family stability, and vice versa. This complementary relationship is strongest between parental income during middle to late childhood and family structure during late childhood.

These results underscore the importance of coordinated policies and interventions that address both economic and familial stability, particularly during critical developmental periods, to maximize the likelihood of positive outcomes for children. Importantly, the strength and timing of these effects may differ by gender, race and ethnicity, parental education, or other sociodemographic characteristics, suggesting an avenue for future research.

complements: two factors that work better together than separately, where the presence of one enhances the effectiveness of the other substitutes: two factors that can replace each other, where having more of one reduces the need for the other

READ THE WORKING PAPER NO. 2025-115 · AUGUST 2025

The Effects of Parental Income and Family Structure on Intergenerational Mobility: A Trajectories-Based Approach bfi.uchicago.edu/working-papers/the-effects-of-parentalincome-and-family-structure-on-intergenerational-mobilitya-trajectories-based-approach

ABOUT OUR SCHOLAR

Steven Durlauf

Frank P. Hixon Distinguished Service Professor; Director, Stone Center for Research on Wealth Inequality and Mobility, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

158

RESEARCH BRIEF • OCTOBER 2025

The Impact of Language on Decision-Making: Auction Winners are Less Cursed in a Foreign Language Based on BFI Working Paper No. 2025-124, “The Impact of Language on Decision-Making: Auction Winners are Less Cursed in a Foreign Language,” by Fang Fu, Amazon Web Services; Leigh H. Grant, Exponent; Ali Hortaçsu, University of Chicago; Boaz Keysar, University of Chicago; Jidong Yang, Renmin University of China; and Karen J. Ye, Queen’s University

Using a foreign language reduces the “winner’s curse” in auctions, as bidders make more strategic decisions and are less likely to overbid when processing information in a nonnative tongue. This effect largely disappears as bidders receive feedback across consecutive auctions, with both native and foreign language bidders converging to similarly poor decision-making patterns driven by observing others’ overbidding behavior. Research on the “foreign language effect” suggests that using a non-native language can change how individuals process information and make decisions. For example, people making moral judgments in a foreign language are more likely to approve utilitarian choices, such as pushing one person off a bridge to save five others, that they would reject when thinking in their native tongue. Similarly, people show reduced emotional responses when making judgments about risk and benefit in a foreign language, leading to more deliberative decision-

making. These effects occur because foreign language use reduces reliance on quick, intuitive, emotionally driven judgments and promotes more deliberative thinking. In this paper, the authors test whether such language effects extend to financial decision-making in market settings. The authors examine this question in the context of blind bidding auctions. They conduct a laboratory experiment in Beijing, China, with 357 native Mandarin Chinese speakers who know English as

Between Figure Difference 1 · Difference Between BidsBids B) Difference from Optimal Bid

Bidding Above the Optimal Strategy

50 Native Language (Chinese)

0 -25 -50 Foreign Language (English)

-75 -100

Sample Information (High-value coins observed)

Bidding Below the Optimal Strategy

Bidding Below the Naïve Strategy

Bidding Above the Naïve Strategy

A) Difference from Naïve Bid

50 25 0 -25 -50 -75 -100

Sample Information (High-value coins observed)

Note: These graphs show difference between individuals’ bids and the Naïve (A) and optimal B) bid by language treatment in the first auction. Vertical bars are 95% confidence intervals.

Note: These graphs show difference between individuals’ bids and the Naïve (A) and optimal (B) bid by language treatment in the first auction. Vertical bars are 95% confidence intervals.

bfi.uchicago.edu

159

a foreign language. Participants bid in ten rounds of auctions with the goal of winning jars filled with a random mix of high-value coins (1 RMB, worth about 15 cents) and low-value coins (0.1 RMB), while avoiding paying more than the value of the jar, or suffering the “winner’s curse”. Before each auction, each participant views a different sample of ten coins drawn from the jar to help estimate its total value. Participants are randomly assigned to complete the entire experiment in either Mandarin or English and in groups of 6 or 12 participants. The authors measure how closely participants’ bids match either a “naïve” strategy (bidding the expected value based on their sample of coins) or a theoretically optimal strategy that accounts for the risk of overbidding. They explore the role of language in influencing participants’ bidding decisions, as well as the role of the number of bidders in an auction and response to feedback across auction rounds. The authors find the following: •

•

Bidders using a foreign language initially make more strategic decisions and are less susceptible to the winner’s curse. In the first auction, 95% of native language bidders bid more than the value of the jar, compared to only 68% of foreign language bidders. Native language bidders bid significantly above the optimal level by an average of 16.05 RMB, while foreign language bidders did not deviate from the optimal bid. The foreign language effect disappeared across consecutive auctions as participants received feedback about others’ bidding behavior. By the second through tenth auctions, bidders in both language treatments adopted similarly poor naïve bidding strategies, with winner’s curse

READ THE WORKING PAPER NO. 2025-124 · SEPTEMBER 2025

The Impact of Language on Decision Making: Auction Winners are Less Cursed in a Foreign Language bfi.uchicago.edu/working-papers/the-impact-of-languageon-decision-making-auction-winners-are-less-cursed-in-aforeign-language

rates converging to around 85% in both groups. •

Feedback about previous winners’ overbidding drove the convergence to poor decision-making. When auction winners overbid relative to the jar’s actual value, other participants increased their own “cursedness” levels in subsequent auctions more than when winners did not overbid. This created a cycle where observing overbidding led to more overbidding.

•

Decision making speed reflected the quality of choices across language treatments. Foreign language bidders took significantly longer to place bids in the first auction when they were making strategic decisions, but bidding speed converged between treatments as both groups adopted faster, more intuitive naïve strategies in later auctions.

•

Group size effects were eliminated by foreign language use. Native language bidders in larger groups (12 participants) overbid significantly more than those in smaller groups (6 participants), replicating prior findings. However, foreign language bidders showed no difference in overbidding between large and small groups, suggesting the foreign language dampened competitive pressures.

These results have important implications for financial decision-making not only in auctions but more broadly in increasingly globalized market settings where participants frequently operate in non-native languages, suggesting both potential benefits and limitations of foreign language use in financial contexts. This interdisciplinary research was a joint effort of researchers from Economics and Psychology and was funded by NSF.

ABOUT OUR SCHOLARS

Ali Hortaçsu

William B. Ogden Distinguished Service Professor of Economics, Kenneth C. Griffin Department of Economics

Boaz Keysar

William Benton Professor in Psychology, Department of Psychology

READ THE JOURNAL ARTICLE JOURNAL OF ECONOMIC PSYCHOLOGY · OCTOBER 2025

The Impact of Language on Decision Making: Auction Winners are Less Cursed in a Foreign Language sciencedirect.com/science/article/pii/S0167487025000571

Written by Abby Hiller • Designed by Maia Rabenold

160

RESEARCH BRIEF • OCTOBER 2025

Why Is Manufacturing Productivity Growth So Low? Based on BFI Working Paper No. 2025-127, “Why Is Manufacturing Productivity Growth So Low?” by Enghin Atalay, Federal Reserve Bank of Philadelphia; Ali Hortaçsu, University of Chicago; Nicole Kimmel, Federal Reserve Bank of Philadelphia; and Chad Syverson, University of Chicago

Nearly all measured total factor manufacturing productivity growth since 1987, and its post-2000s decline, comes from a few computer-related industries. Conventional productivity growth statistics understate the manufacturing sector’s productivity by failing to fully capture quality improvements. TFP growth is understated by 1.6 percentage points in durable manufacturing and 0.5 percentage points in nondurable manufacturing. Official productivity statistics paint a picture of a manufacturing industry in decline. From 1987 to 2009, the Bureau of Labor Statistics (BLS) total factor productivity (TFP) index for manufacturing grew by 1.2% annually, outpacing the 0.9% growth rate of the overall private economy. But between 2009 and 2023, the pattern reversed. Manufacturing TFP fell slightly while private economy TFP continued to rise at 0.8% per year. This reversal is especially concerning given manufacturing’s traditional role as a driver of innovation across the broader economy. In this paper, the authors examine the recent slow growth in manufacturing productivity. They begin by documenting that nearly all measured TFP growth since 1987, and its post-2000s

Figure 1 · Productivity Growth: Manufacturing vs. Rest of Economy

Productivity Growth: Manufacturing vs. Rest of Economy 130 TFP Index (1997=100)

Private Economy

120 Manufacturing 110

100

Manufacturing Excluding Computers & Electronics

1990

2000

2010

2020

Note: ThisThis figure plots the cumulative contribution to manufacturing productivity growth growth for Note: figure plots the cumulative contribution to manufacturing productivity for all industries other than Computer and Electronic Products Manufacturing.

all industries other than Computer and Electronic Products Manufacturing.

total factor productivity: a measure of how efficiently a firm converts inputs (labor, capital, and materials) into output, capturing productivity improvements beyond changes in input usage

bfi.uchicago.edu

161

decline, comes from a few computer-related industries. While productivity growth slowdowns are observed in multiple manufacturing industries, most of the measured sectorwide stagnation is quantitatively explained by productivity changes in Computer and Electronic Products Manufacturing. Building on this result, the authors next scrutinize whether official productivity measures accurately capture productivity for this highly innovative industry. Traditional TFP measures calculate productivity by comparing the real (inflationadjusted) value of what industries produce to the real value of the inputs they consume. To make this calculation, statisticians must convert nominal dollar figures into real quantities, relying on price indices to strip out the effects of inflation. These price indices must capture the frequent technological improvements within Computer and Electronic Products, or else risk undercounting the sector’s production. For example, consider a smartphone that costs $800 today and $800 five years ago. Today’s version likely has a better camera, faster processor, and more storage. When price indices fail to account for these quality improvements, they effectively treat a superior product as having the same price, missing the fact that the real (quality-adjusted) price has fallen dramatically. This measurement error can have cascading effects, threatening to overstate inflation, understate real output growth, and make productivity gains appear smaller than they actually are. The authors hypothesize that this quality-adjustment problem is precisely what is happening in manufacturing statistics, causing conventional measures to understate productivity growth in highly innovative industries. To test this hypothesis, the authors exploit a key institutional difference: the BLS invests more in

quality adjustment for consumer price indices than producer price indices or import price indices. This is important because the last two price indexes are used to construct sector output and productivity, while consumer price indices are not. The authors compare consumer-facing prices (Personal Consumption Expenditures price index) against producer-facing prices (BEA gross output deflators and BLS import price indices) across 212 consumption categories and 414 commodities, using instances where consumer indices show steeper price declines than producer indices to identify unmeasured quality improvements. To translate these price gaps into productivity estimates, the authors use input-output tables tables, which track what each industry buys from other industries, allowing them to trace the impacts of mismeasurement across the entire economy. They find the following: •

Producer and import price indices understate quality growth. Comparing consumer-facing and producer-facing price indices for the same products reveals systematic gaps in rapidly innovating industries. For computers and electronics, consumer price indices show far steeper price declines than producer or import price indices, indicating that official producer deflators fail to capture quality improvements. This means measured inflation is overstated and real output growth is understated in these industries.

•

Manufacturing productivity is systematically understated. Using an input-output framework to account for price mismeasurement in both outputs and inputs, the authors estimate that manufacturing TFP growth is understated by approximately 0.8 percentage points annually. The mismeasurement is concentrated in durable goods manufacturing (understated by 1.6 percentage points) with smaller effects

real: the value of money or interest rates after removing the effects of inflation, showing the true purchasing power nominal: the stated or face value of something, like money or interest rates, without adjusting for inflation price indices: a statistical measure that tracks how the average price of a basket of goods changes over time input-output table: an economic accounting framework that maps which industries supply inputs to which other industries across the entire economy

162

in nondurable manufacturing (0.5 percentage points). No significant mismeasurement appears outside the manufacturing sector. The manufacturing sector has undergone profound transformation over the past quarter century. Employment has collapsed by more than one-quarter since 1997 even as private nonfarm employment grew by more than one-quarter. The sector has become increasingly import-reliant— first from China, then from Vietnam and Mexico— more capital-intensive, and more robot-intensive. Assessments of manufacturing’s evolution, and evaluations of the billions spent on federal support programs like SEMATECH, Manufacturing USA, and the CHIPS Act, hinge on properly measuring real output and productivity.

READ THE WORKING PAPER NO. 2025-127 · SEPTEMBER 2025

Why Is Manufacturing Productivity Growth So Low? bfi.uchicago.edu/working-papers/why-is-manufacturingproductivity-growth-so-low

ABOUT OUR SCHOLARS

Ali Hortaçsu

William B. Ogden Distinguished Service Professor of Economics, Kenneth C. Griffin Department of Economics

Chad Syverson

George C. Tiao Distinguished Service Professor of Economics, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

163

RESEARCH BRIEF • NOVEMBER 2025

Closing Early Math Gaps by Parental Education with Technology at Home Based on BFI Working Paper No. 2025-129, “Closing Early Math Gaps by Parental Education with Technology at Home,” by Daniela Bresciani, Ariel Kalil, Haoxuan Liu, and Susan E. Mayer, University of Chicago Harris School of Public Policy; and Rohen Shah, Yale University

A six-month experiment with 459 diverse Chicago families shows that providing highquality math apps to children of parents without college degrees improved their math skills by 0.17 standard deviations, closing roughly one-third of the initial educationbased achievement gap. The intervention had no effect on children of college-educated parents, who already had access to comparable resources, suggesting that unequal access to quality learning materials is a key driver of early childhood skill gaps. Math skills developed in early childhood predict later academic achievement, yet socioeconomic gaps in these skills emerge even before children start formal schooling. A key contributor to these disparities is variation in how much and how effectively parents engage their children in learning activities at home. While much research focuses on classroom-based interventions, evidence on home-based strategies to improve early math skills remains limited. In this paper, the authors investigate whether providing families with high-quality math learning materials, either digital apps or traditional analog materials, can improve young children’s math skills and reduce achievement gaps by parental education. The authors conduct a six-month randomized controlled trial called About TIME (About Technology

in Math Engagement) with 459 families of preschoolers aged 3-5 from 35 Chicago preschools spanning the socioeconomic spectrum. They randomly assign families to one of three groups: a control group receiving a coloring book unrelated to math, a treatment group receiving a tablet preloaded with four high-quality math apps, or a treatment group receiving analog math materials designed to match the content of the apps. The authors assess children’s math skills at baseline as well as after six months using a standardized measure of their ability to understand and work with numbers. They also collect detailed survey data on parental time investment, attitudes, and experiences with the materials, along with objective app usage data from the tablets.

bfi.uchicago.edu

at's inside this box 164

iPad and charger Experimental Groups Assigned to Families Experimental Groups Assigned to Families Treatment Group: Treatment Group: Tablet with Math Apps Analog Math Materials With four preinstalled apps: Endless Numbers

Funexpected Math

Funexpected MathTango Endless Math Numbers

Montessori Funexpected Preschool Math

Montessori MathTango Funexpected Preschool Math

Montessori Preschool

Control Group: Coloring Book

preinstalled apps: With four preinstalled apps:

ss bers

preinstalled apps:

If you have any questions regarding the About TIME project, please contact Ana Arellano at 773.834.3143 or arellanoa@uchicago.edu.

Tango ss bers

Montessori Preschool

arding the About If you have any questions regarding the About Ana Arellano at TIME project, please contact Ana Arellano at chicago.edu. 773.834.3143 or arellanoa@uchicago.edu.

Tango

arding the About Ana Arellano at The authors compare outcomes across the three hicago.edu.

app group, 74% of non-BA parents reported that the About TIME apps taught math more effectively than their other materials, compared to just 40% of BA parents. Non-BA parents also used the apps about one additional day per week, and objective usage data showed they spent roughly 50% more time on the apps than BA parents. This suggests that the benefits of the About TIME apps differed by parental education because the counterfactual learning environments vary sharply across groups.

experimental groups and find the following: •

Children whose parents did not hold a bachelor’s degree and were assigned to the math app group improved their math skills by 5 percentiles in the national distribution of math skills (0.17 standard deviations), compared to the control group. This gain closed approximately one-third of the initial 25-percentile-point gap between children of BA and non-BA parents. In contrast, the intervention had no detectable effect on children of college-educated parents.

•

Analog materials showed no significant effects for any group.

•

The benefits of About TIME were specific to parental education. When families were divided by household income (above or below the median) or by children’s baseline math scores, no significant treatment effects emerged. This suggests that parental education captures a distinct dimension of family advantage related to access to quality learning resources.

•

Non-BA parents valued and used the apps more than BA parents. Among families in the math

•

Math apps substituted for analog materials, rather than increasing total learning time. Parents in the math app group reduced time spent on analog math activities by about 10 minutes per week while modestly increasing time with math apps, with no change in total math learning time. There was also a spillover effect: families receiving math apps reduced weekly reading time with their children by about 21 minutes.

•

The findings replicate results from a previous study with disadvantaged families. When the About TIME sample was restricted to children from publicly subsidized preschools (matching the recruitment source of the authors’ earlier MPACT

165

study), the math app intervention produced a 0.16 SD improvement—closely matching MPACT’s finding of 0.20 SD. This demonstrates the robustness and replicability of the intervention’s benefits for disadvantaged preschoolers. These findings have important implications for addressing early childhood achievement gaps. The results suggest that unequal access to highquality, easy-to-use math learning materials is a significant driver of skill gaps by parental education. Unlike advantaged families who already possess comparable resources at home, disadvantaged families lack access to such materials and therefore benefit substantially from receiving them. The replication of findings from the earlier MPACT study strengthens confidence that at-home technology can be a scalable tool for improving math outcomes among disadvantaged preschoolers.

READ THE WORKING PAPER NO. 2025-129 · SEPTEMBER 2025

Closing Early Math Gaps by Parental Education with Technology at Home bfi.uchicago.edu/working-papers/closing-early-math-gapsby-parental-education-with-technology-at-home

ABOUT OUR SCHOLARS

Daniela Bresciani

PhD Student, Harris School of Public Policy

Ariel Kalil

Daniel Levin Professor, Harris School of Public Policy; Director, Center for Human Potential and Public Policy; Co-Director, Behavioral Insights and Parenting Lab

Haoxuan Liu

Ph.D. Candidate Harris School of Public Policy

Susan E. Mayer

Professor Emeritus Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

166

RESEARCH BRIEF • NOVEMBER 2025

Debt and Assets

Based on BFI Working Paper No. 2025-92, “Debt and Assets,” by Efraim Benmelech, Northwestern University; Nitish Kumar, University of Florida; and Raghuram Rajan, University of Chicago

Contrary to conventional wisdom, much unsecured debt is implicitly asset backed, and the degree to which unsecured debt is asset backed can change with a firm’s condition and macroeconomic conditions. Asset values can also affect the price of borrowing, notably in adverse economic conditions. All told, the industry practice of classifying debt as “asset based” or “cash-flow based” is overly categorical. Recent research seems to suggest that most US creditors lend money expecting repayment from a company’s cash flows flows, with tangible assets serving merely as a secondary fallback when those flows prove inadequate. A corporation’s assets are less valuable, in other words, than the cash flows it expects to generate, at least when it comes to borrowing. Along these lines, an important recent study shows that 80% of US corporate borrowing takes the form of cash-flow-based debt debt, rather than debt secured by specific assets. This shift toward cash-flowbased lending appears to reflect the increasing sophistication of American financial markets, where earnings-based covenants have become the primary tools for assessing borrowing capacity. The new paper by Benmelech, Kumar, and Rajan challenges such an interpretation by examining the

importance of assets to borrowing for relatively large US firms in recent decades. The critical insight missing from recent analysis is that debt can be backed by assets even when those assets are not explicitly pledged as security. Large, established firms with substantial unpledged assets can issue unsecured debt that is implicitly backed by their tangible assets, providing creditors with confidence while preserving corporate flexibility. This implicit backing represents a strategic choice rather than a limitation; financially strong firms often prefer to avoid secured borrowing, preserving their ability to pledge assets in future situations when financing might be much more difficult. The trick is to detect and measure this implicit asset backing. Such efforts are hamstrung by methodological limitations. Simple correlations

cash flow: the movement of money into and out of a company over a certain period. If the company’s inflows of cash exceed its outflows, its net cash flow is positive. If outflows exceed inflows, it is negative. tangible assets: a physical item with a finite monetary value that can be touched and utilized (tangibility), such as land, buildings, or machinery, and is recorded on a company’s balance sheet cash-flow-based debt: debt that is not covered by specific assets, such as land or a piece of equipment. All unsecured debt is cash-flow-based debt; but so is debt with a general lien against assets. unsecured debt: loans that are not backed by collateral. If the borrower defaults on the loan, the lender may not recover its investment because there are no pledged assets to be seized and sold.

bfi.uchicago.edu

167

between unsecured debt and tangible assets often yield insignificant or even negative correlations, not because assets are unimportant, but because the analysis fails to account for the complex interactions between different types of debt and the varying financial constraints firms face. For example, when the researchers correct for assets already encumbered by secured debt, a strong positive correlation between unsecured debt and available unpledged collateral emerges clearly across the sample of firms. More revealing still is the evidence of how this relationship varies with economic conditions and firm circumstances. During periods of macroeconomic stress, the importance of implicit asset backing becomes more pronounced. Investment-grade firms with substantial unpledged tangibility demonstrate increased debt issuance during economic downturns, and their borrowing spreads reflect the implicit value of their unencumbered assets, particularly during difficult times. This pattern suggests that asset backing is not a fixed characteristic of debt instruments but rather a dynamic feature that responds to changing conditions. Unpledged assets can play different roles for different firms. For example, for financially strong firms operating under normal conditions, creditors primarily look to going-concern value rather than liquidation value for repayment assurance. The presence of substantial unpledged assets serves multiple functions in this context: it prevents destructive competition for collateral among creditors, enables access to debtorin-possession financing during potential reorganizations, and provides monitoring creditors with confidence that their claims could be fully secured if circumstances deteriorate. Even when debt remains unsecured throughout a firm’s reorganization, the going-concern value upon which repayment depends may be

significantly enhanced by the availability of substantial collateralizable assets. This analysis reveals the existence of three distinct but fluid categories of firms in the corporate debt market: •

The first category comprises financially constrained firms that must issue secured debt to borrow, explicitly pledging assets to obtain lender comfort.

•

The second includes firms whose debt appears to be cash-flow-based but is implicitly backed by substantial unpledged assets, with the asset-debt relationship masked.

•

Finally, the third category encompasses firms with such stable cash flows and high going-concern values that assets truly become secondary to debt capacity and borrowing decisions.

Membership in these categories is not permanent. The same debt instrument issued by the same firm can shift along the spectrum from cash-flowbased to asset-backed as corporate conditions and macroeconomic circumstances change. During favorable periods, a firm’s debt may rely primarily on cash flow expectations and going-concern value, but as conditions deteriorate, the implicit backing provided by unpledged assets becomes increasingly important to both debt capacity and pricing. Some unsecured debt even becomes explicitly secured during challenging times, demonstrating the fluid nature of these relationships. This dynamic understanding has profound implications for both financial theory and practical application. The traditional binary classification of debt as either cash-flow-based or asset-based proves inadequate for capturing the sophisticated reality of modern corporate finance. Instead, most debt exists simultaneously in both categories, with the relative emphasis shifting based on firm-specific circumstances and

investment-grade firms: an investment-grade firm is a company with a high credit rating from a major credit rating agency like Standard & Poor’s (S&P), Moody’s, or Fitch. These ratings signify that the firm has a strong financial profile and a low risk of defaulting on its debt obligations. going-concern value: the value of a company generated in its continuing business activities liquidation value: the worth of a company’s business activities and assets if it were to sell them in the event of going out of business debtor-in-possession financing: this financing allows companies that have filed for bankruptcy protection under Chapter 11 to borrow capital to restructure and continue trading. These loans usually have priority over existing debt, equity, and other claims and are facilitated in the hope that the distressed company, with a new cash injection, can save itself, begin making money again, and pay off all its debts.

168

broader economic conditions. This recognition suggests that phenomena such as the sale of assets at fire sale prices, the acceleration provided to investment when real estate prices increase, and the importance of collateral to borrowing in adverse situations remain highly relevant even in the most developed financial markets. Also, debt capacity can indeed depart from strict asset-value constraints during prosperous times.

importance of tangible assets in corporate finance. Rather than being replaced by cashflow-based lending, asset backing has become more subtle and conditional, providing a form of financial insurance that becomes more valuable precisely when it is most needed and preserving flexibility when conditions are favorable. Understanding this dynamic relationship is essential for policymakers, lenders, and corporate managers navigating an increasingly complex financial landscape where the traditional boundaries between different types of debt continue to blur and evolve.

The evolution of corporate debt structures reflects the broader sophistication of modern financial markets, but it also reveals the enduring

Leverage and Tangible Assets The positive association between leverage and tangible assets is consistent with the notion that tangible assets serve as useful collateral that mitigate financing constraints and enhance firms’ debt capacity.

Figure 1 · Residuals of Leverage Against Quartiles of Residual Tangibility

Figure 2 · Residuals of Unsecured Figure 3 · Unpledged Tangibility Leverage Against Quartiles of and Investment-Grade Unsecured Residual Unpledged Tangibility Bond Spreads Around Covid-19 Residuals of Unsecured Leverage Against Quartiles of Unpledged Tangibility and Investment-Grade Unsecured Bond

Residuals of Leverage Against Quartiles of Residual Tangibility Residual Unpledged Tangibility

Spreads Around the Covid-19 Pandemic

1.5 1

-60

-80

-100

-1

-120

0.5 Covid-19 Pandemic

0 -0.5 -1

-2 1

-140

Tangibility Quartile

2 3 Unpledged Tangibility Quartile

Note: This figure plots the average residuals of leverage against quartiles of Note: This figure plots the residuals of unsecured leverage against quartiles of Note: This figure plots the thesteady residuals unsecured residualNote: tangibility. The steady plots pattern the of increasing leverage with increased This figure average residuals of leverage residual unpledged tangibility, revealing pattern of of increasing unsecured tangibility, after correcting for other firm characteristics and macroeconomic leverage with increased unpledged tangibility. leverage against quartiles of residual unpledged against of residual steady conditions, is clear.quartiles (For each figure, please seetangibility. working paper The for more details.)

pattern of increasing leverage with increased tangibility, after correcting for other firm characteristics and macroeconomic conditions, is clear. (For each figure, please see working paper for more details.)

tangibility, revealing the steady pattern of increasing unsecured leverage with increased unpledged tangibility.

READ THE WORKING PAPER NO. 2025-92 · JULY 2025

Debt and Assets bfi.uchicago.edu/working-papers/debt-and-assets

Sept. 2019

Dec. 2019

March 2020

June 2020

Sept. 2020

Note: This figure plots the sensitivity of investment-grade unsecured bond spreads to the issuing firm’s unpledged tangibility share for each month from September Note: This2020 figure the sensitivity of investment-grade 2019 to September (i.e.,plots from six months before March 2020 to six months after March 2020). The coefficient estimate is negative but small before the onset bond spreads thenegative issuing firm’s unpledged of the unsecured Covid pandemic, it becomes muchto more after the onset of the pandemic, slowly returning to normal levels by year end.

tangibility share for each month from September 2019 to September 2020 (i.e., from six months before March 2020 to six months after March 2020). The coefficient estimate is negative but small before the onset of the Covid pandemic, it becomes much more negative after the onset of the pandemic, slowly returning to normal levels by year end.

ABOUT OUR SCHOLAR

Raghuram Rajan

Katherine Dusak Miller Distinguished Service Professor of Finance, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

169

RESEARCH BRIEF • NOVEMBER 2025

The Mortgage Debt Channel of Monetary Policy when Mortgages are Liquid Based on BFI Working Paper No. 2025-139, “The Mortgage Debt Channel of Monetary Policy when Mortgages are Liquid,” by Matthew Elias, University of Chicago and e61 Institute; Christian Gillitzer, The University of Sydney; Greg Kaplan, University of Chicago, e61 Institute, and NBER; Gianni La Cava, e61 Institute; and Nalini Prasad, UNSW Sydney

Despite the aggressive post-pandemic rate hikes, which pushed Australian mortgage rates up by over 4 percentage points and monthly payments up by $13,800, adjustable-rate borrowers did not cut their spending because they tapped into large savings buffers they had built up during the pandemic. Starting in 2022, central banks around the world Figure 1of · Distribution of Interest Rates Distribution Interest Rates raised interest rates more aggressively than they 8% Mortgage Rate had in over 50 years. These policies have direct Reserve Bank impacts on consumers. When the central bank of Australia Begins Tightening Cycle raises interest rates, mortgage rates follow, along 6 with required monthly payments, leaving families Adjustable Rate Mortgages with less money for spending. In this paper, the authors investigate how the post-pandemic 4 monetary tightening affected household spending. The authors study this question in the context of Australia, a particularly interesting setting to study this channel of monetary policy transmission. Most Australian mortgages are adjustable-rate, meaning payments reset quickly when the central bank changes rates. This would normally make consumer spending particularly sensitive to rate hikes. However, Australian adjustablerate mortgages have an unusual feature: “redraw facilities” that allow borrowers to park extra payments and withdraw money anytime, essentially turning their mortgages into savings accounts. Headed into 2022, households had built up unusually large savings in these accounts thanks to pandemic stimulus and lockdowns.

2 Fixed Rate Mortgages

0 Sept. 2021

Dec. 2021

March 2022

June 2022

Sept. 2022

Dec. 2022

March 2023

June 2023

Sept. 2023

Dec. 2023

March 2024

Note: This showsshows the evolution of theofdistribution of interest ratesrates for adjustable andand Note:figure This figure the evolution the distribution of interest for adjustable fixed-rate mortgages over time. vertical slice represents the full of rates at at fixed-rate mortgages over Each time. Each vertical slice represents thedistribution full distribution of rates that point in time, with the dashed horizontal line showing the mean.

that point in time, with the dashed horizontal line showing the mean.

The authors use transaction data from a large Australian bank covering over 83,000 customers’ mortgage accounts, checking and savings accounts from November 2020 through April 2024. They compare borrowers with adjustablerate mortgages to those with fixed-rate

bfi.uchicago.edu

170

mortgages, tracking what happened when the Reserve Bank of Australia raised its policy rate by 425 basis points, from 0.1% to 4.35%, between May 2022 and November 2023. They find the following: •

•

Adjustable-rate borrowers saw their cumulative mortgage payments increase by $13,800 compared to fixed-rate borrowers between August 2022 and April 2024. Despite this large increase in repayments for adjustable-rate borrowers, there was little change in non-durable, durable and services spending for adjustable-rate mortgagors relative to fixed rate borrowers. Adjustable-rate borrowers covered 70% of their higher payments by drawing down their savings (mostly from mortgage redraw accounts). The remaining 30% came from other sources. Before rates started rising, households had built up substantial savings cushions. During the pandemic, extra mortgage payments averaged 42% of required payments, compared to 26% in the five years before the pandemic. Only about 7% of adjustable-rate borrowers were living paycheck to paycheck when rates started rising in 2022, down from 13% in 2018. The authors’ estimates suggest that if the Reserve Bank had kept rates at zero through

READ THE WORKING PAPER NO. 2025-139 · NOVEMBER 2025

The Mortgage Debt Channel of Monetary Policy when Mortgages are Liquid bfi.uchicago.edu/working-papers/the-mortgage-debtchannel-of-monetary-policy-when-mortgages-are-liquid

Figure 2 · Mortgage Payments Spiked, But Spending Stayed Payments Flat Mortgage Spiked, But Spending Stayed Flat 100% of Income

Increase in Required Repayments

90 Average Spending, Non-Mortgagor

80 70 60 50

Average Spending, Mortgagor

This gap shows households absorbed higher payments without cutting spending

40 30

Mortgage Repayments

20 10

January 2021

July 2021

January 2022

July 2022

January 2023

July 2023

January 2024

Note: ThisThis figure shows average spending of mortgagors and non-mortgagors as a share Note: figure shows average spending of mortgagors and non-mortgagors as a share of of mean income for each group. Repayments is mean mortgage repayments by mortgagors, mean income for each group. Repayments is mean mortgage repayments by mortgagors, as as a share of mean mortgagor income.

a share of mean mortgagor income.

2024 (as originally signaled), total household spending would have been at most 1% higher at its peak, far below the 5% that standard economic models would predict. This research underscores the importance of structural features that make mortgages liquid. The authors caution that this result is somewhat specific to this episode, when Australian mortgage holders happened to have unusually large savings buffers. In other periods when those buffers are smaller, rising rates might pack more punch.

ABOUT OUR SCHOLARS

Matthew Elias

PhD Student in Economics, Kenneth C. Griffin Department of Economics

Greg Kaplan

Alvin H. Baum Professor, Kenneth C. Griffin Department of Economics and the College

Written by Abby Hiller • Designed by Maia Rabenold

171

RESEARCH BRIEF • NOVEMBER 2025

Who Pays for Tariffs Along the Supply Chain? Evidence from European Wine Tariffs Based on BFI Working Paper No. 2025-137, “Who Pays for Tariffs Along the Supply Chain? Evidence from European Wine Tariffs,” by Aaron B. Flaaen, Federal Reserve Board of Governors; Ali Hortaçsu, University of Chicago; Felix Tintelnot, Duke University; Nicolás Urdaneta, Duke University; and Daniel Xu, Duke University

Although foreign producers partially absorbed the 2019 US tariffs on European wines by lowering their prices, domestic markups amplified the cost as it moved through the supply chain, ultimately causing US consumers to pay more in dollar terms than the government collected in tariff revenue. Price effects took nearly a year to fully materialize at retail, and the wine industry engaged in “tariff engineering” by relabeling products to avoid duties. Who bears the burden of tariffs? In light of recent tariffs on imports, policymakers, researchers, and the public have debated how tariff costs are distributed among foreign producers, domestic importers, trade intermediaries, and final consumers. While existing research suggests that import prices (inclusive of tariffs) tend to increase

with tariffs, it remains unclear how these costs are transmitted to consumers. This paper resolves this disconnect by tracing tariff impacts on prices throughout the entire supply chain, from foreign producers through importers, distributors, and retailers to final consumers.

Figure 1 · Who Pays for Tariffs? The Hidden Markups Behind a $5 Bottle of Wine

bfi.uchicago.edu

172

Figure 2 · Alcohol by Volume Switching

The authors examine US tariffs imposed on European wines in October 2019 as part of the Airbus-Boeing subsidy dispute. The policy levied a 25% tariff specifically on still wines with ≤14% alcohol by volume (ABV) from France, Germany, Spain, and the United Kingdom. Using confidential transaction-level data from a major wine importer matched to exporter and downstream distributor and retail prices, the researchers compare price changes for tariffed wines (still wines ≤14% ABV) against a control group of non-tariffed wines (still wines >14% ABV and sparkling wines) from producers that sold no tariffed products.

only $1.19 in tariffs—a dollar pass-through exceeding 100%. Even accounting for statistical uncertainty across all stages, the consumer dollar cost per dollar of tariff revenue exceeded 68% with 90% confidence. •

Price effects emerged gradually, taking nearly a year to reach consumers. Import prices began declining three months after tariffs took effect, as foreign suppliers adjusted their pricing. The importer’s prices to distributors increased around the same time. However, retail prices did not rise until approximately 10-12 months after the tariffs were imposed, and remained elevated well beyond when tariffs were suspended in March 2021. This lag structure reflects inventory management, contract timing, and the multiple stages goods traverse before reaching consumers.

•

Tariff engineering created compositional bias in trade statistics. Immediately after tariffs took effect, the share of new wine label applications for products >14% ABV from France jumped by 40 percentage points. Roughly one-quarter of this increase came from existing products switching their reported ABV from ≤14% to >14%, crossing the tariff threshold without necessarily changing the wine itself. This strategic relabeling demonstrates how firms adapt product characteristics to minimize tariff exposure and shows that standard unit value measures from customs data can produce misleading pass-

They find the following: •

Consumers paid more than the tariff in dollar terms. Foreign producers lowered their prices by 5.2% following the 25% tariff, absorbing roughly one-quarter of the tariff burden. However, because the producer’s price decline was much smaller than the 25% tariff rate, the importer still faced a net cost increase—paying a lower pre-tariff price but a higher tariffinclusive price overall. As these costs moved through the supply chain, domestic markups amplified the price increase. The importer raised prices to distributors by 5.4%, absorbing some of the tariff through lower margins but passing most of the cost downstream. Retail prices ultimately rose by 6.9%. For a wine initially priced at $5 at the border, consumers paid $1.59 more per bottle while the government collected

173

through estimates when product composition shifts systematically. These findings offer important lessons for policymakers evaluating trade policy, particularly as they interpret pass-through estimates for tariffs imposed in 2025. Conventional measures of tariff pass-through, expressed in percentage terms, can substantially understate the actual cost burden on consumers when goods pass through multiple distribution stages with significant markups. The timing of price adjustments also matters, as the nearly year-long lag before retail prices fully adjust means tariffs can influence inflation well after implementation. Finally, the systematic product reclassification to avoid tariffs demonstrates how firms strategically respond to trade policy through product adaptation rather than just pricing adjustments, potentially limiting revenue collection and distorting trade statistics.

More On This Subject In “Why is Trade Not Free? A Revealed Preference Approach,” Rodrigo Adão and coauthors show that redistributive trade protection accounts for a significant fraction of tariff variation in the US and causes large monetary transfers between US individuals, mostly driven by differences in the social value of transfers across individuals employed in different sectors. In “Production, Relocation, and Price Effects of US Trade Policy: The Case of Washing Machines,” Ali Hortaçsu and coauthors show that 2018 global tariffs applied to all washers imported to the US increased prices about 12 percent for both washers and dryers, a complementary good not subject to tariffs. In “Tariffs, Trade, and a Misused Model,” Brent Neiman discusses how the Trump administration misused his research to justify tariff policy.

READ THE WORKING PAPER NO. 2025-137 · OCTOBER 2025

Who Pays for Tariffs Along the Supply Chain? Evidence from European Wine Tariffs bfi.uchicago.edu/working-papers/who-pays-for-tariffs-alongthe-supply-chain-evidence-from-european-wine-tariffs

ABOUT OUR SCHOLAR

Ali Hortaçsu

William B. Ogden Distinguished Service Professor of Economics, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

174

RESEARCH BRIEF • DECEMBER 2025

Comparing the Impacts of Cash vs. SNAP on Consumption of Drugs and Alcohol Based on BFI Working Paper No. 2025-145, “Paternalistic Social Assistance: Evidence and Implications from Cash vs. In-Kind Transfers,” by Anna Chorniy, Mount Sinai; Amy Finkelstein, Massachusetts Institute of Technology; and Matthew J. Notowidigdo, University of Chicago

Emergency department visits for drug and alcohol use increase by 20-30% following cash benefit receipt but do not respond to food stamp receipt. This is the first direct comparison showing that cash and food stamps affect consumption differently in the same population.

This paper brings new evidence to this debate. The authors provide the first direct test of whether cash transfers and food stamps lead to different consumption patterns in the same population. Using two decades of data from South Carolina, the authors track individuals receiving both cash benefits from Supplemental Security Income (SSI) and in-kind benefits from SNAP. They link benefit receipt records to detailed health care utilization data, examining how outcomes change in the days following each transfer’s scheduled monthly payout.1

Figure 1 · Effects of SNAP and SSI on Drug and Alcohol ED Visits Effects of SNAP and SSI on Drug and Alcohol ED Visits Change in Drug/Alcohol ER Visits Per 10K

Why do governments provide food stamps instead of cash? Classic economic theory suggests cash is superior because it allows recipients to optimize their spending. Yet in-kind programs like SNAP (Supplemental Nutrition Assistance Program) dominate US welfare spending. Voters and policymakers prefer such programs because they worry cash will be spent on drugs, alcohol, or other “inappropriate” purchases. But is this concern justified?

Emergency department (ED) visits for drug and alcohol use increase by 20-30% in the week

0.6 0.4

Cash Benefits (SSI)

0.2 0 -0.2 -0.4

Food Vouchers (SNAP) Day of Payout

-0.6 -0.8 -1 -10

-8

-6

-4

-2 0 2 4 Days Before and After Payout

Note: This change in emergency department visits visits for drug alcohol use around Note: Thisfigure figureshows showsthe the change in emergency department forand drug and alcohol use around the the time of benefit receipt. The x-axis shows days relative to benefit payout (day 0 = payout day). time of benefit receipt. x-axis shows days relativeand to benefit (day 0effect = payout day). The green The green line shows the The effect for SSI (cash) recipients, the blue payout line shows the for SNAP (food stamps) measuredand perthe 10,000 where person-days line shows therecipients. effect forOutcomes SSI (cash)are recipients, blueperson-days, line shows the effect for SNAP (food stamps) represents the total number of days individuals were observed (e.g., 10 people observed for 10 days recipients. Outcomes are measured per 10,000 person-days, where person-days represents the total = 100 person-days). number of days individuals were observed (e.g., 10 people observed for 10 days = 100 person-days).

following SSI receipt, but do not respond to SNAP receipt. Even after adjusting for the fact that SSI benefits are about four times higher than SNAP benefits, the authors confirm that cash and SNAP appear to have significantly different impacts on consumption of drugs and alcohol.

They find the following: •

1 0.8

•

Fills of prescription drugs for new illnesses also increase by 20-40% following SSI receipt but do not respond to SNAP receipt.

1 SSI benefits are paid on the first of the month (or the preceding weekday if the first falls on a weekend). SNAP benefits are paid on one of 15 days between the 1st and 19th of the month, determined by the last digit of the recipient’s case number.

in-kind transfer: Government assistance provided as specific goods or services (such as food vouchers, housing subsidies, or health care) rather than unrestricted cash. Recipients can only use these benefits for designated purposes. SNAP (Supplemental Nutrition Assistance Program): The US federal program that provides food assistance to low-income individuals and families. Formerly known as food stamps, SNAP provides electronic benefits that can be used only to purchase eligible food items at authorized retailers.

bfi.uchicago.edu

175

This approach reveals the following: •

•

When recipients have self-control problems, the government should provide some of each benefit type. The worse people’s self-control issues, the more welfare assistance should come as food stamps rather than cash. Additionally, the stronger people’s mental accounting, the less welfare assistance should come as food stamps. This is because when mental accounting is strong, recipients increase their food spending nearly dollar-for-dollar with their SNAP benefit. Since each SNAP dollar is so effective at increasing food consumption, the government needs less SNAP to achieve its target food consumption level and can provide more of the transfer as cash. The authors also consider whether a ‘sin tax’ on sin tax drugs and alcohol could replace the need for food stamps. They find that even with access to sin taxes, providing some benefits as SNAP may still be optimal. This is because a uniform sin tax treats everyone identically, while people vary

Figure 2 SNAP · Effects and Effects of and of SSISNAP on First FillsSSI on First Fills 150

Change in New Prescription Fills Per 10K

These findings reveal an important result: Cash and food vouchers are not used interchangeably. Since most SNAP recipients already spend more on food than their benefit amount, food stamps should free up cash for other purchases. If this were true, SNAP would increase spending on temptation goods, just as cash does. The fact that it doesn’t suggests recipients mentally treat SNAP as designated “food money.” The authors incorporate this mental accounting, accounting along with self-control problems (over consumption of temptation goods) into a framework for selecting the optimal combination of cash and SNAP for a fixed-budget transfer program.

100 50 Food Vouchers (SNAP)

0 -50

Cash Benefits (SSI) Day of Payout

-100 -150 -10

-8

-6

-4

-2 0 2 4 Days Before and After Payout

Note: This figure shows the change in first fills of prescriptions around the time of benefit Note: This figure shows the change in first fills of prescriptions around thereceipt. time of benefit First fills are prescriptions in a therapeutic class for which the recipient had no fills in the previous receipt. First fills are in a therapeutic classThe forx-axis which the recipient had no fills six months, capturing newprescriptions medical needs rather than routine refills. shows days relative to the benefit payout (day 0 = payout day). The green shows needs the effect for SSI (cash) recipients, and The six months, in previous capturing new line medical rather than routine refills. the blue line shows the effect for SNAP (food stamps) recipients. Outcomes are measured per 10,000 x-axis shows days relative to benefit payout (day 0 = payout day). The green line shows the person-days, where person-days represents the total number of days individuals were observed. effect for SSI (cash) recipients, and the blue line shows the effect for SNAP (food stamps) recipients. Outcomes are measured per 10,000 person-days, where person-days represents the total number of days individuals were observed.

dramatically in their self-control problems and mental accounting behavior. SNAP, by contrast, allows for targeted assistance. Recipients who struggle more with self-control can receive a higher share of benefits as food stamps, while those with better self-control receive more cash. The findings have direct implications for ongoing policy debates about the mix of cash versus in-kind assistance. While the authors focus on paternalistic social policy in the United States, similar policies exist worldwide. For instance, in Brazil, concerns that a large share of a cash transfer program for the poor (Bolsa Familia) was being spent on on-line gambling recently prompted the government to prohibit use of cash transfer program cards for online betting. The framework developed here could be adapted to analyze such policies whenever self-control problems and mental accounting shape spending decisions.

mental accounting: The psychological phenomenon whereby people treat money differently depending on its source or designated use, even when all money is objectively fungible (interchangeable). For example, people may treat a tax refund differently from regular wages, or treat “food money” differently from other income, leading to spending patterns that deviate from standard economic predictions. sin tax: A tax levied on goods or activities considered harmful or undesirable, such as tobacco, alcohol, or gambling. Also called excise taxes or Pigouvian taxes, these are designed to discourage consumption by increasing prices, while potentially generating revenue to address negative externalities associated with the taxed goods.

READ THE WORKING PAPER NO. 2025-145 · NOVEMBER 2025

Paternalistic Social Assistance: Evidence and Implications from Cash vs. In-Kind Transfers bfi.uchicago.edu/working-papers/paternalistic-socialassistance-evidence-and-implications-from-cash-vs-inkind-transfers

ABOUT OUR SCHOLAR

Matthew Notowidigdo

David McDaniel Keller Professor of Economics and Business and Public Policy Fellow, Chicago Booth

Written by Abby Hiller • Designed by Maia Rabenold

176

RESEARCH BRIEF • DECEMBER 2025

Dynamic Competition for Sleepy Deposits Based on BFI Working Paper No. 2025-128, “Dynamic Competition for Sleepy Deposits,” by Mark L. Egan, Harvard University; Ali Hortaçsu, University of Chicago; Nathan A. Kaplan, Harvard University; Adi Sunderam, Harvard University; and Vincent Yao, Georgia State University

Most bank depositors rarely shop for better rates and stay with the same bank for nearly a decade, on average. This “sleepiness” accounts for more than half of banks’ deposit franchise value while also creating stability in the banking system. People tend to leave their money in the same bank account over long durations, even when they could earn more by moving it. This “sleepiness” is considered a critical feature of the banking sector, as it bolsters stability and allows banks to use deposit funds toward long-term loans. At the same time, regulators have scrutinized banks for not fully passing through interest rate increases to savers, even as they quickly raise lending rates. In this paper, the authors study the implications of sleepy deposits for bank competition, value, and financial stability in the US banking sector. The authors begin by documenting depositor behavior using a novel dataset from Fiserv, a major financial technology company that processes accounts for banks and credit unions. The data cover 12 million deposit accounts from 89 financial institutions, tracking when accounts open and close, and why customers leave. They combine these data with detailed information on deposit volumes at every bank branch in America and weekly data on the interest rates banks offer. The authors document the following concerning depositors’ shopping behaviors: •

Depositors rarely switch accounts. Ninety-four percent of depositors keep their bank account each year, consistent with estimates from practitioners. New accounts make up 5-15% of total accounts, implying that the average life of an account is 8-9 years.

Figure 1 · Total Deposit Franchise Value in the Banking Sector Total Deposit Franchise Value in the Banking Sector $1.75 Trillion 1.5 1.25 1 0.75

Baseline

0.5 0.25 0

Counterfactual

2005

2010

2015

2020

Note: This figure shows time series of aggregate deposit franchise value under the authors’ Note: This figure shows time series of aggregate deposit franchiseare value under the authors’ baseline estimates and under a counterfactual policy in which all depositors always active.

baseline estimates and under a counterfactual policy in which all depositors are always active.

•

Less than 20% of account closures are a result of depositors shopping for better terms. The majority are driven by account inactivity, moving, and death.

•

Turnover is higher among business and trust accounts compared to individual accounts. Larger accounts, measured by account balances, also appear to be more active. In contrast, accounts held by elderly individuals are less active. Somewhat surprisingly, accounts set up for online banking tend to experience less turnover. Part of this appears to be driven by moving, which is inherently a less important factor for online depositors.

•

Sleepiness is negatively correlated with the lagged federal funds rate—people “wake up” (relatively speaking) when interest rates

bfi.uchicago.edu

177

increase. A one percentage point increase in the short rate reduces the fraction of depositors who are inactive by 0.4 percentage points. To understand how this sleepiness shapes bank competition, the authors develop a model of the supply and demand for sleepy deposits. In the model, depositors are either “awake” (shopping for accounts) or “asleep” (staying put), and banks compete knowing that most customers will stay asleep for long stretches. Instead of constantly battling over every depositor with the best rates, banks face a dynamic choice: should they “invest” by offering attractive rates to attract new customers (knowing those customers will likely stick around for years), or “harvest” by offering meager rates to their existing sleepy customers (who probably won’t bother switching)?

roughly constant regardless of concentration. In concentrated markets, dominant banks harvest their sleepy base with low rates, but this creates opportunities for smaller banks to compete aggressively for active shoppers. •

Sleepiness accounts for more than half of banks’ deposit franchise value, roughly 58% on average.

•

Banks with low-quality deposit services as well as banks with high marginal costs benefit most from sleepiness. Moving from the 25th to the 75th percentile of product quality decreases sleepiness dependence by about 10.4 percentage points, while moving from the 25th to the 75th percentile of marginal costs increases sleepiness dependence by about 3.4 percentage points. This is because sleepiness allows low-quality, high-cost banks to charge relatively high markups without losing too many depositors in the near-term.

•

Sleepiness creates stability in the banking sector. For two large global banks in the US, the probability of default after the Federal Reserve’s 2022-2023 hiking cycle would have increased to more than 20% in a counterfactual without sleepy depositors.

The model reveals the following: •

•

Sleepiness raises profit margins for banks. Average markups (the spread between what banks earn on loans and what they pay on deposits) would fall by 53% if all depositors were always actively shopping for new accounts. Sleepiness also makes markups procyclical, meaning they rise when interest rates rise. Average markups hover around 31 basis points when rates are near zero but jump to 127 basis points when short-term rates reach 5%. This is because when rates are high, banks with large market shares choose to “harvest” their sleepy depositors by keeping rates low. Dynamic competition eliminates the usual relationship between market concentration and markups. In a standard model, banks in concentrated markets would charge higher markups than banks in competitive markets. But with sleepy depositors, markups stay

READ THE WORKING PAPER NO. 2025-128 · SEPTEMBER 2025

Dynamic Competition for Sleepy Deposits bfi.uchicago.edu/working-papers/dynamic-competitionfor-sleepy-deposits

These findings carry important implications for bank regulation. Regulators in several countries, most notably the UK’s Financial Conduct Authority (FCA), have sought to increase the extent to which banks pass through interest rates to their depositors. The authors’ estimates suggest this would benefit depositors by raising the rates they earn by about 36 basis points. But it would come at substantial cost to financial stability, particularly during periods of monetary tightening. In this way, policymakers face a tradeoff between protecting depositors from low rates and maintaining a stable banking system.

ABOUT OUR SCHOLAR

Ali Hortaçsu

William B. Ogden Distinguished Service Professor in Economics and the College, Kenneth C. Griffin Department of Economics

Written by Abby Hiller • Designed by Maia Rabenold

178

RESEARCH BRIEF • DECEMBER 2025

Social Pressure Drives Parents to Adopt AI that May Harm Students Based on BFI Working Paper 2025-144, “Social Dynamics of AI Adoption,” Leonardo Bursztyn, University of Chicago; Alex Imas, University of Chicago; Rafael Jiménez-Durán, Bocconi University; Aaron Leonard, University of Chicago; and Christopher Roth, University of Cologne

Parents feel pressured to adopt AI for their children’s education if they know that other children are using the technology, despite its uncertain long-term consequences. This social pressure outweighs parents’ personal desire to restrict AI use in classrooms. As students, parents, and schools grapple in real time with the presence of artificial intelligence (AI) in classrooms, one powerful incentive is driving its use by students: parents’ fear that their children will fall behind because other children are using AI. This fear of falling behind outweighs the considerable uncertainty surrounding the effects of AI on students, who may benefit from AI in the short run but who may also suffer long-run negative effects on their cognitive development and human capital outcomes. This drive for short-run gains at the risk of long-run costs can result in what the authors call rat race dynamics that drive the unrestricted adoption of new technologies. (This notion is akin to time consistency in policymaking.) To understand this short vs. long run tradeoff, the authors ask a key question: Do parental decisions to adopt educational AI tools for their kids reflect informed judgments about their potential risks for human capital formation, or are they primarily driven by social factors and anxiety about their children falling behind their peers? To answer this question, the authors investigate the adoption of AI tools through incentivized, preregistered experiments involving more than 2,000 parents of teenagers from the United States, Canada,

Figure 1 · Parents’ Willingness to Pay by Information Treatment and Peer Adoption

Note: This figure displays the mean willingness of parents to pay for a three-month subscription for premium AI (worth $60), conditional on peer adoption of AI tools (the percentage of other students who use AI) and the information treatment. This figure shows that parents are highly responsive to social context. When asked to state their willingness to pay (WTP) for premium AI under different scenarios of peer AI adoption (20%, 40%, 60%, or 80%), average WTP increases as take-up rises. When peer take-up increases from 20% to 80%, WTP increases by more than 60%. This effect is both economically large and highly statistically significant.

and the United Kingdom. The authors measure parental demand for advanced AI tools by eliciting parents’ willingness to pay (WTP) for a three-month subscription to a premium unrestricted AI education plan. This allows them to study how adoption rates among teenagers’ peers influence parental demand, and to show how beliefs about AI’s impact on

time consistency: the problem of time consistency looms large for policymakers. A time consistent policy is one where future a policymaker cannot act on an incentive to revoke a previously established policy. On the other hand, a policy that lacks time consistency would offer a future policymaker both the incentive and the means to break a policy commitment. The economists Fynn E. Kydland and Edward C. Prescott were awarded the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel in 2004 in part for their formative work on time consistency in economic policy. Their 1977 paper, “Rules Rather Than Discretion: The Inconsistency of Optimal Plans,” was hugely influential in monetary policymaking. In his Nobel lecture, Prescott cited the work of UChicago economist and Nobel Laureate, Robert Lucas, as especially important in the development of their ideas. See here for a useful primer on time consistency.

bfi.uchicago.edu

179

•

While information about the potential longrun AI-risk leads to a large negative shift in incentivized beliefs about the effects of AI on cognitive skills, such information does little to curb demand; that is, parents are still largely affected by teenage peer adoption.

•

Consistent with rat race dynamics, information about long-run harm increases parents’ preference for banning AI in education for all students.

•

Finally, the authors provide further evidence for these findings by revealing that a substantial portion of parents who support an AI ban still justify allowing their child to use AI due to fears of them falling behind.

What I think that you think can shape my actions UChicago economist Leonardo Bursztyn, one of the authors of this paper, has co-authored related work on the social dynamics of decision-making. This new paper provides insights into how the fear of falling behind motivates parents to take actions in line with other parents’ choices. In previous work, Bursztyn et al. provide insights into how the perceived ideas of others can affect the choices/actions that people take. Bursztyn’s paper, “Misperceptions About Others,” reveals that misperceptions about others’ views are widespread, that these misperceptions are disproportionately concentrated on one side relative to the truth, that these misperceptions are exaggerated when they pertain to “outsiders,” that people tend to think that “insiders” believe as they do. In another paper, “Misperceived Social Norms: Female Labor Force Participation in Saudi Arabia,” Bursztyn et al. show how misperceptions about others restrict women’s ability to work outside the home. By custom, Saudi men decide whether women in their families work outside the home, and privately, most men believe that women should be allowed to work. Those men, though, also think that other men do not share their views, so they are disinclined to allow women in their families to join the labor force. However, the authors show that when men are informed that other men agree on women and work, then those men are more open to the idea.

cognitive skills shape demand for advanced AI tools. Further, by randomly assigning parents to either a control group receiving information emphasizing AI’s short-run educational benefits, or a treatment group that additionally highlights potential long-run risks, they provide key insight into this technological rat race. They find the following: •

Parents’ WTP for AI tools increases by more than 60% as the proportion of their children’s peers who use AI increases from 20% to 80%. (See accompanying figure.)

READ THE WORKING PAPER NO. 2025-144 · NOVEMBER 2025

Social Dynamics of AI Adoption bfi.uchicago.edu/working-papers/social-dynamics-ofai-adoption

These findings confirm the authors’ rat race hypothesis: peer adoption ignites parental anxiety about falling behind, which accelerates more adoption despite potential long-term drawbacks. The rat race persists. Further, the incentives driving the rat race are so strong that individual level policy interventions are likely insufficient. Rather, to achieve socially optimal outcomes, educators, parents, and school boards could act in a coordinated fashion to employ AI tools. These lessons apply beyond education. There is a global rat race to employ AI among firms, countries, and institutions, and fear of falling behind may risk long-run costs that far outweigh short-run gains. For example, rat race dynamics could generate overinvestment in AI technology such as data centers as the fear of falling behind leads firm to invest even if the individual investment is unprofitable. Collectively, this is how investment bubbles form. Understanding how rat race dynamics interact with systemic technological change is imperative if we hope to manage the complex trade-offs between short-run gains and long-run consequences.

ABOUT OUR SCHOLARS

Leonardo Bursztyn

The Saieh Family Professor of Economics, Kenneth C. Griffin Department of Economics

Alex Imas

Roger L. and Rachel M. Goetz Professor of Behavioral Science, Economics and Applied AI and Vasilou Faculty Scholar, Chicago Booth

Written by David Fettig • Designed by Maia Rabenold

180

RESEARCH BRIEF • DECEMBER 2025

Stop Using Test Scores to Measure Test Results Based on BFI Working Paper No. 2025-143, “Do Test Scores Misrepresent Test Results? An Item-by-Item Analysis,” by Jesse Bruhn, Brown University; Michael Gilraine, Simon Fraser University; Jens Ludwig, University of Chicago; and Sendhil Mullainathan, Massachusetts Institute of Technology

Collapsing students’ responses to individual test questions to single scores discards useful information about teacher efficacy and student knowledge that predicts outcomes like graduation, discipline, and future earnings. Imagine a teacher evaluating an in-class math exercise. The teacher is likely to analyze each student’s answers: Who could use more help on long division? Who has mastered their multiplication tables? They are less likely to simply sort their students by their overall math ability based on their percentage of correct responses. Despite this intuition, when it comes to standardized testing, we tend to follow the first approach. We assume aggregate test scores sufficiently capture students’ ability, and use scores to inform high stakes decisions like which students to give extra help to or which teachers to hire or fire. Schools invest as much as 18% of student time spent on testing and preparation. In this paper, the authors examine whether this approach throws away useful information. They ask: Do individual responses to test questions better predict student outcomes than aggregate scores do? Using data from five million Texas students spanning eight years, the authors connect responses to 1.31 billion individual test questions with student outcomes ranging from course grades and disciplinary problems to graduation, college attendance, and earnings. They compare predictions made using individual questions to

Figure 1 · Teachers Ranked ‘Excellent’ Overall Show Surprising Weaknesses; Teachers Ranked ‘Poor’ Overall Teachers Ranked 'Excellent' Overall Show Surprising Weaknesses; Have Hidden Strengths Teachers Ranked 'Poor' Overall Have Hidden Strengths Ranked by Overall Score

Teacher Effectiveness by Test Question

Highest

Student Ranking Overall Rank one is highest

Student Ranking by Category

1000 2000 3000

Lowest Note: Each vertical line represents one of 2,280 fourth-grade math teachers in Texas in 2016. Teachers are Note: Each lineonrepresents one oftest 2,280 fourth-grade math teachers in Texas color-coded byvertical their ranking the overall average score (left column), then re-ranked for each test in 2016. content area are whilecolor-coded keeping their original color. If a single captured all teacher each Teachers by their ranking on measure the overall average test effectiveness, score (left column), then column would show the same smooth gradient from blue (top) to red (bottom). Instead, the dramatic color re-ranked content while keeping their original color.have If a specific single measure mixing revealsfor thateach 'good'test teachers have area specific weaknesses and 'struggling' teachers strengths—information that is completely lost when we look only at average scores. This hidden structure captured all teacher effectiveness, each column would show the same smooth gradient from represents 66% of the predictable variation in teacher performance.

blue (top) to red (bottom). Instead, the dramatic color mixing reveals that ‘good’ teachers have specific weaknesses and ‘struggling’ teachers have specific strengths—information that is completely lost when we look only at average scores. This hidden structure represents 66% of the predictable variation in teacher performance.

those made using aggregate scores—both simple averages and Item Response Theory scores (a slightly more sophisticated aggregate that weights harder questions more heavily than easier ones). For teachers, the authors measure effectiveness using

bfi.uchicago.edu

181

both approaches to determine whether aggregation hides meaningful performance patterns.

consistently differentiate teachers based on their patterns of comparative advantage. Using these item categories, rather than aggregate test scores, to identify the bottom 5% of teachers produces 21% more graduates for the same number of teacher replacements.

They find the following: •

•

Aggregation destroys substantial information about both students and teachers. When teachers are ranked by their effectiveness using aggregate scores, then re-ranked based on individual test questions, the rankings shift dramatically. The correlation between the two rankings is only 0.66 for teachers and 0.75 for students, far from the 1.0 that would indicate perfect agreement. The lost information matters for high-stakes teacher decisions. When identifying which teachers fall in the bottom 5% for their effectiveness at improving student outcomes, aggregate scores and item-level data point to different teachers. The two approaches disagree 51.6% of the time when measuring teachers’ impact on student class failure rates, 42.4% for student disciplinary infractions, 44.9% for student graduation, and 39.6% for student college attendance. Question-level data improves decisions about which teachers to replace. Because test questions change year-to-year, the authors develop a method to categorize similar items across years by identifying which types of questions

•

Switching to item-level analysis is highly costeffective. Schools already collect individual question responses, but discard them before analysis. The additional cost of storing and analyzing item-level rather than aggregate data is negligible, yielding an infinite marginal value of public funds at implementation costs below $4 million annually.

It’s time to rethink how we use testing data in both research and practice. This study shows that aggregating test responses into single scores discards information that could improve critical decisions about students and teachers. Education is not alone in this problem. Aggregation occurs throughout economics—in health, housing, transportation, crime, public finance, consumer finance, and even inflation measurement. Evidence that aggregation destroys valuable information in education suggests we should question whether it’s harmless in these other domains.

Marginal Value of Public Funds (MVPF): The MVPF is designed to measure long-run policy effectiveness. It is calculated as the ratio of two numbers: the benefits that the policy provides, divided by the government cost. The numerator (benefits) captures the extent to which the policy improves the lives of beneficiaries (described by economists as individuals’ “willingness to pay”), and the denominator reflects net government cost.

READ THE WORKING PAPER NO. 2025-143 · NOVEMBER 2025

Do Test Scores Misrepresent Test Results? An Item-by-Item Analysis bfi.uchicago.edu/working-papers/do-test-scoresmisrepresent-test-results-an-item-by-item-analysis

ABOUT OUR SCHOLAR

Jens Ludwig

Edwin A. and Betty L. Bergman Distinguished Service Professor, Harris School of Public Policy

Written by Abby Hiller • Designed by Maia Rabenold

KEEP UP WITH BFI

bfi.uchicago.edu