Journal of Personal Finance Volume 16 Issue 2

Volume 16 Issue 2 2017 www.journalofpersonalfinance.com

Journal of Personal Finance

Techniques, Strategies and Research for Consumers, Educators and Professional Financial Consultants

IARFC INTERNATIONAL ASSOCIATION OF REGISTERED FINANCIAL CONSULTANTS

IARFC INTERNATIONAL ASSOCIATION OF REGISTERED FINANCIAL CONSULTANTS

Biltmore Conference April 17 – 19, 2018 Asheville, NC

IARFC Invites You The International Association of Registered Financial Consultants invites you to mark your calendar for April 17-19, 2018 for the IARFC Biltmore Conference held on the Biltmore Estate in Asheville, NC. Opportunities will be available to network, attend CE sessions and judge the Finals of the IARFC Image used with permission from The Biltmore Company, Asheville, North Carolina

National Financial Plan Competition while enjoying the lavish Biltmore Mansion, grounds and gardens.

“We chose the Biltmore Estate for the combination of a beautiful backdrop and the successful entrepreneurial spirit of the Vanderbilts,” related IARFC Chairman H. Stephen Bailey, MRFC. “It’s not your ordinary big city convention-type visit. At the Biltmore you can attend your CE sessions, relax in the mountain setting and

Keynote Speaker Ric Edelman, RFC®

2018 IARFC BILTMORE Conference

& National Financial Plan Competition Asheville, NC April 17 – 19, 2018

appreciate the splendor of a unique period of history.”

To learn more call 800.532.9060 or visit www.IARFC.org

Journal of Personal Finance

Volume 16, Issue 2 2017 The Official Journal of the International Association of Registered Financial Consultants Š2017, IARFC. All rights of reproduction in any form reserved.

2

Journal of Personal Finance

Contents Editors’ Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Do Self-Control Measures Affect Saving Behavior?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Gui Jeong Kim, PhD, Senior Researcher, Samsung Life Insurance, Seoul, Korea Sherman D. Hanna, PhD, Professor, Department of Human Sciences, The Ohio State University, Columbus, OH We examine the effects of self-control mechanisms on saving behavior using the 2013 Survey of Consumer Finances (SCF), following the assumptions of research that analyzed the 1998 SCF. Self-control mechanisms include saving goals, foreseeable expenses, and saving rules. We find a positive effect of having one or more saving rules on the likelihood of saving, and weak effects of having retirement as a saving goal and of having children/ family as a saving goal on saving. However, it is not clear that the measures of self-control reported in previous research really provide useful ways to increase the likelihood of saving. We discuss implications for financial planning advice.

What’s Your Risk Appetite? Helping Financial Advisors Better Serve Clients (by Quantifying Kahneman-Tversky’s Value Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Sid Muralidhar, Grade 11, Thomas Jefferson High School for Science and Technology, Great Falls, Virginia Emerson Berlik, Grade 11, Thomas Jefferson High School for Science and Technology, Round Hill, Virginia This paper presents a methodology to allow advisors to quantify risk tolerance of clients, over gains and losses, based on the Kahneman-Tversky survey. Once a formal and quantitative estimate of an individual’s risk appetite can be determined, and its evolution tracked over time, advisors can design effective investment portfolios to cater to the client’s specific risk tolerance. The paper extends this individual-level risk diagnostic and applies it over various subgroups and demonstrates that (a) teens are more risk-seeking than adults when it pertains to losses; (b) among investment professionals, women are more conservative than men when it pertains to gains; and (c) even within these subgroups, every individual is unique and neither expected utility theory nor prospect theory appropriately capture the diversity in risk tolerance. This paper seeks to make Kahneman-Tversky’s research on prospect theory/behavioral economics, and their value function practical and user-friendly, thus improving investment decision making.

Life Quality and Health Costs in Late Retirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Yuanshan Cheng, Ph.D., Assistant Professor, Winthrop University, Rock Hill, SC Philip Gibson, Ph.D., Assistant Professor, Winthrop University, Rock Hill, SC Tao Guo, Ph.D., Assistant Professor, William Paterson University, Wayne, NJ Individuals are living longer due to the advancement of medical technology and nutrition quality. Are the elderly enjoying retirement in those extended years with good quality of life or are they simply alive? Using data from the Health and Retirement Study (HRS) and the Consumption and Activities Mail Survey (CAMS), this study contributes to the literature by presenting empirical evidence on how individuals spend time in retirement. The results show that retirees on average do not spend their time significantly different throughout retirement. Most life tasks such as reading the paper or magazines, listening to music, playing sports or exercising, visiting others, and house cleaning are similar among retirees in different age groups. We also present evidence that retirees on average experience a spike in medical expenses late in retirement. We compare systematic withdrawal strategies with and without health costs risk quantifying the impact on portfolio sustainability.

The Impact of Product Knowledge and Quality of Care on Long-term Care Insurance Demand: Evidence from the HRS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Jacob Lumby, Texas Tech University, Lubbock, Texas Christopher Browning, PhD, BS Program Co-Director, Assistant Professor of Personal Financial Planning, Texas Tech University, Lubbock, Texas Michael S. Finke, PhD, Dean and Chief Academic Officer, The American College of Financial Services, Bryn Mawr, PA

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Using a unique module in the Health and Retirement Study (HRS), this paper considers three important factors that may influence consumer demand for long-term care insurance (LTCI): preference for high quality care, potential costs, and knowledge. In addition, this paper proposes a new method for examining insurance demand.

Only those individuals who are considering purchasing LTCI in the near future (who don’t currently own a LTCI policy) are included in the analysis. By focusing on this group, this paper attempts to determine the factors that are most relevant to the LTCI purchase decision when the consumer is most heavily considering it. Our findings imply that consumers deeply care about the provision for high quality long-term care, and suggest that widespread informational deficiencies currently suppress the demand for private long-term care insurance.

Why A QLAC in an IRA Is a Terrible Way to Defer the Required Minimum Distribution (RMD) Obligation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 Michael E. Kitces, MSFS, MTAX, CFP®, CLU, ChFC, RHU, REBC, CASL, Partner and the Director of Wealth Management for Pinnacle Advisory Group, Columbia, Maryland The longevity annuity has become increasingly popular in recent years as a potential new vehicle for retirement income, as its ability to delay payments to an advanced age like 85 allows for a significant accumulation of mortality credits. And since the introduction of Treasury Regulations in 2014, a so-called “qualified longevity annuity contract” (QLAC) can even be purchased inside of an IRA or other retirement account, allowing a portion of a retiree’s RMDs to be deferred from 70½ to as late as age 85. However, as it turns out the unique nature of a longevity annuity’s payment structure is not very hospitable as an RMD deferral strategy. The fact that it can take until a retiree’s late 80s just to break even and recover principal means the retiree risks significant forgone growth by trying to merely defer RMDs through the use of a QLAC. And of course, the RMDs will still eventually happen anyway, as the QLAC merely defers when payments begin. In fact, ironically, if the retiree does live, the accelerated payments of a QLAC in the later years can actually deplete an IRA even faster than normal IRA RMDs. Ultimately, this doesn’t mean that the longevity annuity (or a QLAC inside an IRA) is a bad deal. The ability to accumulate mortality credits still means it can be very effective as a fixed income alternative for those who fear they may not have enough money to fund a retirement well beyond their life expectancy. And if a retiree intends to spend all of his/her assets anyway, and the only available dollars for retirement are held in an IRA or other retirement account, the QLAC is an effective means to engage in such a strategy. Nonetheless, the bottom line is that while a QLAC may be a valid way to use a retirement account to hedge against longevity—and defer RMDs along the way—it’s still not very effective as an RMD avoidance or deferral strategy. Just because you can buy a longevity annuity inside a retirement account as a QLAC doesn’t mean you should.

Is Deferring Social Security the Lowest Cost Option for Adding Guaranteed Income? . . . . . . . . . . . . . . . . . . 69 David A. Littell, JD, ChFC, Chairholder of the Joseph E. Boettner Chair in Research, Co-Director of the New York Life Center for Retirement Income, The American College of Financial Services, Bryn Mawr, PA Kirk S. Okumura, MSFS, ChFC, Academic Director of the Financial Services Certified Professional® (FSCP®) Program, The American College of Financial Services, Bryn Mawr, PA Editor’s Note: Although the Journal of Personal Finance is a refereed, research journal that targets the practitioner market, the editors believe there is a role in the literature for complex cases that are instructive or allow practitioners to compare their insights into a planning scenario with how others would approach the same situation.

In last year’s Fall Issue, we published the winning case for the IARFC National Financial Plan Competition, which was a multifaceted planning case. In this issue, we are delighted to present an analysis of a retirement planning issue involving Social Security that is extremely important in today’s world where so many people are now having to make decisions about when to start the Social Security retirement benefits. The selection and publication of cases is on an editorial basis rather than a refereed basis. We would like to publish one case with each issue, and welcome submissions of any planning situations our readers have encountered such that they feel the cases would provide a beneficial learning opportunity for others in the profession. We also invite readers to submit comments on any of the case presentations.

Wade Pfau and Walt Woerheide, Co-Editors CE Exam for Members of the IARFC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

3

Journal of Personal Finance

4

Call for Papers Journal of Personal Finance (www.JournalofPersonalFinance.com) Overview The Journal of Personal Finance is seeking high quality submissions that add to the growing literature in personal finance. The editors are looking for original research that uncovers new insights—research that will have an impact on advice provided to individuals. The Journal of Personal Finance is committed to providing high quality article reviews in a single-reviewer format within 60 days of submission. Potential topics include: •

Household portfolio choice

•

•

Retirement planning and income distribution

Investment research relevant to individual portfolios

•

Household credit use

•

Individual financial decision-making

•

•

Household risk management

Professional financial advice and its regulation

•

Life-cycle consumption and asset allocation

•

Behavioral factors related to financial decisions

•

Financial education and literacy

Please check the “Submission Guidelines” on the Journal’s website (www.JournalofPersonalFinance. com) for more details about submitting manuscripts for consideration.

Contact Wade Pfau and Walt Woerheide, Co-Editors Email: jpfeditor@gmail.com www.JournalofPersonalFinance.com

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

5

Journal of Personal Finance Volume 16, Issue 2 2017 Co-Editors Wade Pfau, Ph.D., The American College Walt Woerheide, Ph.D., ChFC, CFP™, RFC®, The American College

Editorial Board Sarah Asebedo, Ph.D., Texas Tech

Karen Eilers Lahey, Ph.D., The University of Akron

H. Stephen Bailey, HB Financial Resources

Douglas Lamdin, Ph.D., University of Maryland Baltimore County

David Blanchette, Ph.D., Morningstar

Jean M. Lown, Ph.D., Utah State University

Benjamin F. Cummings, Ph.D., The American College

Lew Mandell, Ph.D., University of Washington

Dale L. Domian, Ph.D., CFA, CFP™, York University

Carolyn McClanahan, MD, CFP™, Life Planning Partners

Ric Edelman, Ric Edelman, Inc.

Yoko Mimura, Ph.D., California State University, Northridge

Michael S. Finke, Ph.D., CFP™, RFC®, The American College

Robert W. Moreschi, Ph.D., RFC®, Virginia Military Institute

Joseph W. Goetz, Ph.D., University of Georgia

David Nanigian, Ph.D., Mihaylo College at Cal State Fullerton

John Grable, University of Georgia

Barbara M. O’Neill, Ph.D., CFP™, CRPC, CHC, CFCS, AFCPE, Rutgers

Michael A. Guillemette, Ph.D., University of Missouri

Jeff Rattiner, JR Financial Group

Clinton Gudmunson, Ph.D., Iowa State University

Cliff Robb, Ph.D., Kansas State

Tao Guo, Ph.D., William Patterson University

Sandra Timmerman, Ed.D., Retirement Resource Center

Sherman Hanna, Ph.D., The Ohio State University

Jing Jian Xioa, Ph.D., University of Rhode Island

Douglas A. Hershey, Ph.D., Oklahoma State University

Rui Yao, Ph.D., CFP™, University of Missouri

Michael Kitces, Pinnacle Advisory Group

Yoonkyung Yuh, Ewha Womans University Seoul, Korea

Mailing Address: IARFC Journal of Personal Finance 1046 Summit Drive, P.O. Box 506 Middletown, OH 45042

Postmaster: Send address changes to IARFC Journal of Personal Finance, 1046 Summit Drive, P.O. Box 506 Middletown, OH 45042

Permissions: Requests for permission to make copies or to obtain copyright permissions should be directed to the Co-Editors.

Certification Inquiries: Inquiries about or requests for information pertaining to the Registered Financial Consultant or Registered Financial Associate certifications should be made to IARFC, 1046 Summit Drive, P.O. Box 506 Middletown, OH 45042.

Disclaimer: The Journal of Personal Finance is intended to present timely, accurate, and authoritative information. The editorial staff of the Journal is not engaged in providing investment, legal, accounting, financial, retirement, or other financial planning advice or service. Before implementing any recommendation presented in this Journal readers are encouraged to consult with a competent professional. While the information, data analysis methodology, and author recommendations have been reviewed through a peer evaluation process, some material presented in the Journal may be affected by changes in tax laws, court findings, or future interpretations of rules and regulations. As such, the accuracy and completeness of information, data, and opinions provided in the Journal are in no way guaranteed.

The Editor, Editorial Advisory Board, the Institute of Personal Financial Planning, and the Board of the International Association of Registered Financial Consultants specifically disclaim any personal, joint, or corporate (profit or nonprofit) liability for loss or risk incurred as a consequence of the content of the Journal.

General Editorial Policy: It is the editorial policy of this Journal to only publish content that is original, exclusive, and not previously copyrighted.

Subscription requests should be addressed to: IARFC Journal of Personal Finance 1046 Summit Drive, P.O. Box 506 Middletown, OH 45042 editor@iarfc.org 1-800-532-9060 Subscription Rates, 1yr, 2 issues, add $15 for delivery outside the U.S. Individual Subscription: Member $45, NonMember $65 Institutional: $120, 3 copies, ea. issue Single Copies: Member $25, Non-Members $35 The Journal of Personal Finance (ISSN 15406717) is published in the U.S. in the months of March and October by the International Association of Registered Financial Consultants (IARFC).

6

Journal of Personal Finance

Editors’ Notes Welcome to the Fall 2017 issue of the Journal of Personal Finance. We are excited to bring you another six articles representing cutting edge research and current thinking in the field of personal finance.

In our first article, Gui Jeong Kim and Sherman Hanna, Ph.D., examine the effects of self-control mechanisms on saving behavior.

These mechanisms include saving goals, foreseeable expenses, and saving rules. They find a positive effect of having one or more saving rules on the likelihood of saving, and weak effects of having retirement as a saving goal and of having children/family as a saving goal. They finish their analysis with a discussion of implications for financial planning advice.

Next, Sid Muralidhar and Emerson Berli present a methodology to allow advisors to quantify the risk tolerance of clients over gains and losses, based on a Kahneman-Tversky survey. This may help advisors to design effective investment portfolios that cater to the client’s specific risk tolerance. They note that every individual is unique and neither expected utility theory nor prospect theory appropriately captures the diversity in risk tolerance. Their objective is to make Kahneman-Tversky’s research on prospect theory/ behavioral economics more practical and user-friendly, thus improving investment decision-making.

Our third article is by Yuanshan Cheng, Ph.D., Philip Gibson, Ph.D., and Tao Guo, Ph.D. They investigate whether the elderly are enjoying retirement with good quality of life as their lifespans extend, or, whether they are simply alive. Their study contributes to the research literature by presenting empirical evidence on how individuals spend time in retirement. The results show that retirees on average do not spend their time in significantly different ways throughout retirement. Most life tasks such as reading the paper or magazines, listening to music, playing sports or exercising, visiting others, and house cleaning are similar among retirees in different age groups.

Next, Jacob Lumby, Christopher Browning, Ph.D., and Michael S. Finke, Ph.D., consider three important factors that may influence consumer demand for long-term care insurance (LTCI). These include a preference for high quality care, potential care costs, and consumer knowledge. In addition, their article proposes a new method for examining insurance demand. Their findings imply that consumers deeply care about the provision for high quality long-term care, and suggest that widespread informational deficiencies currently suppress the demand for private long-term care insurance.

In our fifth article, Michael E. Kitces, MSFS, MTAX, CFP®, CLU, ChFC, RHU, REBC, CASL, investigates the prudence of using qualified longevity annuity contracts (QLAC) as a means to extend required minimum distributions (RMD) in retirement. Contrary to conventional wisdom, he argues that the unique nature of a longevity annuity’s payment structure is not very hospitable as an RMD deferral strategy. His bottom line is that while a QLAC may be a valid way to use a retirement account to hedge against longevity—and defer RMDs along the way—it’s still not very effective as an RMD avoidance or deferral strategy.

Finally, David Littell and Kirk Okumura provide a case study about the efficacy of delaying the Social Security claiming age. They consider a hypothetical individual eligible for the maximum allowable Social Security benefit at the current full retirement age of 66 and assume that this person retires at age 65. They compare the costs of two equivalent options the retiree will have. These include deferring Social Security to age 70 and buying 5 years of additional income to cover the deferral period with a single premium immediate annuity, or to claim Social Security at age 65 and buy a commercial deferred income annuity that covers the additional income that would have been provided if benefits were deferred to age 70. By making this comparison they can determine which is the better deal, buying more income by deferring Social Security or by purchasing a commercial annuity. They find that for most singles and married couples it makes sense to use a portion of the assets earmarked for purchasing guaranteed income to defer Social Security.

We hope you enjoy the current issue of the Journal of Personal Finance.

— Wade Pfau — Walt Woerheide

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Do Self-Control Measures Affect Saving Behavior?

Gui Jeong Kim, PhD, Senior Researcher, Samsung Life Insurance, Seoul, Korea Sherman D. Hanna, PhD, Professor, Department of Human Sciences, The Ohio State University, Columbus, OH

Abstract We examine the effects of self-control mechanisms on saving behavior using the 2013 Survey of Consumer Finances (SCF), following the assumptions of research that analyzed the 1998 SCF. Self-control mechanisms include saving goals, foreseeable expenses, and saving rules. We find a positive effect of having one or more saving rules on the likelihood of saving, and weak effects of having retirement as a saving goal and of having children/family as a saving goal on saving. However, it is not clear that the measures of self-control reported in previous research really provide useful ways to increase the likelihood of saving. We discuss implications for financial planning advice.

Key Words: behavioral life cycle hypothesis, saving behavior, saving goals, saving rules, self-control, Survey of Consumer Finances

7

Journal of Personal Finance

8

Introduction

reason why people do not save enough for their retirement period is a lack of self-control to save for future consumption.

The normative life-cycle hypothesis tries to explain people’s saving behavior (Modigliani & Brumberg, 1954; Browning & Crossley, 2001). In this theory, a rational household will try to smooth consumption over a lifetime using saving or borrowing strategies. However, it is difficult for households to estimate future income, and calculating the optimal saving strategy is challenging. Many empirical studies have shown that people do not save enough before retirement (Kim & Hanna, 2015; Hanna, Kim, & Chen, 2016). An aggregate measure of personal savings, personal saving as a percent of disposable personal income, generally decreased from 1992 to 2005, and while the rate increased after the Great Recession, the rate is currently still substantially lower than the rates in years before 1992 (Figure 1). Rha, Montalto, and Hanna (2006) noted some criticisms of this aggregate measure, but the general decrease in the personal savings rate has often been cited in discussions about low levels of savings by U.S. households. An alternative theory to explain these empirical results is the behavioral life-cycle hypothesis (Shefrin & Thaler, 1988).

According to this theory, households do not behave rationally because of a lack of self-control. Thaler (1994) and Thaler and Shefrin (1981) suggested that individuals might have lower saving rates because they have low willpower. Also, people’s preferences vary over time due to lack of self-control and so people may postpone saving for retirement (Fisher & Montalto, 2010; Laibson & Harris, 2001; Laibson, Repetto, & Tobacman, 1998). Those earlier studies indicated that one

Rha et al. (2006) studied the effect of psychological factors, especially the role of self-control, on saving behavior. They used data from the 1998 Survey of Consumer Finances (SCF) and found that people with a motivation for saving had a higher probability of saving. Rha et al. (2006) also found that households with saving rules were more likely to save than those without saving rules. The psychological variables improved the explanatory power of the model significantly.

Griesdorn, et al. (2014, p. 25) stated: “Some of the strongest evidence on self-control and savings comes from a pioneering study by Rha et al. (2006) which demonstrated the strong link between savings and self-control mechanisms (mental accounting and savings rules) practiced by households.”

Since 1998, the U.S. experienced the Great Recession and the composition of the U.S. population has changed. For instance, in the 20 years before 1998, the rate of having retirement and education related reasons for saving had consistently increased, possibly due to the aging of the “baby boom” generation and the increasing number of children of the baby-boom generation approaching college age (Kennickell, Starr-McCluer, & Surette, 2000). In 1998, the U.S. was in the seventh year of an economic expansion, and the stock market had increased substantially over these seven years, with only short periods of decrease. In 2013, the U.S. economy was also in an economic expansion (Bricker et al., 2014) but memories

Figure 1. Personal Savings as a Percent of Disposable Personal Income, 1952–2016

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

9

Figure 2. Percent of Households Reporting Saving (Spending less than Income), and Mean Household Income, 1992–2013 SCF

of the “Great Recession” were still salient. Between 2010 and 2013, household income growth was limited for households not in the top part of the income distribution, and median net worth did not increase (Bricker et al., 2014), despite positive signs of recovery from the stock market. Retirement plan participation in 2013 was lower than in 2007 for families in the bottom half of the income (Bricker, et al., 2014). Therefore, households’ attitudes about saving might have changed over time. A number of macroeconomic shocks have taken place, including increases in mean household income from 1992 to 2001, then a further increase between 2004 to 2007 (Figure 2).

The proportion of households that saved (spending less than income) generally moved with changes in mean household income, going from 55 percent in 1995 to 56 percent in 1998 to 59 percent in 2001, but decreased to 52 percent by 2010, with a small increase in 2013 (53 percent).

Thus, the main objective of the current research is to update the empirical analysis using the most recent data available, and to ascertain whether the effects of behavioral factors on saving behavior changed between 1998 and 2013. While this is a replication, Strömbäck, et al. (2017) stated “…the relationship between self-control and financial behavior is still inconclusive.” The Open Science Collaboration (2015) found that many studies could not be replicated, and stressed the importance of replication studies (see also Handwerk, 2015). In addition to repeating the same analyses with different datasets, it is also useful to reconsider assumptions made in previous studies. As many authors have noted, many U.S. households do not save

enough to meet their retirement needs (Hanna, et al., 2016), so the question of why some households do not save continues to be an important question for research.

Using logistic regression, we found a positive effect of having one or more saving rules on the likelihood of saving, similar to the Rha et al. (2006) result with the 1998 SCF. We found weak effects of having a saving goal for retirement on saving, while Rha et al. found a stronger effect of having that goal. Unlike Rha et al. (2006), having precautionary, purchase, and education saving goals and foreseeable major expenses had no significant effects on the likelihood of saving, whereas having a goal of saving for children had a weak positive effect on saving in the 2013 SCF. We discuss some of the limitations of the Rha et al. approach.

Literature Review There have been many studies that analyzed factors related to household saving. One commonly used indicator of saving is based on a question in the SCF, whether a household reported spending less than income. Studies using this measure of saving have employed a variety of theoretical frameworks. Yuh and Hanna (2010) analyzed factors affecting the likelihood that households spent less than income, but their theoretical framework was based only on a normative economic model, taking into account the life-cycle savings model and optimal saving under uncertainty. Heckman and Hanna (2015) analyzed saving behavior of low-income households with the SCF, and used a model based on institutional theory, which states that

10

Journal of Personal Finance

institutional environments and conditions affect financial decisions of low-income households. Lee and Hanna (2015) also used the same measure of saving as a dependent variable, but their model was based on relating saving goals to Maslow’s hierarchy of needs.

According to the behavioral life-cycle hypothesis (Shefrin & Thaler, 1988), households do not behave rationally because of a lack of self-control, and may postpone saving for retirement (Fisher & Montalto, 2010; Laibson & Harris, 2001; Laibson, et al. 1998). As a solution for the lack of self-control, Shefrin and Thaler (1988), argued that a system of mental accounting and having rules can explain household behavior better than the normative life-cycle model. A number of authors have proposed models based on behavioral concepts, both to explain household behavior and to try to improve household decisions. Kim, Lee, and Hong (2016) discussed research on how commitment mechanisms can help households save more for retirement. Mental accounting and framing devices, while not consistent with the normative life-cycle model, may be ways that consumers can overcome their failure to save enough, for instance, by identifying a savings account as being only for emergencies (Griesdorn, et al., 2014). Using simple decision rules may be also a commitment device.

Rha et al. (2006) analyzed saving (spending less than income) in the 1998 SCF, and included variables based on the behavioral life cycle hypothesis. Rha et al. assumed that variables available in the SCF dataset, including whether households used a saving rule, whether they had reported particular goals for saving, and whether they stated that they had a major expense in the next 5 to 10 years, were all commitment devices in terms of the behavioral life-cycle hypothesis. They found that having one or more saving rules had a strong positive effect on the likelihood of saving. Having foreseeable major expenses had a positive effect on the likelihood of saving, as did having saving goals for retirement, for precautionary, or future purchases. Having a saving goal for education was negatively related to the likelihood of saving.

Methodology Dataset and Analysis In the current study, the 2013 Survey of Consumer Finances (SCF)—a nationally representative sample of U.S. households and the most recent data available—is used for the empirical

analysis. The survey provides detailed information on U.S. families’ balance sheets and sociodemographic characteristics.

Bricker et al. (2014) provide a detailed overview of the survey.

The SCF is the best U.S. survey for our research objectives because it has the key variables related to household perceptions and expectations, as well as very detailed information about household finances. This study includes all households in the 2013 SCF, and the total sample size is 6,015.

Our descriptive analyses were weighted using all five implicates to represent the U.S. population, while our logistic regression was not weighted following the methodological recommendation by Lindamood, Hanna & Bi (2007). There is some controversy about weighting of multivariate analyses with SCF datasets, but as Shin and Hanna (2016) showed, unweighted analyses produce more conservative tests of hypotheses.

For both descriptive analyses and the logistic regression, the repeated imputation inference (RII) method was used for proper estimation of standard errors (Lindamood, et al., 2007; Montalto & Sung, 1996). Following Rha et al. (2006), we compared a saving logit without behavioral life cycle (BLC) variables to a model with BLC variables added, using a likelihood ratio test. The procedure is shown in Appendix 2.

Measurements of Variables Dependent Variable To measure saving, we used the SCF question asking whether spending was greater than income, was about the same as income, or was less than income over the past year. It was coded as 1 if households reported that their spending, excluding any investments and a home or automobile purchase, was less than income over the past year (indicating households did save), and coded as 0 if households reported that their spending was equal to or more than income over the past year (indicating households did not save).

Independent Variables Following Rha et al. (2006), we included saving goals, having saving rules, and having foreseeable major expenses as independent variables representing self-control mechanisms.

Rha et al. categorized saving goals into five types: retirement, precautionary, children, purchases, and education, based on respondent listing of saving objectives, as shown in Appendix 1. The respondents could provide up to a maximum of six reasons (c.f., Lee & Hanna, 2015). If respondents reported any

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

of the five types of goals listed in Appendix 1, that saving goal variable was coded as 1, otherwise it was coded as 0.

For instance, if a respondent had a saving goal coded as 17 or 22 in the original SCF variable, that household was coded as having a retirement saving goal. Saving rules included the responses of “save one income and spend other,” “spend regular income and save other,” and “save regularly,” and the savings rule variable was coded as 1 if respondents reported having one or more of the rules and coded as 0 otherwise.

Having a foreseeable expense was coded as 1 if respondents reported a foreseeable major expense in the next 5 to 10 years and coded as 0 otherwise.

Other control variables included financial asset, home ownership, consumer debt and household income as financial variables; expectations about future income and interest rates, planning horizon and willingness to take risk as expectation variables; number of years until retirement, age, education, race/ethnicity, and household type as demographic characteristics.

Results Descriptive Results A key independent variable in the Rha et al. (2006) study was whether the household had one or more saving rules, which the authors used as a proxy for a self-control measure. In order to obtain more insight into that variable, we analyzed the distribution of each of the questions used in that measure.

In 2013, about 35 percent of households reported that they saved whatever was left at the end of the month, 2 percent reported that they spent one member’s income and saved another member’s income, 5 percent reported that they spent regular income and saved other income, and about 40 percent reported that they saved regularly each month (Table

11

1). It is not clear from these components of the Rha et al. saving rule variable what it represents, since the most common rule seems almost the same as saving.

In the 2013 SCF, 53 percent of all households reported spending less than income (Table 2). For the bivariate patterns, significance levels indicate whether the rate of saving (spending less than income) was significantly different for a particular category compared to the reference category, to make comparisons to the logistic regression easier. So, for instance, about 67 percent of households were homeowners, and 33 percent were renters. The rate of saving for homeowners was about 59 percent, which was significantly different than the rate of saving for renters, 42 percent. For each set of variables, a means test was performed using the RII technique.

Saving goals were analyzed based on whether the household listed that goal for any of their saving goals responses, so a household could have multiple goals (Lee & Hanna, 2015).

Those who had a retirement saving goal were more likely to report saving than those who did not list that goal (61 percent versus 47 percent). Those who had a precautionary saving goal were slightly more likely to report saving than those who did not list that goal (54 percent versus 52 percent). Those who had a saving goal for a purchase were less likely to report saving than those who did not list that goal (50 percent versus 54 percent). Those who had a saving goal for education were slightly more likely to report saving than those who did not list that goal (55 percent vs 53 percent). There was no difference in saving behavior between those who listed a saving goal related to children or family (other than education) and those who did not.

There was no difference in saving behavior between those who reported having a foreseeable major expense in the next 5 to 10 years and those who did not. However, there was a substantial difference in saving behavior between those who had

Table 1. Distribution of having saving rules, by specific saving rule, 2013 SCF Saving rule variable Save whatever is left at the end of the month (X3017) Save income of one family member, spend the other (X3018) Spend regular income, save other income (X3019)

Percent having rule 34.88% 2.42% 5.28%

Save regularly by putting money aside each month (X3020)

39.93%

Have at least one saving rule

45.98%

Question in the SCF: Which of the following statements come closest to describing your saving habits?

12

Journal of Personal Finance

Table 2. Bivariate analysis, distribution of independent variables, and percent of households saving in categories, 2013 SCF Variables Distribution Percent that saved p-value Entire sample 100% 53.02% Behavioral life-cycle variables Saving goals Retirement 41.02 61.02 <0.000 No response for this goal 58.98 47.45 reference Precautionary 47.69 53.91 0.003 No response for this goal 52.31 52.20 reference Children (other than education) 11.12 53.06 0.952 No response for this goal 88.88 53.01 reference Purchase 22.44 49.90 <0.000 No response for this goal 77.56 53.92 reference Education 16.64 55.45 <0.000 No response for this goal 83.36 52.53 reference Have a foreseeable major expense 55.98 53.03 0.948 Do not have a foreseeable major expense 44.02 53.00 reference Saving rules: Have one or more 45.98 71.89 <0.000 No saving rules 54.02 36.95 reference Financial characteristics Financial asset Less than $700 20.01 28.53 reference $700 to $6,799 19.99 46.12 <0.000 $6,800 to $36,919 20.01 62.02 <0.000 $36,920 to $181,099 20.00 72.47 <0.000 $181,100 or more 19.99 87.26 <0.000 Home owner 66.76 58.67 <0.000 Renter 33.24 41.66 reference Consumer debt: Have debt 63.02 52.18 <0.000 No consumer debt 36.98 54.44 reference Perceived pension adequacy: Yes 50.52 59.68 <0.000 No 49.48 46.21 reference Household income Less than $23,000 20.00 32.22 $23,000 to $39,999 21.92 46.88 <0.000 $40,000 to $61,999 18.10 56.17 <0.000 $62,000 to $102,999 20.06 72.75 <0.000 $103,000 or more 19.92 89.35 <0.000 Expectation variables Expect income increase 9.80 68.77 <0.000 Don’t expect income increase 90.20 51.31 reference Expect interest rate increase 76.77 54.02 <0.000 Don’t expect interest increase 23.23 49.70 reference

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

13

Table 2, continued. Bivariate analysis, distribution of independent variables, and percent of households saving in categories, 2013 SCF Variables

Distribution

Percent that saved

p-value

Next few months

27.95

41.53

reference

Next year

14.76

48.59

<0.000

Next few years

25.89

53.69

<0.000

Planning horizon

Next 5â&#x20AC;&#x201C;10 years

20.24

61.59

<0.000

Longer than 10 years

11.16

70.51

<0.000

Average risk

36.31

60.42

reference

No risk

46.61

42.73

<0.000

Above average risk

14.08

67.00

<0.000

3.01

57.57

<0.000

Willingness to take risk

Substantial risk

Demographic Characteristics Age Under 30

11.88

53.56

reference

30 to 39

17.10

56.16

0.027

40 to 49

18.41

51.77

0.114

50 to 59

19.92

52.07

0.175

60 to 69

16.52

56.61

0.008

70 to over

16.17

48.21

<0.000

Less than high school

17.95

42.00

reference

High school

31.30

47.11

<0.000

Some college

18.90

48.88

<0.000

Bachelor degree

19.37

65.09

<0.000

Post Bachelor degree

12.48

71.22

<0.000

White

67.96

57.28

reference

Black

14.28

42.28

<0.000

Hispanic

13.29

42.20

<0.000

4.47

54.64

0.053

Single with children

12.10

36.01

reference

Single without children

30.75

48.66

<0.000

Married with children

30.52

57.06

<0.000

Married without children

26.63

61.14

<0.000

Education

Race/ethnic status of respondent

Asian/other Household type

Weighted analysis; RII technique applied.

Reference category used in the mean test is top category of each variable.

2nd column represents proportion of household by each category.

3rd column represents proportion of saving by each category.

4th column represents significance of the mean difference from reference category.

Journal of Personal Finance

14

saving rules, with 72 percent reporting saving, and those who did not, with only 37 percent reporting saving.

The percent of households saving increased with wealth and income. The proportion of households who reported saving in the top 20 percent of financial wealth was 87 percent, more than three times the 28.5 percent rate in the lowest 20 percent of financial wealth (Table 2). The proportion in the top 20 percent of income was 89 percent, much higher than the 32 percent rate in the lowest 20 percent of income. Homeowners, those who expected pension adequacy, those who expected household income to increase faster than inflation, and those who expected interest rates to increase were more likely to be savers than the corresponding groups who did not have these characteristics. The proportion of savers increased with the planning horizon and with risk tolerance.

The highest portion of savers by age group was for the 60 to 69 age group, while those in the 70 or more age group had the lowest proportion of savers. The proportion of savers generally increased with education. For racial/ethnic groups, households with a White respondent had the highest proportion of savers, while Blacks and Hispanics had the lowest proportions. Married households with and without children had a higher proportion of savers than comparable single households.

Multivariate Results The results from logistic regression showed the effects of explanatory variables on the likelihood of saving (Table 3).

Among self-control mechanisms (saving goals, saving rules,

Table 3. Multivariate Logistic Analysis for Saving, 2013 SCF Variable

coefficient

odds ratio

Behavior life-cycle variables Have saving goals Retirement Precautionary Children (other than education) Purchase Education Have a foreseeable major expense Have saving rules Financial variables Financial assets ($100,000) Home ownership Consumer debt Perceived pension adequacy Household income (reference category = Less than $23,000) $23,000 to $39,999 $40,000 to $61,999 $62,000 to $102,999 $103,000 or more Expectation variables Expect income increase Expect interest rate increase Planning horizon (reference category = Next few months) Next year Next few years Next 5 years Longer than 10 years Willingness to take risk (reference category = No risk) Average risk Above average risk Substantial risk

0.1173* 0.0938 0.1653* 0.0214 –0.0013 –0.0439 1.1281***

1.124 1.098 1.180 1.022 0.999 0.957 3.089

0.0012* 0.2344*** –0.2645*** 0.3711***

1.001 1.264 0.768 1.449

0.4492*** 0.6314*** 1.0715*** 1.9745***

1.567 1.880 2.919 7.205

0.3921*** –0.0665

1.480 0.936

0.0573 0.1114 0.1458 0.4185***

1.059 1.118 1.157 1.520

0.0194 0.0682 0.2223

1.020 1.071 1.249

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

15

and having a foreseeable major expense), saving rules had the largest effect on the likelihood of saving. Among the saving goals, saving for retirement and saving for children/family had positive effects (significant at the 10 percent level) on saving, while the Rha et al. (2006) results for the 1998 SCF showed no significant effect of having a saving goal for children/family on whether the household saved. Foreseeable major expenses were not significantly related to saving, while the Rha et al. showed a significant positive effect on saving. Appendix 2 shows the results of a Likelihood Ratio test for whether the model with BLC variables added was better than the model without BLC variables. As the test indicated, the model with BLC variables included was significantly better than the one without BLC variables.

Among financial variables, income had significantly a positive effect on saving, which is consistent with standard life cycle theory for consumption smoothing. Financial asset and perceived pension adequacy variables were also significantly and positively related to the likelihood of saving. Households who had positive debt were less likely to save than households who did not have debt. Households who owned a home were more likely to save than renter households.

Households who expected income to increase faster than prices in the future were more likely to save than those who did not, while expectation of interest rates increasing had no a significant effect on saving. The planning horizon, considered a proxy for time preference, had no significant effect on saving, except for the category of longer than 10 years compared to

Table 3, continued. Multivariate Logistic Analysis for Saving, 2013 SCF Variable

coefficient

odds ratio

Demographic characteristics Number of years until retirement

0.0136***

1.014

Age (reference category = under 35) 35 to 44

–0.2349**

0.791

45 to 54

–0.4066***

0.666

55 to 64

–0.1780

0.837

65 to over

–0.2352*

0.791

High school

0.0443

1.045

Some college

0.0375

1.038

Bachelor degree

0.1600

1.174

Post Bachelor degree

0.0870

1.091

Education (reference category = Less than high school)

Race/ethnicity (reference category = White) Black

–0.2537***

0.776

Hispanic

–0.2958***

0.744

Asian/other

–0.1580

0.854

Single without children

0.4231***

1.527

Married with children

0.2750**

1.317

Married without children

0.4537***

1.574

Household type (reference category = Single with children)

Intercept Max-rescaled R-Square –2 Log L Unweighted analysis, significance levels based on RII technique p–value is based on two tail test. *** significant at 1%; ** significant at 5%; * significant at 10%

–1.5794*** 0.3378 31983.464

Journal of Personal Finance

16

the next few months. Willingness to take risk also did not significantly affect saving behavior in our model.

The number of years until retirement was positively related to the likelihood of saving. Age had a negative effect on the likelihood of saving. Education did not have significant effects. Black and Hispanic households were less likely to save than White households. Married households with and without children and single households without children were more likely to save than single households with children.

Discussion and Implication In order to focus only on the effects of the BLC variables in the 2013 SCF and to compare our results to those of Rha et al.

(2006), we created Table 4. The most important BLC variable both in our analysis of the 2013 SCF and in the Rha et al.

(2006) analysis of the 1998 SCF was having saving rules. In particular, having one or more saving rules had a large positive significant effect on the likelihood of saving. In our analysis, a household with saving rules had odds of saving three times as high as a household without saving rules (Table 3).

Having a saving goal related to retirement had a significantly positive effect on saving in 1998, but the effect in 2013 was not significant at the conventional 5 percent level based on a two-tail test. Having a saving goal related to children/family (other than for education) had a weak positive effect in 2013 but not in 1998. Having a saving goal for education had no

significant effect in 2013, whereas it had a significant negative effect in 1998. The lack of any effect of having a saving goal for education in 2013 might reflect complex factors, for instance, those stating that saving for education was a goal might be in a situation where saving is difficult, for instance, having expenses related to having children at home. Precautionary and purchase saving goals and having foreseeable major expenses might reasonably be expected to positively affect the likelihood of saving, but neither had significant effects in 2013, while both had significant positive effects in 1998.

Articulation of specific saving goals did not have much effect on the likelihood of saving in 2013, with only two types of goals—retirement and children/family—having weak effects that were significantly different from zero. Having one or more simplistic saving rules was the only behavioral factor that had a strong effect on saving.

For both survey years, having saving rules had a strong positive effect on saving. To some extent, it would be plausible to repeat the conclusions of Rha et al. (2006), that clients might benefit from being urged to implement saving rules. However, our analysis of the distribution of different types of saving rules (Table 1) suggests that getting clients to implement saving rules may be almost as challenging as implementing saving. Rha et al. suggested that helping clients identify saving goals might help in getting clients to save, but our results suggest that this may have only a weak

Table 4. Comparison of the effects of self-control mechanisms on saving behavior in 2013 & 1998 SCF 2013 Variables

Coefficient

1998 Odds Ratio

Coefficient

Odds Ratio

Behavior Life-cycle Variables Have saving goals Retirement

0.1173*

1.124

0.2308***

1.260

Precautionary

0.0938

1.098

0.1281***

1.137

Children/Family (other than education)

0.1653*

1.180

0.0472

1.048

Purchase

0.0214

1.022

0.1799***

1.197

Education

–0.0013

0.999

–0.3044***

0.738

–0.0439

0.957

0.0855***

1.089

3.089

0.9585***

2.608

Having a foreseeable major expense Have saving rules

1.1281***

a. Result from Rha, Montalto and Hanna (2006) *** significant at 1%; ** significant at 5%; * significant at 10%

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

effect, and only for retirement and for children/family goals.

There may be many “tricks” to help get clients to start saving (e.g., Ambrose, 2017), but we did not find strong evidence that there are any simple ways to get clients to exercise

17

self-control that would lead to more saving. It is possible that only externally provided “nudges” such as default enrollment in employer retirement plans (Thaler & Benartzi, 2004) can effectively increase the likelihood that households save.

Appendix 1. Measurement of Saving Goal in the SCF If household reasons for saving is 17 or 22, then code 1 as retirement.

If household reasons for saving is 23, 24, 25, 32, 92, 93, then code 1 as precautionary.

If household reasons for saving is 3, 5, 6, then code 1 as children/family (other than education).

If household reasons for saving is 12, 13, 14, 15, 16, 27, 29, 30, 9, 18, 20, 41, then code 1 as purchases.

If household’ reasons for saving is 1, 2, then code 1 as education.

1. Children’s education; education of grandchildren

23. Reserves in case of unemployment

2. Own education; spouse/partner’s education; education—not known for whom

24. In case of illness; medical/dental expenses

3. “For the children/family”, n.f.s.; “to help the kids out”; estate 5. Wedding, Bar Mitzvah, and other ceremonies (except 17)

25. Emergencies; “rainy days”; other unexpected needs; for “security” and independence 26. Investments reasons (to get interest, to be diversified, to buy other forms of assets)

6. To have children/a family

27. To meet contractual commitments (debt repayment, insurance, taxes, etc.), to pay off house

9. To move (except 11)

28. “To get ahead”; to advance standard of living

11. Buying own house (code “summer cottage” in 12)

29. Ordinary living expenses/bills

12. Purchase of cottage or second home for own use

30. Pay taxes

13. Buy a car, boat or other vehicle

31. No particular reason (except 90, 91, 92)

14. Home improvements/repairs

32. “For the future”

15. To travel; take vacations; take other time off

33. Like to save

16. Buy durable household goods, appliances, home furnishings; hobby and recreational items; for other purchases not codable above or not further specified; “buy things when we need/want them”; special occasions

40. Don’t wish to spend more

17. Burial/funeral expenses

91. Wise/prudent thing to do; good discipline to save; habit

41. To give gifts; “Christmas” 90. Had extra income; saved because had the money left over—no other purpose specified

18. Charitable or religious contributions 92. Liquidity; to have cash available/on hand 20. “To enjoy life” 93. “Wealth preservation”; maintain lifestyle 21. Buying (investing in) own business/farm; equipment for business/farm

-1. Don’t/can’t save; “have no money”

22. Retirement/old age

-7. Other 0. Inap. (/no further responses)

18

Journal of Personal Finance

Appendix 2. Comparison of Unrestricted and Restricted Model Using a Likelihood Ratio Test The null hypothesis is that exclusion variables’ coefficients from the unrestricted model is set equal to zero.

The likelihood-ratio statistic is ΔG2 = −2 log L from restricted model − (−2 log L from unrestricted model) The degrees of freedom is k (the number of coefficients in question).

The p-value is P(χ2k ≥ ΔG2).

2.1 In our study, ΔG2 = 33604.794 – 31983.464 = 1621.33 with df = 7.

This matches Likelihood Ratio

1621.33, 7, <.0001

Where, restricted model excludes BLC variables, unrestricted model includes BLC variables.

Thus, the model with BLC variables is preferred to the model without BLC variables.

Handwerk, B. (2015). Scientists replicated 100 psychology studies, and fewer than half got the same results. Smithsonian, Ambrose, E. (2017). Saving tricks that work. AARP Bulletin, 58 (5), August 27. Retrieved June 9, 2017 from Smithsonian.com.

20.

Hanna, S. D., Kim, K.T., & Chen, S. C.-C. (2016). Retirement Bricker, J., Dettling, L. J., Henriques, A., Hsu, J. W., Moore, K. B., savings. in J. Xiao, Handbook of Consumer Finance Research, Sabelhaus, J., Thompson, J., & Windle, R. A. (2014). Changes Springer Publishing, 2nd edition, pp. 33-43. in U.S. Family finances from 2010 to 2013: Evidence from the Heckman, S.J. & Hanna, S.D. (2015). Individual and institutional Survey of Consumer Finances. Federal Reserve Bulletin, 100(4), factors related to low-income household saving behavior.

1-41.

Journal of Financial Counseling and Planning, 26(2).

Browning, M., & Crossley, T. F. (2001). The life-cycle model of consumption and saving. The Journal of Economic Perspectives, Kennickell, A. B., Starr-McCluer, M., & Surette, B. J. (2000). Recent changes in U.S. family finances: Results from the 1998 Survey of 15(3), 3-22 Consumer Finances. Federal Reserve Bulletin, 86 (1). 1-29.

Fisher, P. J., & Montalto, C. P. (2010). Effect of saving motives and horizon on saving behaviors. Journal of Economic Psychology, 31, Kim, K. T. & Hanna, S. D. (2015). Does financial sophistication matter in retirement preparedness? Journal of Personal Finance, 92-105.

14(2), 9-20.

Griesdorn, T. S., Lown, J. M., Devaney, S. A., Cho, S. H., & Evans, D. A. (2014). Association between behavioral life-cycle constructs Kim, K. T., Lee, J., & Hong, O. E. (2016). The role of self-control and financial risk tolerance of low-to moderate-income house- on retirement preparedness of U.S. households. International holds. Journal of Financial Counseling and Planning, 25(1), 27-40. Journal of Human Ecology, 17(2), 31-42.

References

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

19

Laibson, D., & Harris, C. (2001). Dynamic choices of hyperbolic consumers. Econometrica, 69(4), 935–957.

Shefrin, H. M., & Thaler, R.H. (1988). The behavioral life-cycle hypothesis. Economic Inquiry, 26(4), 609-643.

Laibson, D., Repetto, A., & Tobacman, J. (1998). Self-control and saving for retirement. Brookings Papers on Economic Activity, 1, 91–196.

Shin, S. H. & Hanna, S.D. (2016). Accounting for complex sample designs in analyses of the Survey of Consumer Finances. Journal of Consumer Affairs, published online, DOI: 10.1111/joca.12106.

Lee, J. M., & Hanna, S. D. (2015). Saving goals and saving behavior from a perspective of Maslow’s Hierarchy of needs.

Journal of Financial Counseling and Planning, 26(2).

Lindamood, S., Hanna, S. D., & Bi, L. (2007). Using the Survey of Consumer Finances: Methodological consideration and issues.

Journal of Consumer Affairs, 41 (Winter), 195–214.

Modigliani, F., & Brumberg, R. (1954). Utility analysis and the consumption function: An interpretation of cross-section data.

In K. Kurihara (Ed.). Post-Keynesian economics. New Brunswick: Rutgers University Press.

Strömbäck, C., Lind, T., Skagerlund, K., Västfjäll, D., & Tinghög, G. (2017). Does self-control predict financial behavior and financial well-being? Journal of Behavioral and Experimental Finance, in press.

U.S. Federal Reserve Board (2014). Excel file based on public data. https://www.federalreserve.gov/econres/files/scf2013_ tables_public_real.xls. Retrieved June 8, 2017.

Thaler, R. H. (1994). Psychology and savings policies. American Economic Review, 84(2), 186-192.

Montalto, C. P., & Sung, J. (1996). Multiple imputation in the 1992 Survey of Consumer Finances. Journal of Financial Counseling and Planning, 7, 133-146.

Thaler, R. H., & Benartzi, S. (2004). Save more tomorrow: Using behavioral economics to increase employee saving. Journal of Political Economy, 112(1), 164-187.

Open Science Collaboration. (2015). Estimating the reproducibility of psychological science. Science, 349(6251), aac4716.

Thaler, R. H., & Shefrin, H. M. (1981). An economic theory of self-control. Journal of Political Economy, 89(2), 392–406.

Rha, J., Montalto C. P., & Hanna S. D. (2006). The effect of self-control mechanisms on household saving behavior. Financial Counseling and Planning, 17(2), 3–16.

Yuh, Y. & Hanna, S. D. (2010). Which households think they save? Journal of Consumer Affairs, 44 (1), 70-97.

20

Journal of Personal Finance

What’s Your Risk Appetite? Helping Financial Advisors Better Serve Clients (by Quantifying Kahneman-Tversky’s Value Function)

Sid Muralidhar, Grade 11, Thomas Jefferson High School for Science and Technology, Great Falls, Virginia Emerson Berlik, Grade 11, Thomas Jefferson High School for Science and Technology, Round Hill, Virginia

Abstract This paper presents a methodology to allow advisors to quantify risk tolerance of clients, over gains and losses, based on the Kahneman-Tversky survey. Once a formal and quantitative estimate of an individual’s risk appetite can be determined, and its evolution tracked over time, advisors can design effective investment portfolios to cater to the client’s specific risk tolerance. The paper extends this individual-level risk diagnostic and applies it over various subgroups and demonstrates that (a) teens are more risk-seeking than adults when it pertains to losses; (b) among investment professionals, women are more conservative than men when it pertains to gains; and (c) even within these subgroups, every individual is unique and neither expected utility theory nor prospect theory appropriately capture the diversity in risk tolerance. This paper seeks to make Kahneman-Tversky’s research on prospect theory/behavioral economics, and their value function practical and user-friendly, thus improving investment decision making.

Key Words risk appetite; risk tolerance, risk aversion, risk seeking, loss aversion, Kahneman-Tversky, behavioral finance, value function, utility function

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Introduction Individuals increasingly face financial decisions earlier in life and on a bigger scale than ever before (student debt1, mortgage loans, retirement2, etc.). The financial decisions they make are influenced by their financial literacy and their behavioral biases, and sadly the data on financial literacy is not promising as Lusardi and Mitchell (2014) demonstrate that many adults cannot answer basic questions about compounding, inflation, or diversification3. With respect to behavioral biases, the seminal Kahneman-Tversky (1979) paper demonstrated that investors were not risk averse as postulated by expected utility theory and identified biases like loss aversion (i.e., a tendency to gamble losses). But they did not provide a formal framework for individuals to determine their respective behavioral biases and this poses a challenge for advisors and individual investors. For example, many investment strategies are designed to be age-based (i.e., rotating a portfolio into a greater allocation to bonds as one ages), but as Merton (2014) notes, every investor is unique and it is possible that there could be biases based on initial wealth, gender and ambition that could make these generic strategies suboptimal for a large segment of the population. Further, robo-advisors ask very naïve questions to gauge risk appetite. For example, a robo-advisor’s only question about risk tolerance to establish a portfolio is whether the investor cares more about maximizing gains, minimizing losses or both equally. The purpose of this paper is to address this challenge and help advisors establish an individual’s risk tolerance over gains and losses using Kahneman-Tversky seminal survey. Once the risk tolerance of an individual can be formalized and robustly quantified, it then allows the risk tolerance to be tracked over time to allow for changes in tolerance to be addressed by changes in portfolios. Furthermore, this method of formalizing risk tolerance allows us to examine the importance of age, gender and financial literacy in risk tolerance by examining aggregate subgroup results.

Kahneman-Tversky’s multidisciplinary work, blending economics and psychology, forms the basis of this paper’s proposed risk assessment toolkit. It also draws inspiration

1.

2.

3.

Muralidhar and Pamecha (2016).

Merton (2014).

For example, only one-third of the population polled was able to answer all three basic questions posed.

21

from Andrew Lo’s adaptive markets hypothesis (2017), which argues that the psychological tendencies of individuals, and their evolution, need to be considered in portfolio management. Further, Choi, Fisman, Gale, and Kariv (2007) attempt to extract risk preferences of individuals using budget constraints through an innovative graphical interface. They conclude that, “the behavior of subjects is generally complex and we found it impossible to classify in a simple taxonomy.” This quote also motivates this research, in an attempt to provide individual-level risk assessments.

This paper attempts to extend Kahneman-Tversky’s work by introducing the Risktyle model, formally quantifying the Kahneman-Tversky value function4. Risktyle is flexible and can be applied to individual or group responses and we provide four case studies of individuals to demonstrate how investors can differ dramatically from one another and even relative to age and experience-based assumptions. Interestingly, when applied to aggregate teen and investment professionals (pros), Risktyle highlights unique behavioral biases, especially differences in risk appetite between pro men and pro women (especially on gains), and teens and pros (especially on losses).

On an individual basis, using the Risktyle model to track and educate users on their evolving behavioral biases, as they age and improve their financial literacy, can potentially ensure precious savings are not gambled away. For example, should the equity market decline as it did in 2008, investors who are risk averse on losses might want to cut their losses; others who are loss averse might be willing to hold the portfolio or even increase their allocation to risky assets. Similarly, insurance purchases might be influenced by whether investors are riskseeking or risk averse with respect to losses.

The paper is structured as follows: Section I reviews expected utility theory (EUT) and prospect theory (PT) and discusses their key assumptions of the risk appetite of the population.

Section II describes the Risktyle model and demonstrates its efficacy in explaining the Kahneman-Tversky results, thereby potentially validating its use in quantifying the value function and estimating risk appetite on gains and losses. The four case studies in Section III demonstrate how the Risktyle model, when applied to different individuals, highlights that each investor is unique and potentially contradicts the traditional assumed risk appetite. Section IV summarizes the data from sub groups to demonstrate how gender, age and literacy affect 4.

To find out your risk appetite using the Kahneman-Tversky survey, visit Risktyle.com.

22

Journal of Personal Finance

risk appetite. Section V examines areas of future research to extend this approach and Section VI concludes.

assumption forms the basis of all portfolio recommendations derived from models using modern portfolio theory (MPT).

Section I—Risk Perspectives of EUT and PT and the Value Function

Prospect Theory: The Alternative to EUT

Daniel Kahneman won the Nobel Prize in 2002 for his work on prospect theory, “for having integrated insights from psychological research in economic science, especially concerning human judgment and decision making under uncertainty” (The Nobel Foundation n.d.). The main purpose of Kahneman and Tversky (1979) was to critique expected utility theory (EUT), the prevailing model for economic behavior. They sought to show that EUT is based on a representative, rational, and consistent individual (an “econ”), and failed to capture actual human behavior. If EUT’s assumptions were not acceptable, it follows that its recommendations must be overturned as well. So, Kahneman-Tversky proposed an alternative model of choice for decisions influenced by risk, replacing the traditional utility function with a value function. This section briefly reviews EUT and PT and sets the stage for the new approach.

Expected Utility Theory In 1738, Daniel Bernoulli (1954) proposed “utility” as the total satisfaction of consuming goods and services. He introduced an expected utility hypothesis: the value of an item or outcome is not based on its price, but rather the utility it yields, and the theory of diminishing marginal utility—the level of utility gained is less at higher levels of overall utility. Bernoulli claimed rational behavior can be described as maximizing utility, rather than monetary outcomes, which accounts for risk aversion (1954). This was a bold claim at the time.

Von Neumann and Morgenstern (1944) formalized EUT, requiring individual preferences to satisfy five axioms (continuity, transitivity, completeness, monotonicity, and independence). If satisfied, individuals prefer actions that maximize utility, thus being “rational.” EUT defines a “gamble” as being a probability distribution over a known, finite set of outcomes (e.g., 90 percent of $4,000 or 10 percent of $0). Expected value was defined as probability multiplied by outcome. In effect, the principal assumption with respect to risk tolerance was that investors were risk averse (i.e., when given two gambles with the same expected return but with different risks, the investor prefers the less risky outcome). This simple

PT is based on responses to contrived gambles presented to test subjects. University students and faculty in Israel were asked to choose between two gambles in a question5. Each question was followed or preceded by a question Kahneman-Tversky considered the “same” choice. Figure 1a (Kahneman-Tversky’s #3) and Figure 1b (Kahneman-Tversky’s #4) shows two examples of gambles (or Options), and taken together, an example of “same” prospects (or questions).

Option A

Option B

80% chance of winning $4,000

100% chance of winning $3,000

Option A

Option B

20% chance of winning $4,000

25% chance of winning $3,000

Figure 1a and Figure 1b. Two related questions in the KahnemanTversky survey (Kahneman-Tversky’s #3, 4). Each is a prospect made up of two gambles (Option A and Option B). Figure 1b divides the probabilities in Figure 1a by 4.

In Figure 1a, the individual is offered a higher sum of money with a lower probability of winning (Option A), as opposed to a guaranteed chance of winning a lower sum of money (Option B). In Figure 1b, the probability of winning in both Options in Figure 1a are simply divided by 4. Kahneman-Tversky claimed the options were the “same” and that EUT “econs” should recognize them as such and pick the one with the highest expected value (Option A) in both questions.

Option A

Option B

80% chance of winning $4,000

100% chance of winning $3,000

Option A

Option B

80% chance of losing $4,000

100% chance of losing $3,000

Figure 2a and Figure 2b. Two related questions in the KahnemanTversky survey (Kahneman-Tversky’s #3, 3’). Figure 2b multiplies the outcomes of Figure 2a by –1. Figure 2b is the prime of Figure 2a, which is Figure 1a repeated.

5.

No details are provided on the age, sex or financial sophistication of the respondents, which we address later. Not all respondents were asked the same questions; number of responses ranged from a low of 66 for some questions to a high of 95 for others. Further, the experiment was replicated at other educational institutions validating the original results.

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Kahneman-Tversky also examined decisions on uncertain losses by flipping the prospect of “winning” to “losing.” The purpose of this switch is to analyze the effect of decision making with losses rather than gains. Again, EUT’s “econs” should recognize this simple transformation and be consistent, risk averse and rational (i.e., pick B in Figure 2b). Similarly, Kahneman-Tversky examined gambles with very low (Kahneman-Tversky’s #8, 14, 8’, 14’) and high probabilities (Kahneman-Tversky’s #1, 3, 7, 14, 3’, 7’, 14’), and very small and large gains (Kahneman-Tversky’s #1, 2, 3, 4, 7, 8, 13, 14) and losses (Kahneman-Tversky’s #3’, 4’, 7’, 8’, 13’, 14’). The relevant questions for this paper are provided in Appendix I.

Kahneman-Tversky’s Results and the Value Function Kahneman-Tversky’s main findings were: (1) people tend to overweight certainty—the certainty effect; (2) people are affected by the prospect of losses differently than the prospect of gains—loss aversion; and (3) people make decisions based on the potential of gains and losses rather than final expected value. These results violated the axioms of EUT; hence, EUT was found inadequate for modeling human behavior (as would be the portfolio models of MPT) and needed to be replaced with PT. In the certainty effect, people overweight outcomes that are considered “certain,” relative to outcomes that are merely probable, contradicting EUT. For example, consider a 45 percent chance of winning $6,000 vs. a 90 percent chance of winning $3,000 (Kahneman-Tversky’s #7), and a 0.1 percent chance of winning $6,000 vs. a 0.2 percent chance of winning $3,000 (Kahneman-Tversky’s #8). EUT predicts indifference among gambles in Kahneman-Tversky’s #7 and Kahneman-Tversky’s #8, as they have the same expected value. However, a statistically significant 86 percent of respondents chose the more certain outcome in Kahneman-Tversky’s #7 (Option B) and 73 percent of respondents chose the outcome which offered the larger gain in Kahneman-Tversky’s #8 (Option A).

Kahneman-Tversky also observed that people behave differently when faced with gains and losses. In Figure 2a, 80 percent of respondents erred on the side of caution and picked Option B. However, when the prospect was flipped from “winning” to “losing” (Kahneman-Tversky’s #3’), the results also flipped. Now, in Figure 2b, 92 percent of respondents preferred to risk a greater loss (Option A) to the guaranteed

23

loss. The term loss aversion was coined to show that people are risk averse in relation to gains, whereas they are risk-seeking in relation to losses (Kahneman and Tversky, 1979). Over the years, more research has been conducted on this “loss aversion” behavior that was demonstrated among Kahneman and Tversky’s respondents. Benartzi and Thaler (1997) showed people are not averse to the variability of a potential loss, but rather the actual loss itself. They showed people inherently are averse to losses. This is further supported by Kahneman, Knetsch, and Thaler (1991), which found that people tend to value things greater if they own them, called the endowment effect, showing loss aversion can be a product of underweighting opportunity costs.

Kahneman-Tversky claimed there is a “reference point” relative to which people base and value on gains and losses, rather than the final outcome. “[O]ur perceptual apparatus is attuned to the evaluation of changes or differences rather than to the evaluation of absolute magnitudes” (Kahneman and Tversky, 1979). Interestingly, the reference point is never clearly identified in their paper. Kahneman-Tversky recommended replacing the utility function with a value function, defined on deviations from the reference point and generally concave for gains and commonly convex for losses, and steeper for losses than for gains (Figure 3).

However, Kahneman and Tversky did not ask their respondents to quantify by how much more they preferred one option to the alternative presented. Thus, the value function they proposed is hypothetical and was not formally quantified, thereby limiting its broad adoption. In their representation, there is an implicit assumption that all respondents have the same value function/risk appetite. Choi, Fisman, Gale, and Kariv (2007) show instead, using a very different technique from the one proposed in this paper, that there is heterogeneity of individual behavior and risk appetite 6.

Moreover, since the responses to any question were always less than 100 percent, it further suggests that there are exceptions to both the assumption that all investors are risk averse (as in EUT) or that all investors are loss averse (as in PT). Hence,

6.

In their experiment they are able to show different individual behaviors among their subjects. For example: one always chooses safe portfolios (infinite risk aversion), one displays the logarithmic von Neumann-Morgenstern utility function, one displays risk neutrality, and some display mixed behavior—clearly not consistent with the standard assumption of EUT.

24

Journal of Personal Finance

it is important to develop a methodology to capture each individual’s value function.

Figure 3. Kahneman and Tversky’s Value Function (Kahneman & Tversky, 1979).

Since Kahneman-Tversky’s groundbreaking work, further research has been conducted on prospect theory and its implications.

Shefrin and Statman (1985) found people sell winning stocks too early and hold on to stocks that have lost value too long.

These circumstances can be avoided knowing one’s behavioral biases.

Camerer (1995) explains there are two types of tests in economic theory: destructive (testing if the old theory is robust and replicable) and constructive (proposing an alternate theory) ones. We hoped to have achieved both by showing the behaviors of individuals cannot be generalized for an entire population and that everyone has his or her own, widely-different behavior as it pertains to decisions under uncertainty.

On Dan Ariely’s website, he says of behavioral economics, “…we repeatedly and predictably make the wrong decisions in many aspects of our lives and that research could help change some of these patterns” and the Risktyle model may help change some of these patterns!

Section II—Risktyle: Assigning “Values” to Kahneman-Tversky’s Value Function The primary purpose of this paper is to present Risktyle, a formal model to estimate Kahneman-Tversky’s value function, based on the responses of individuals to Kahneman-Tversky’s original survey. By establishing a formal value function, this

paper then allows advisors to be able to quantify their client’s respective risk tolerance, see the magnitude of the risk-seeking or risk-averse behavior, determine if there are differences in risk appetite over gains and losses, and finally, be able to track the evolution over time. This clarity on the risk appetite of their clients provided by this approach is potentially an improvement to current methods of bucketing clients into broad portfolios based on age, wealth or responses to single/very broad questions about risk tolerance (as noted earlier as being used by robo-advisors).

Additionally, because Risktyle is formal and flexible, it can be applied to individuals, subgroups, and entire populations, and one can draw valuable conclusions about behavioral biases and risk-appetite of each. Making individuals or even groups of individuals aware of their risk appetite has interesting implications for not only portfolio selection, but also for other decisions with social impact.

For example, some have argued that girls tend to be more risk averse than boys and this can affect career choices and even test scores such as the SAT (Baldiga 2014). By making girls aware of their risk biases, research has shown that they could be educated into not always choosing the safe option which could improve test scores and their future career prospects (Kachnowski 2017). Further, many colleges admit students for entrepreneurship courses and do not have a clear way of distinguishing which students have the appropriate risk appetite to succeed as risk-takers (as SAT scores do not provide such clarity). Hence, this formalization of the value function can have positive externalities outside finance.

Replicating Kahneman-Tversky’s Original Survey (with More Clarity on Respondents) We replicated the Kahneman-Tversky experiment, surveying 378 respondents (compared to a range of 66–95 respondents in Kahneman-Tversky for specific questions).

In addition, prior to conducting the survey, each respondent was asked basic questions about age, gender, financial literacy and current occupation. This additional level of detail on each respondent allows Risktyle to be applied not only on individual responses, but also allows us to examine whether Kahneman-Tversky might have missed some unique insights by focusing strictly on aggregate responses, without examining potential dispersion of results based on

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

25

subgroups—teens versus adults, men versus women, financially literate versus those not familiar with finance concepts.

Risktyle Model—A Form of Risk-Adjusted Performance

Our survey addresses some inadvertent biases in KahnemanTversky’s research. First, Kahneman-Tversky’s respondents were limited to university faculty and students, and did not include investment professionals or individuals from other walks of life. This has interesting implications for the advisor community. Second, Kahneman-Tversky’s survey does not provide additional information (i.e., age, gender) on the respondents, which could further mask interesting differences in subpopulations. Third, Kahneman-Tversky reports aggregate results, thus making generalizations about the population and ignoring individual or subgroup results.

As Choi, Fisman, Gale, and Kariv (2007) point out, the pairwise choices presented in Kahneman-Tversky’s (1979) survey were created for the specific purpose of violating the axioms it sought to expose, thereby limiting the use of the data collected. However, Kahneman-Tversky’s questions provide us the probabilities and outcomes of each of these options, presented to the respondents, and this paper seeks to show that is adequate to extract risk appetite. Thus, research on Kahneman-Tversky’s results must go a step further to extract important information from the decisions made by the subjects.

A. The Experiment: To be consistent with Kahneman-Tversky’s results, Risktyle uses 14 of Kahneman-Tversky’s original 20 questions (Kahneman-Tversky #1, 2, 3, 4, 7, 8, 13, 14, 3’, 4’, 7’, 8’, 13’, 14’ —See Appendix I), avoiding the word and multistep problems for simplicity. All respondents were given the same questions so this allows for comparisons of either individuals or subgroups. Our survey provided the same instructions to respondents that Kahneman-Tversky provided (e.g., no correct answer, treat as a real example) and used the same research method of randomizing questions to prevent bias. In all, 297 American teens (aged 13–177 consisting of 89 females, 208 males) and 81 global investment professionals8 (41 women, 40 men) were polled to get a relatively diverse population, compared to the university respondents Kahneman-Tversky surveyed.9 Teens are categorized by age, gender, and economic knowledge (proficient, fair, and little to none). Professionals are categorized by age, gender, and field of work (e.g., financial consultant, asset manager, investment banker). Finally, this paper reports aggregate population results consistent with Kahneman-Tversky, but it goes further to examine the results of subgroups (e.g., pro men vs. pro women) and individuals. The individual results demonstrate the potentially wide dispersion of risk appetite, masked by the Kahneman-Tversky aggregate, as do the subgroup results.

One of the attractive features of MPT is that two investments can be compared to one-another by examining their risk-adjusted return either on an absolute basis (Sharpe 1994 or Modigliani 1997), or by examining them on a relative basis (Sharpe 1994 or Muralidhar 1999). In short, the Sharpe ratio is calculated by dividing the expected return of the investment (either absolute or relative) by the appropriate volatility of the investment, and the investment with the higher ratio is typically preferred to the investment with the lower ratio if one is risk-averse. The calculation of the M-square measure of risk-adjusted performance is a bit more complex and requires normalizing for differences in volatility and makes the comparison in terms of risk-adjusted returns. This analysis is very easy to conduct for a professional, but the average individual typically does not know the expected return of a stock or bond investment. Moreover, the knowledge of Sharpe ratios or M-square is beyond the understanding of a financially illiterate individual.

Hence, it is hard to use MPT-based examples to establish the risk-tolerance of individuals.

7.

8.

9.

The return rate of the teens group was 74 percent, with 297 respondents.

The return rate for investment professionals was 84 percent, with 81 respondents.

Subsequent to the writing of this paper, we have managed to solicit responses from an additional 100 non-investment professional adults, but do not report the results here. See Section V.

On the other hand, the simple gambles posed by KahnemanTversky in Figure 1 and 2 lend themselves to a much broader audience and can be easily answered by teens, adults and investment professionals alike. However, the challenge is now on the researcher to convert the responses provided to these gambles into a reasonable measure of risk-tolerance. We attempt to do so by developing a formal equation to plot the value function that Kahneman-Tversky hypothesized was representative of individual behavior. The Risktyle model attempts to do so by converting each option in a gamble into a form of a risk-adjusted performance measure so that once it is quantified, it can be compared to other risky gambles (both in terms

26

Journal of Personal Finance

of sign and magnitude), and potentially aggregated across subgroups.

One goal of Risktyle is to explain why individuals appear to answer questions the way they did in Kahneman-Tversky’s original survey (i.e., different than EUT). The Risktyle model examines individual gambles in each question/prospect, relative to its alternative, and tries to formalize the likely psychological process conducted by the individual in selecting that gamble.

This is a very tall order and Risktyle is, at best, a dynamic first step, charting new territory in quantifying risk aversion and risk-seeking investment behavior using Kahneman-Tversky’s hypothetical gambles.

The Risktyle model derives/assigns a numerical “value” for each gamble (i.e. Option A and Option B) in each prospect. The attractiveness of Option A is influenced by Option B and vice versa, each effectively serving as a reference point to the other.

Once the “value” of every gamble in each prospect is established, one can plot the value function of each individual, based on the gamble they choose for every prospect. Simply put, the model is calibrated strictly to Kahneman-Tversky’s questions and it is possible it may not work for a different set of questions.

This is discussed in Section V. In every Kahneman-Tversky gamble, there is a unique probability (p e.g., 80 percent) and a unique outcome (x e.g., $4,000), and effectively a complementary probability (1–p e.g., 20 percent) and an alternative outcome (usually zero for both gambles, with a few exceptions). Comparing the (xA , pA) of Option A to the (xB , pB) of Option B is potentially an applesto-oranges comparison, because xA ≠ xB and pA ≠ pB in all of Kahneman-Tversky’s prospects, even if the expected value is the same. Instead, Risktyle starts by comparing gambles in a prospect by defining relative risk based on complementary probabilities (1–p) for the exact same alternative outcome (i.e., the zeros). The (1–p) is easier to compare than p, because they are both anchored to the same outcome—zero.

Prospects in which the probabilities and outcomes are the same do not need to be considered for two reasons: (i) Kahneman-Tversky’s questionnaire did not contain any such problems, and (ii) when probabilities are exactly the same (i.e., 50 percent vs. 50 percent) or when outcomes are the same (i.e., $4,000 vs. $4,000), it is assumed individuals will choose the higher sum of money or probability, or in the case of equal sums they will be indifferent.

In EUT, to be considered rational, an individual should pick gambles with the highest expected value. Kahneman-Tversky shows that humans do not make choices based on expected value, and Risktyle attempts to explain the behavior. Recall, in Risktyle, the reference point for Gamble A is Gamble B and viceversa. Because Kahneman-Tversky never chose identical p’s and x’s in every question, we can do a few simple things in our model which is a five step process.

Define the base variables as follows: Gain/Loss Probability Expected Value (EV)

Gamble A xA pA EVA = xA x pA

Gamble B xB pB EVB = xB x pB

Step 1: Calculate gamble risk (relative to its reference point). Gamble risk has two terms: (i) the probability (1–p) of obtaining the outcome common to both gambles (zero), (ii) added to the difference between the probability of the other gamble and its reference (or pA – pB). The higher this value, the riskier the gamble is relative to the reference gamble. The second term (or pA – pB) serves two purposes. Not only does it ensure the overall gamble risk is not 0, if p is 100 percent, but it also turns gamble risk into a relative calculation. This proves useful later in the model, to derive the final value of a bet (or “individual gamble strength”). Negative gamble risk is possible as one gamble can be much more certain than another.

Once the gamble risk is calculated, each value is adjusted, by taking the minimum of 1 and the gamble risk of each gamble calculated before. This ensures relative risk cannot exceed 1, causing gamble risk to be bound within [–1, 1]. This is useful in problems such as Kahneman-Tversky #14, in which the probability variance between gambles is enormous (0.1 percent vs.

100 percent).

Step 2: Calculate relative gamble risk by dividing each respective gamble by the absolute value of its reference point, within each prospect, creating a truly relative value from one gamble to another. The absolute value of the divisor negates the bias caused by a negative adjusted gamble risk. Ignoring

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

27

the absolute value could incorrectly flip the order of risk between the two gambles. By design of the model, the relative gamble risk of A is always greater than the relative gamble risk of B. This is because Gamble A has a lower probability of a nonzero outcome, thus a greater probability of uncertainty, leading to Gamble A having greater relative gamble risk.

Step 3: Calculate relative expected value, defined as the expected value of one gamble divided by the expected value of the reference point within a prospect. Expected value is made relative to manipulate it against relative risk.

Step 4: Calculate individual gamble strength (IGS), which is similar in principle to risk-adjusted performance calculations in traditional finance. IGS measures the relative risk-adjusted

value of each gamble, by subtracting the relative gamble risk from the relative expected value, and multiplying that value by the relative size of the bet. The more negative (positive) the value of the IGS, the greater (lower) the risk involved for the possible reward in relation to the alternative gamble. Note that IGS is only comparable within a prospect, not across prospects10.

The IGS score of each gamble is the critical number needed to evaluate risk tolerance and provide valuable insights on behavioral biases of individuals, subgroups or even entire populations. The IGS scores and expected value of each gamble in all 14 selected prospects are listed in Table 1. A plot of the expected value and IGS score of a respondentâ&#x20AC;&#x2122;s answers is the first step to estimate a value function as shown later.

10.

An IGS(A) =1 in say Kahneman-Tversky #7 is not identical/comparable to and IGS(A) = 1 in Kahneman-Tversky #8 because it depends also on the IGS(B), expected value, etc., in both questions.

Table 1. Calculating the IGS Value and Expected Value of Kahneman-Tverskyâ&#x20AC;&#x2122;s Questions.

Journal of Personal Finance

28

However, for the purpose of explaining Kahneman-Tversky’s results, Risktyle goes one step further: it calculates the prospect strength (PS) of each prospect (second column in Table 1). PS measures the total disparity in IGSA and IGSB, calculated by taking the absolute value of IGSB subtracted by the IGSA. To explain Kahneman-Tversky, for gains (losses), when PS is greater (less) than some threshold, investors choose B (choose A). The model reveals this threshold is 2, and it is possible for different questions and populations, this threshold could differ (a point for further research).

using the Risktyle model (column 5). Questions with a prime in column 1 indicate questions with expected losses. The perfect correspondence between the Risktyle model (column 5) and Kahneman-Tversky (column 2), potentially explains why the universe of Kahneman-Tversky’s respondents answered the way they did and validates the use of the Risktyle model to examine individual risk appetite.

EUT

RelativeBet Strength

KT 1

B

A

2.49

B

KT 2

A

A

0.07

A

KT 3

B

A

2.32

B

KT 4

A

A

0.28

A

KT 7

B

Equal

4.39

B

KT 8

A

Equal

0.01

A

KT 13

B

Equal

6.38

B

KT 14

A

Equal

1.00

A

KT 3’

A

B

2.32

A

KT 4’

B

A

0.28

B

KT 7’

A

Equal

4.39

A

KT 8’

B

Equal

0.01

B

KT 13’

A

Equal

6.38

A

KT 14’

B

Equal

1.00

B

The Risktyle Model: Validating Kahneman-Tversky The Risktyle model seeks to identify risk-seeking or risk-averse behavior by quantifying it. One might credibly ask: how do we know that the Risktyle model actually works? While many of our individual analyses of respondents appear satisfactory to them (see Section III), we determined that one form of validation would be to see if we can use this model to replicate Kahneman-Tversky’s results in a rational/consistent manner.

Kahneman-Tversky initially hypothesized that humans were influenced by a so-called “certainty effect,” and were, by definition, loss averse. Keeping those assumptions in mind, Risktyle focuses on certainty and risk-adjusted value. For convenience, the questions were ordered so Gamble A had a higher probability of uncertainty and Gamble B had a higher probability of the given outcome. In Table 2, we provide each of the questions taken from Kahneman-Tversky (column 1), the answer from the original Kahneman-Tversky (Column 2), the answer assuming EUT (Column 3), the PS of each (column 4), and the responses

PS Values

Analysis

Risktyle Predicted Answer

KT Original

KT Questions

Table 2: Explaining Kahneman-Tversky responses using the Risktyle model’s PS value

Table 3 explains this perfect correspondence by highlighting the PS value (column 1), what it implies in IGS values (column 2), the Kahneman-Tversky effect (column 3), and the behavior observed in Kahneman-Tversky’s population (column 4). This table and the embedded explanation validates Risktyle as a useful model to explain the risk attitudes of individuals, subgroups, and populations.

KahnemanTversky Effect

Behavior Observed in Kahneman-Tversky Population

Gains: PS > 2 (Kahneman-Tversky’s #1, 3, 7, 13)

Great disparity in IGS

Certainty effect

High expected value/high probability; pick bets that offers them the most certainty.

Gains: PS < 2 (Kahneman-Tversky’s #2, 4, 8, 14)

Little disparity in IGS

Certainty effect

Low expected value/low probability; pick bets that offer them the biggest outcome.

Losses: PS > 2 (Kahneman-Tversky’s #3’, 7’, 13’)

Great disparity in IGS

Loss aversion

High negative expected value/high probability; pick riskiest bet.

Losses: PS < 2 (Kahneman-Tversky’s # 4’, 8’, 14’)

Little disparity in IGS

Loss aversion

Low negative expected value/low probability; pick bets that offer lowest absolute loss.

Table 3. Specific Kahneman-Tversky behaviors perfectly explained by the Risktyle model

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

29

Section III—Four Individual Case Studies Using Risktyle Probably the most powerful aspect of the Risktyle model for advisors is it permits individual risk appetite analysis, based on individual responses to the questionnaire. In this section, we show wide dispersion across individual behavior of the four respondents. Individual risk appetites can be very different and hence generalizations about the shape of the value function (Figure 3) may not be appropriate. As a result, Risktyle can serve as a better tool to empower individuals to understand their individual biases.

The next stage of research plots the IGS of the responses for each of the 14 questions (vertical axis) relative to the expected value for each independent subgroup (horizontal axis). As noted earlier, because Kahneman-Tversky did not have a formal definition of “value” and did not ask respondents to quantify “value,” we use the IGS (the relative risk-adjusted value of each gamble) as a proxy for value. IGS serves as the “value” on the y-axis that Kahneman-Tversky proposed, and appears to capture the heterogeneity that Choi, Fisman, Gale, and Kariv (2007) identified and measured. Risktyle differs from the Choi, Fisman, Gale, and Kariv’s (2007) approach in that it is able to capture risk aversion and both on gains and losses, a critical aspect of Kahneman-Tversky’s findings. These 14 points help plot the “value function.” Quantifying Kahneman-Tversky answers (column 2, Table 1) into numerical scores—using the IGS values provided in Table 1 and the answers selected by the respondents—reveals an interesting chart (Figure 4). The Risktyle model’s estimate of the value function is a polynomial trendline of order 2, derived from a matrix formulation of the multiple regression model, and forces it to pass through the origin for these 14 data points. It is not a perfect fit because only 14 data points are used, and Kahneman-Tversky’s questions were not designed to be evenly distributed along the expected value (horizontal) axis, which opens an interesting avenue for future research discussed in Section V. However, the chart for Kahneman-Tversky’s aggregate responses appears to show loss aversion (for losses) and risk aversion (for gains), very similar to the shape hypothesized by Kahneman-Tversky (Figure 3), further validating the use of the Risktyle model to understand risk appetite. For example, by examining Figure 4, the advisor might conclude that the investor might not be inclined to buy insurance but rather gamble losses, and this could start a useful dialog between advisor and client.

Figure 4. Risktyle model “Value Function” for Aggregate KahnemanTversky responses.

We now present four case studies to show the diversity in value functions across difference age groups, financial literacy and gender.

Case Study 1: Risk Seeking/Loss Averse Behavior in an Older Female (Age 60–80) The general assumption in age-based portfolios is that as one ages, the portfolio should be rotated into safer assets.

However, the individual modeled in Figure 5a, defies this simple assumption. She consistently chose risky options for the gambles and one of the possible explanations for this behavior is that she can be considered affluent, with no major financial responsibilities. In this case, wealth and remaining life expectancy may be the primary determinant of risk appetite, not age.

Figure 5a: Risk Appetite of Female Adult (Loss Averse; Risk-Seeking on Gains)

30

Journal of Personal Finance

Case Study 2: Risk Averse on Gains and Losses in Investment Professional Male (Age 40–60) Unlike Case Study 1, this respondent demonstrated very risk averse behavior on both gains and losses (Figure 5b). Even though the net worth of this respondent was similar to the individual in Case Study 1, the risk aversion in this individual was probably driven by his health condition (heart patient) and the fact that he has 2 teenage sons whose education he has to finance. In this case, health, remaining life expectancy, and financial obligations determine risk appetite.

Figure 5c: Risk Appetite of Female College Student (Loss Averse; Risk Averse on Gains)

Figure 5b: Risk Appetite of Male Investment Professional (Risk Averse on Losses and Gains)

Case Study 3: Risk Averse on Gains; Loss Averse in Adult Female (Age 20–30) This individual is a business school student and displays the exact same profile as the aggregate Kahneman-Tversky respondent (Figure 5c). She manages her own small investment portfolio and has an interest in pursuing a career in finance.

Case Study 4: Risk Averse on Losses, Risk Seeking on Gains in Teen Male (Age 15–20) One of the authors of this paper has a profile that is very different from the other three case studies. This individual has little to no net worth but is a member of the Investment Club and has above average financial literacy (and understands the pernicious effects of student debt). The behavior exhibited by this individual is very different from the average teen (Figure 5d), and is risk averse on losses, but risk-seeking on gains.

Figure 5d: Risk Appetite of Male Teen (Risk Averse on Losses and Risk-Seeking on Gains) In short, these four case studies show individual risk appetite matters can be shaped by wealth, health, financial obligations, and must be tracked as every individual is unique and risk appetite could evolve as one ages or gains financial literacy!

Section IV—Subgroup Analysis: Impact of Gender, Age and Literacy To push the results further, Risktyle allows us to examine subgroup results as well since the survey was run on different groups of two potential financially literate extremes in a universe of people: 297 teens and 81 investment professionals, both sorted by gender. The broad results are: (A) the responses of pros, in aggregate, differ from Kahneman-Tversky’s universe (and women more so than men); (B) using aggregate results, the risk appetite of teens significantly differs from pros (as it pertains to losses), and further, pro-women significantly differ

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

31

from pro-men (as it pertains to gains); and (C) subgroup value functions explain the previous result by highlighting the differences in risk-appetite by size of gamble.

A. Aggregate Comparison: Pros and Teens vs. KahnemanTversky: Table 4 provides the aggregate response to Kahneman-Tversky questions (column 1) assuming EUT (column 2), Kahneman-Tversky universe (column 3) and current respondents (remaining columns). Figures in parentheses in the header indicate the number of respondents and figures in parentheses within the table indicate percentage of respondents who chose this answer. Table 4 highlights the fact that there are compelling differences between Kahneman-Tversky’s universe and ours, most significantly among the Investment Professionals (Kahneman-Tversky’s #4, 14 and 13’) in column 4, pro men (Kahneman-Tversky’s #13, 14 and 13’) and pro women (Kahneman-Tversky’s #2, 4, 14 and 13’). In the teen category, Kahneman-Tversky’s aggregate results synchronized perfectly with teens, though teen males differed from teen females for one answer (Kahneman-Tversky’s #4’). By generalizing results to apply to entire populations, Kahneman-Tversky may have made a similar simplifying assumption to that of Bernoulli, and von Neumann, and Morgenstern. Table 4 suggests shades of grey among any population, which requires further examination, and one has to look at individual value functions as opposed to a macro view.

Question

EUT

KT

Pro (81)

Further, there were several responses within each subgroup that did not meet the 90 percent confidence interval, giving greater support to the case for individual analysis in Section III. Once again, the major takeaway is that individuals need personalized guidance.

B. Aggregate Risktyle Scores for Pros and Teens (based on Gender) on Gains and Losses: Table 4 shows dispersion across subgroups in the aggregate results. Hence, a better tool is needed to describe and understand each subgroup.

The IGS numerical score permits adding these values across any subcategory (e.g., across gains for pro men) to allow more clarity on the characteristics of each subpopulation based on their selection for different types of gambles. Table 4 shows that pro men chose “B” for Kahneman-Tversky #1, and Table A.1 gives “B” an IGS score of 1.48; Kahneman-Tversky #2 = –0.04 and so on. Adding these, the total gain IGS of pro men can be calculated just for gambles with gains (and similarly for losses).

This pair of values (i.e., gains and losses) can then be plotted on a graph for each subgroup as in Figure 6: teen males (1 for gains; –10.9 for losses), pro men (–4.5 for gains, –4.5 for losses), teen females (1 for gains; –11.2 for losses), and pro women (2.2 for gains; –4.5 for losses). The more negative (positive) a sub group, the more risky (safe) its choices. The risk appetite for gains is on the vertical axis and the risk appetite for losses is plotted on the horizontal axis. The range of the axes are

Pro Men (40)

Pro Women (41)

Teen (297)

Teen Male (208)

Teen Female (89)

KT1

A

B (82%)*

B (83%)*

B (75%)*

B (90%)*

B (60%)*

B (60%)*

B (61%)*

KT2

A

A (83%)*

A (59%)*

A (73%)*

B (54%)

A (67%)*

A (68%)*

A (65%)*

KT3

A

B (80%)*

B (89%)*

B (80%)*

B (98%)*

B (70%)*

B (67%)*

B (79%)*

KT4

A

A (65%)*

B (52%)

A (60%)*

B (63%)*

A (61%)*

A (64%)*

A (55%)

KT7

A/B

B (86%)*

B (91%)*

B (85%)*

B (98%)*

B (82%)*

B (83%)*

B (80%)*

KT8

A/B

A (73%)*

A (65%)*

A (63%)*

A (68%)*

A (81%)*

A (79%)*

A (85%)*

KT13

A/B

B (82%)*

B (57%)

A (50%)

B (63%)*

B (75%)*

B (76%)*

B (75%)*

KT14

A/B

A (72%)*

B (58%)*

B (55%)

B (61%)*

A (59%)*

A (57%)*

A (65%)*

KT3’

B

A (92%)*

A (78%)*

A (65%)*

A (90%)*

A (71%)*

A (67%)*

A (81%)*

KT4’

B

B (58%)

B (65%)*

B (73%)*

B (59%)*

B (51%)

B (52%)

A (52%)

KT7’

B/A

B (92%)*

B (83%)*

B (78%)*

B (88%)*

B (68%)*

B (67%)*

B (69%)*

KT8’

B/A

A (70%)*

A (73%)*

A (78%)*

A (68%)*

A (52%)

A (51%)

A (53%)

KT13’

A/B

A (70%)*

B (75%)*

B (70%)*

B (80%)*

A (62%)*

A (63%)*

A (61%)*

KT14’

A/B

B (83%)*

B (65%)*

B (68%)*

B (63%)*

B (69%)*

B (68%)*

B (72%)*

Table 4. Summary of the results. Darker shading in column KT symbolizes Kahneman-Tversky results differing from EUT. Lighter shading in other columns symbolizes results differing from Kahneman-Tversky. “*” represents at least a 90 percent significance for given N.

32

Journal of Personal Finance

determined by Kahneman-Tversky’s questions and the Risktyle model (more negative = more risk seeking).

(Figure 4), suggesting a balance between risks taken for gains and losses. The aggregate scores plotted in Figure 6 seem to bear this out.

Figure 6: Comparing the Risk Appetite of Males and Females (sorted

by Age and Proficiency) using the IGS aggregates for Gains and Figure 7a. Pro Women’s Risktyle Model “Value Function”.

Losses for these subgroups.

Two things stand out in Figure 6: First, teen males and females tend to take similar risky bets when it comes to losses, significantly riskier than both professional subgroups, but both opt for the safe bet when it comes to gains. Vedantam (2001) reported that some teens are more likely to take risks than adults because of the rapid changing of the prefrontal cortex during adolescence. Allen (2014) reported that these changes may causes a “hormonal jolt” in some teens, causing them to embrace risky behavior. This has interesting implications for the design of financial literacy programs.

Second, the older respondents, intentionally the more financially literate and sophisticated, reveal a very interesting dispersion. While professional men and women tend to favor less risky gambles relative to teens in the context of losses, represented by the arrows moving to the right—which is supported

by Gaechter, Johnson, and Hermann (2007) which finds loss aversion decreases with education—professional men tend to prefer more risky gambles with respect to gains, represented by the downward-sloped arrow to the rectangle, and professional women prefer the safer alternative, signified by the arrow pointing to the triangle moving up. Butcher (2015) finds that women make better traders, even though they are found to be more risk averse, which our results seem to confirm.

Aggregate IGS gains and losses scores hide information about the risk appetite over different gambles, so Section C displays the Risktyle value function of these subgroups.

C. Evaluating Risk Appetite of Subgroups: For pro women (Figure

7a), the curve is much flatter than the Kahneman-Tversky curve

For pro men (Figure 7b), the tipping over of the curve, as it pertains to large gains, suggests extremely risky behavior (i.e., risk seeking) when there is a potential for a big outcome. This is what the aggregate score IGS score for gains in Section B (Figure 6) could not cleanly detect.

Figure 7b. Pro Men Risktyle Model “Value Function”.

Section V—Areas of Future Research The model was calibrated based on Kahneman-Tversky’s initial set of questions. We would look to enhance the model in the future to add more questions for an even more in-depth analysis, because as noted earlier, Kahneman and Tversky (1979) do not balance the questions evenly across the expected value or the IGS axes. A more even distribution of these questions would allow for better plotting of Risktyle’s estimated value function. Further, this paper fit a simple regression line through all 14 points. This allows for generalization of the individual’s overall behavior, but not specific behavior point. An alternate

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

approach would be to fit different lines through the respective IGS values for losses and gains, separately. This would model a “two function” approach that behavior towards losses and gains are separate behaviors rather than a single one.

This approach was not used in the paper as, on average, this technique creates two parabolic functions on either side of the y-axis, dissimilar to Kahneman-Tversky’s value function.

Further, since we have this data on individuals based on current responses, and there appear to be interesting changes in risk appetite as one ages and gains financial literacy, future research will try to consider a longitudinal survey where the evolution of individual risk appetite will be tracked over multiple years to see if there are any trends based on the evolution of individuals as opposed to based on our subgroups.

This model appears to have other applications as well. The dean of a business school with an entrepreneurship program stated they look for risk-seekers and risk-takers for their entrepreneurship program. However, they have no solid way of identifying these students. By being able to quantify risk behavior in individuals, this could be an option in potentially selecting students for programs. Additionally, research suggests that teenage girls choose professions that tend to be less “risky” and if educated about their bias, appear to consider other alternatives (Kachnowski 2017).

Finally, we have started to expand our database to include subgroups not considered for this study (e.g., adult males and females who are not financial professionals). The goal of this research would be to complete a “Risk GAP Map” (i.e., risk appetite sorted by gender, age and proficiency or GAP).

Section VI—Conclusion People around the globe are making financial decisions that make or break the quality of their lives at increasingly younger ages. Student loans are expected to financially ruin many of its debtors (FinAid 2016). The student debt crisis is only one indication that many individuals are making a variety of risky decisions without sufficient financial knowledge. As Merton (2014) notes, “To begin with, putting relatively complex investment decisions in the hands of individuals with little or no financial expertise is problematic. Research demonstrates that decision making is pervaded with behavioral biases.”

Financial decisions, such as financing student loans and saving for college, housing, healthcare, life insurance purchases, and

33

retirement, indicate a critical need for a user-friendly risk assessment tool to gauge risk appetite and take an individual’s financial temperature. Risktyle is one potential tool, and it complements Kahneman-Tversky’s ground-breaking theory by quantifying their value function. Risktyle is the first attempt to explain Kahneman-Tversky’s findings via a simple, quantifiable model. Essentially, Risktyle creates a potential value function of prospects by calculating the IGS value of each gamble and educating the users as to their biases.

Compelling results are evident from polling diverse groups, based on gender, age, and level of financial proficiency.

The findings reveal important differences in individual and subgroup behavior regarding risk, based on certain biases, and it is too simple, even “irrational,” to have a single model for individual behavior regarding risk. For example, investment professionals, both men and women, differ meaningfully from Kahneman-Tversky’s respondents, whereas teens, both male and female, prove very similar to Kahneman-Tversky’s results.

Second, Risktyle reveals a behavioral transition in males and females as they age; both become more risk averse to losses.

Males, however, become more risk-seeking, and females become more risk-averse when it comes to gains. Moreover, individual behavior over specific bets was explained by graphing their Risktyle value function. This shows that individuals behave vastly different from one another in how they face risk and establishes that a macro-based perspective on diverse populations may potentially disadvantage certain individuals. Further, as the case studies showed, health, wealth, and financial obligations may also determine risk appetite.

Risktyle can be used as an effective diagnostic tool. Similar to medical check-ups that track a person’s physical health, Risktyle can track individuals current approach to risk, how they change their behavior over time because of changes in life circumstances, and whether or not they want to alter their investment behavior and/or need to advance their financial literacy. For example, if a young adult shows up as very risk averse, they would be potentially advised to increase their portfolio risk so that they may reach a point later in life where they are happy with the money they can retire with. Similarily, one who is loss averse might shun the purchase of insurance which could be detrimental long-term.

Choi, Fisman, Gale, and Kariv (2007) stated, “Because uncertainty is endemic in a wide variety of economic circumstances, models of decision making under uncertainty play a key role

Journal of Personal Finance

34

in every field of economics.” This paper presents a robust mathematical model of decision making under uncertainty that displays the uniqueness of individuals.

Risktyle could make sure an individual’s portfolio is not out of style with his or her risk appetite!11 11.

To find out your risk appetite using the Kahneman-Tversky survey, visit Risktyle.com

Appendix I: Kahneman and Tversky’s 14 Selected Questions Kahneman-Tversky #1 Option A

Kahneman-Tversky #2 Option B

33% chance of winning $2,500 66% chance of winning $2,400

100% chance of winning $2,400

Option A

Option B

33% chance of winning $2,500

34% chance of winning $2,400

67% chance of winning $0

66% chance of winning $0

1% chance of winning $0 Kahneman-Tversky #3

Kahneman-Tversky #4

Option A

Option B

Option A

Option B

80% chance of winning $4,000

100% chance of winning $3,000

20% chance of winning $4,000

25% chance of winning $3,000

Kahneman-Tversky #7

Kahneman-Tversky #8

Option A

Option B

Option A

Option B

45% chance of winning $6,000

90% chance of winning $3,000

0.1% chance of winning $6,000

0.2% chance of winning $3,000

Kahneman-Tversky #13 Option A 25% chance of winning $6,000

Kahneman-Tversky #14 Option B

Option A

Option B

25% chance of winning $4,000

0.1% chance of winning $5,000

100% chance of winning $5

25% chance of winning $2,000

Kahneman-Tversky #3’

Kahneman-Tversky #4’

Option A

Option B

Option A

Option B

80% chance of losing $4,000

100% chance of losing $3,000

20% chance of losing $4,000

25% chance of losing $3,000

Kahneman-Tversky #7’

Kahneman-Tversky #8’

Option A

Option B

Option A

Option B

90% chance of losing $3,000

45% chance of losing $6,000

0.2% chance of losing $3,000

0.1% chance of losing $6,000

Kahneman-Tversky #13’ Option A 25% chance of losing $6,000

Kahneman-Tversky #14’ Option B

Option A

Option B

25% chance of losing $4,000

0.1% chance of losing $5,000

100% chance of losing $5

25% chance of losing $2,000

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Acknowledgements Thank you to Marie Pillai for reaching out to several of her esteemed colleagues to take the survey. Thank you to Sachin Muralidhar for approaching several of his peers to take the survey and provide the teen perspective. Thank you to Scott Becker for reviewing our calculations. Thank you to Dr. Ronald Mainieri, Dr. Richard Monteverde, Ms. Jo-Ann Muir, Harish Neelakandan, Dr. Ramchand, Dr. Towle, Dr. Woerheide, and Dr. Pfau for their mentorship and feedback on presenting our ideas with more clarity and focus. A heartfelt thank you to our parents, for their support and trust in allowing us to fully immerse ourselves in this research while neglecting our usual chores! Finally, a big thank you to Professor Daniel Kahneman, Professor Andrew W. Lo, Professor Hersh Shefrin, Professor Meir Statman, and Professor Shachar Kariv for their kind words of encouragement; we are truly standing on the shoulders of giants!

Works Cited Allen, A. 2014, September 2. Risky Behavior by Teens Can Be Explained in Part by How Their Brains Change. The Washington Post. http://www.highbeam.com /doc/1P2-37134211. html?refid=easy_hf Baldiga, K. 2014. Gender Differences in Willingness to Guess. Management Science, 60(2), 434–448. doi:10.1287/ mnsc.2013.1776 Bernoulli, D., 1954 (1738). Exposition of a New Theory on the Measurement of Risk, Econometrica, 22: 23–36.

Butcher, S. 2015. Data proves that women make far better traders than men. Retrieved http:// news.efinancialcareers.com/us-en/197066/ exclusive-figures-show-women-make-far-better-traders-men Camerer, Colin F. 1995. Individual decision making. In A. E. Roth and J. Kagel (Eds.), Handbook of Experimental Economics. Princeton: Princeton University Press.

Camerer, Colin F. 2001. Prospect Theory in the Wild: Evidence from the Field. In: Choices, Values, and Frames. Contemporary Psychology. No.47. American Psychological Association , Washington, DC, pp. 288–300.

35

Choi, S., Fisman, R., Gale, D., & Kariv, S. (2007). Consistency and Heterogeneity of Individual Behavior under Uncertainty.

American Economic Review, 97(5), 1921–1938. doi:10.1257/ aer.97.5.1921 Experimental Economics Center. n.d. Von Neumann-Morgenstern EUT. http://www.econport.org/content/handbook/decisions-uncertainty/basic/von.html FinAid. 2016. Student Loan Debt Clock. http://www.finaid.org/ loans/studentloandebtclock.html.

Gaechter, S. and Johnson, Eric J. and Herrmann, A. July 2007.

Individual-Level Loss Aversion in Riskless and Risky Choices. IZA Discussion Paper No. 2961. Available at SSRN: https://ssrn.com/ abstract=1010597 IDEAS. n.d. Top 1‰ Research Items by Number of Citations. https://ideas.repec.org/top/top.item.nbcites.html Lusardi, A., and Mitchell, Olivia S. 2014. The Economic Importance of Financial Literacy: Theory and Evidence. Journal of Economic Literature, 52(1): 5–44.

Kahneman, D., Knetsch, J. L., & Thaler, R. H. 1991. Anomalies: The Endowment Effect, Loss Aversion, and Status Quo Bias.

Journal of Economic Perspectives, 5(1), 193–206. doi:10.1257/ jep.5.1.193 Kahneman, D., Tversky, A. 1979. Prospect Theory: An Analysis of Decision under Risk (PDF). Econometrica. 47 (2): 263. doi:10.2307/1914185. ISSN 0012-9682.

Martin, R. 2004. The St. Petersburg Paradox. In Edward N. Zalta.

The Stanford Encyclopedia of Philosophy (Fall 2004 ed.). Stanford, California: Stanford University. ISSN 1095-5054. Retrieved 2016-12-20.

Merton, R. C. 2014, July & Aug. The Crisis in Retirement Planning. https://hbr.org/2014/07/the-crisis-in-retirement-planning Modigliani, F. 1997. “Risk-Adjusted Performance.” Journal of Portfolio Management. 1997 (Winter): 45–54.

Muralidhar, S. and A. Pamecha. (2016). “Student Tuition: Unburdening the Debt for Youth in America.” Investments and Wealth Monitor.

36

Journal of Personal Finance

Muralidhar, A. 1999. An Explanation for the Discount/Premium Puzzle in Currency Markets—the Impact of Agency Problems and Capital Constraints. JP Morgan Fleming Asset Management, Investment Insight.

Sharpe, William F. 1994. “The Sharpe Ratio.” Journal of Portfolio Management. 1994 (Fall): 49–58.

Shefrin, H., & Statman, M. (1985). The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence.

The Journal of Finance, 40(3), 777. doi:10.2307/2327802 The Economist. 2015, January 02. That ranking. http:// www.economist.com/blogs/freeexchange/2015/01/ influential-economists The Nobel Foundation. n.d. The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2002. http:// www.nobelprize.org/nobel_prizes/economic-sciences/ laureates/2002/

About the Authors Emerson Berlik: Emerson Berlik is a senior at Thomas Jefferson High School for Science and Technology, where he is currently the co-president of his school’s chess team and mobile app club. He is the co-founder and the Chief Technical Officer of www.Risktyle.com.

Sidharth Muralidhar: Sidharth Muralidhar is a senior at Thomas Jefferson High School for Science and Technology, and is a co-founder and the Chief Executive Officer of www.Risktyle. com. He has previously published an article on an innovative solution to the student debt crisis in IMCA’s Investments & Wealth Monitor (2016), is a two-time MIT INSPIRE finalist in Economics, placing 3rd in 2017, and was a member of the topranked U.S. team at the Knowledge@Wharton Global Investment Competition in 2017.

To identify your risk appetite go to www.Risktyle.com.

Vedantam, S. 2001, June 3. Are Teens Just Wired That Way? Researchers Theorize Brain Changes Are Linked to Behavior. The Washington Post. http://www.highbeam.com/doc/1P2-439122. html?refid=easy_hf

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

37

Life Quality and Health Costs in Late Retirement

Yuanshan Cheng, Ph.D., Assistant Professor, Winthrop University, Rock Hill, SC Philip Gibson, Ph.D., Assistant Professor, Winthrop University, Rock Hill, SC Tao Guo, Ph.D., Assistant Professor, William Paterson University, Wayne, NJ

Abstract Individuals are living longer due to the advancement of medical technology and nutrition quality. Are the elderly enjoying retirement in those extended years with good quality of life or are they simply alive? Using data from the Health and Retirement Study (HRS) and the Consumption and Activities Mail Survey (CAMS), this study contributes to the literature by presenting empirical evidence on how individuals spend time in retirement. The results show that retirees on average do not spend their time significantly different throughout retirement. Most life tasks such as reading the paper or magazines, listening to music, playing sports or exercising, visiting others, and house cleaning are similar among retirees in different age groups. We also present evidence that retirees on average experience a spike in medical expenses late in retirement. We compare systematic withdrawal strategies with and without health costs risk quantifying the impact on portfolio sustainability.

38

Journal of Personal Finance

Introduction Due to the recent shift from defined benefit plans to defined contribution plans, more attention is being given to the study of retirement. Many retirees spend close to one-third of their lives after exiting the labor market in retirement. During this time, many rely on their accumulated savings for consumption.

Recent advances in medical treatment and nutrition science have increased the life expectancy of retirees. However, the quality of life during these extended years is still relatively unknown. As more baby boomers enter retirement, academics and practitioners have a stake in learning how individuals are spending their time in retirement. Practitioners have little evidence to rely on when and how much the average retiree can expect to spend on out-of-pocket medical expenses during retirement. Compared to the Consumer Price Index, the higher inflation rate associated with health care costs adds more uncertainty to the retirement planning process. This paper uses the data from the Health and Retirement Study (HRS) and the Consumption and Activities Mail Survey (CAMS) to examine the time use and consumption patterns of the elderly. The results reveal that on average, the quality of life among the elderly is good and remains stable until the very end. While we can demonstrate that retirees in general experience a stable quality of life, we also find that the average amount and the volatility of out-of-pocket medical expenses increase dramatically late in retirement. Empirical results of the HRS are incorporating into a simulation model to assess the impact of health cost risk on portfolio sustainability. Our findings provide implications for policy makers, financial advisors, and retirees regarding both the quality of life and the health care cost risks in retirement.

Literature Review In the U.S., many baby boomers are entering retirement without the savings needed to maintain their pre-retirement standard of living (Moore and Mitchell, 1997). As the number of baby boomers entering retirement increases, understanding the quality of life for retirees is needed for retirement studies.

Most studies on retirement planning focused on two areas, improving financial satisfaction and health satisfaction of the elderly (Wise, 2014). For example, using historical financial market return data, Bengen (1994) was among the first to study sustainable withdrawal rates in retirement. Since then, many other studies have set out to identify a sustainable rate of retirement spending using different methodologies

and assumptions. For a thorough review of the literature on sustainable withdrawal rate, see Kitces (2014). In addition to sustainable withdrawal rates, other studies have explored ways to reduce longevity risk. For instance, Benartzi, Previtero and Thaler (2011) suggested that households should annuitize a portion of their financial wealth at retirement to hedge against the risk of outliving their assets. Likewise, Finke and Pfau (2015) explored the benefit of incorporating annuities into retirement distribution strategies. More recently, researchers have begun to study the use of Home Equity Conversion Mortgages (HECM) as a source of additional liquidity in retirement (Davidson & Taylor, 2015; Tomlinson, Pfeiffer, & Salter, 2016). However, these studies focus on financial sustainability and pay little attention to the health condition of retirees and how they spend their time throughout retirement.

Individuals’ health conditions not only affect their life expectancy but also impact their health care expenditures. Thus, incorporating health care cost in retirement planning is important. Feenberg and Skinner (1992) argued that health care risk is the largest risk faced by many retirees. Examining the cost of health care in retirement, Goda, Shoven, and Slavov (2011) illustrated that households’ medical expenses grow at a much faster rate than Social Security benefits. The high cost of health care creates a financial burden for many households, in particular for the elderly who need the constant care of medical professionals. As expected, the burden of high medical expenses in retirement negatively affects portfolio sustainability. Using bootstrap simulation, Drew, Walk, and West (2016) looked at the consequences of unexpected health care costs coupled with longevity in retirement. They found that when households are required to pay for unplanned medical expenses, it increases the risk of an early portfolio depletion.

How people spend time has significant implications on their happiness and life quality. In his seminal article, Becker (1965) introduced the standard household economic theory where he argued that both time and consumption are important factors that influence the utility derived by individuals. Concerning spending, Hurd and Rohwedder (2003) demonstrated that time use and home production are the primary reasons for the consumption shortfall during the transition to retirement, known as the “retirement consumption puzzle.” More recently, Kalenkoski and Oumtrakool (2014) were among the first to explore how time use changes for individuals before and after retirement. Building on the prior literature, we present empirical evidence on time use of the elderly and draw implications of

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

their quality of life and habit over different retirement stages.

Additionally, we explore how out-of-pocket health care costs change throughout retirement. Using this information, we present a simulation model that examines the impact of outof-pocket healthcare cost on retirees’ portfolio.

Data and Methodology The data for this study comes from the Health and Retirement Study (HRS) and the Consumption and Activities Mail Survey (CAMS). The HRS is a longitudinal panel study from the Survey Research Center at the University of Michigan. It is widely used in studies related to retirement issues. Starting from September 2001, households from the HRS were randomly selected to participate in the CAMS database. Since then, CAMS data have been collected biannually to create multiple waves. This study uses multiple waves of the CAMS data to analyze how retirees spend time in retirement and draw implications about their life qualities. Health care costs are analyzed using the RAND imputed HRS core data. We use RAND HRS because it offers more observations and covers more areas of health costs when compared with CAMS.

Cross-sectional Analysis of Time Use In Table 1, we present the descriptive statistics from the latest wave of CAMS, which comes from the 2015 database. The HRS focuses on the elderly, which makes it the most appropriate dataset for the purpose of this study. Respondents are divided into four categories based on their age. Respondents whose age is greater than or equal to 50, but less than 65 are categorized as “Close to Retirement”; those whose age is greater than or equal to age 65 but less than age 75 are classified as “Early Retirement”; those whose age is greater than or equal to age 75 but less than age 85 are classified as “Mid Retirement”; and those whose age is greater than or equal to age 85 is grouped into “Late Retirement”. Even though some respondents who are older than age 65 are still partially employed, grouping respondents this way helps us to analyze retirees during different stages of retirement. The results in Table 1 show the summary statistics for the time-use variables from the CAMS database for all retirement categories. As shown, there are no significant differences in hours spent watching television, reading the paper or magazines, reading books, listening to music, and sleeping. Across those different retirement groups, we observe similarities in the time spent walking, playing sports or exercising, visiting others, and communicating with

39

others. Some differences can be seen in the time spent using computers. However, this might be attributed to computer use habits of the different cohorts. The amount of time allocated to shopping, meal preparation, and pet care appears to be fairly consistent until late retirement. As individuals get older, the number of hours spent helping others, taking care of grandchildren, doing volunteer work, attending religious activities and other meetings slowly declines.

Panel C of Table 1 shows the variation in spending from CAMS. Consistent with the expectations that individuals like to travel early in retirement, we find that those who are between the ages of 65 and 75 appear to spend the most on trips and vacations. On average, we observe that the money spent on trips and vacation is five times greater during early retirement when compared with late retirement.

In examining the amount of money dedicated to medical expenses, we find that the average amount of money spent on over-the-counter and prescription drugs appears to be consistent throughout retirement. The stable out-of-pocket cost for prescription drugs can be attributed to the coverage offered through Medicare. Unsurprisingly, out-of-pocket spending on health care services is consistent among our first three groups.

However, it increases significantly during late retirement.

CAMS does not include all out-of-pocket health costs, therefore, we turn to the imputed out-of-pocket medical cost from the RAND HRS for health cost analysis.

Table 2 presents the average and standard deviation of outof-pocket medical expenses incurred by retirees. The variable we use to measure out-of-pocket medical costs is RwOOPMD from the RAND HRS. It covers hospital costs, nursing home costs, doctor visits costs, dental costs, outpatient surgery costs, average monthly prescription drug costs, home health care costs, and special facilities costs. It is imputed by RAND Corporation. We sort individuals according to their household wealth (wealth including house and wealth not including house) into five groups. For individuals who have little wealth, it is very likely they will depend on the government health care system which includes both Medicare and Medicaid. Hence, we focus on those individuals who will rely on their household wealth to pay for much of their medical expenses. Table 2 shows households that are in the top two quintiles of wealth. The results reveal that both the average and the standard deviation of out-of-pocket medical expenses remain relatively stable and have a sharp increase late in retirement. This is consistent across all wealth levels. The risk that retirees and financial

Journal of Personal Finance

40

Table 1: Average by Different Age Groups (CAMS 2015) Panel A (Hours spent last week) Watching programs or movies/videos on TV Reading newspapers or magazines Reading books Listening to music Sleeping and napping (including at night) Walking Participating in sports or other exercise activities Visiting in-person with friends, neighbors, or relatives Communicating by telephone, letters, email, Facebook, Skype, or other media with friends, neighbors, or relatives Working for pay Using the computer Praying or meditating House cleaning Washing, ironing or mending clothes Yard work or gardening Shopping or running errands Preparing meals and cleaning up afterward Personal grooming and hygiene, such as bathing and dressing Caring for pets Physically showing affection for others through hugging, kissing, etc.

Close to Retire

Early Retirement

Mid Retirement

Late Retirement

N = 2233

N = 1456

N = 1200

N = 329

19.43 3.23 3.38 8.77 43.85 8.82 2.52 6.77

23.54 4.70 4.35 6.04 46.01 6.51 2.28 7.36

24.10 5.80 4.19 4.80 46.96 6.32 2.22 7.18

23.12 6.61 4.30 4.04 51.57 5.82 1.71 6.91

7.64

6.69

5.82

4.27

20.84 12.94 4.85 5.12 2.65 2.06 3.97 6.35 7.34 3.93 4.15

5.35 8.74 4.23 4.87 2.41 2.53 4.16 6.40 6.64 2.97 2.72

1.71 5.29 4.72 3.99 2.20 2.42 3.74 6.14 6.80 2.14 2.68

0.03 2.70 4.59 2.21 1.44 1.15 2.48 4.67 6.65 0.92 2.29

6.46

5.73

3.40

1.41

13.57

7.51

4.25

1.36

2.83

3.43

2.72

1.49

3.50 1.45

4.05 1.66

4.57 1.75

3.73 1.61

3.51

3.71

3.63

3.42

10.08 4.64 1.48 1.50 2.04

7.18 6.59 1.57 1.39 2.36

8.78 5.84 1.26 0.76 2.37

6.41 6.75 1.07 0.64 2.16

3.83

3.47

2.53

0.62

1.74 4.63

1.63 6.50

1.28 5.85

0.90 5.93

1,716

2,041

1,272

402

1,895 1,756

1,643 2,154

1,680 1,952

1,980 1,845

631

957

936

940

1,386

1,122

1,157

1,573

257

190

222

319

Panel B (Hours spent last month) Helping friends, neighbors, or relatives who did not live with you and did not pay you for the help Taking care of grandchildren Doing volunteer work for religious, educational, health-related, or other charitable organizations Attending religious services Attending meetings of clubs or religious groups Taking care of finances or investments, such as banking, paying bills, balancing the checkbooks, doing taxes, etc.

Treating or managing an existing medical condition of your own Playing cards or games, or solving puzzles Attending concerts, movies, or lectures, or visiting museums Singing or playing a musical instrument Doing arts and crafts projects, including knitting, embroidery, or painting Making home improvements, including painting, redecorating or making home repairs Working on, maintain, or cleaning your car(s) or vehicles(s) Dining or eating outside the home (not related to business or work) Panel C (Money spent last year, $) Trips and vacations: including transportation, accommodations, and recreational expenditures on trips Gasoline Health insurance: out-of-pocket, including Medicare supplemental insurance Prescription and nonprescription medications: out-of-pocket costs, not including what’s covered by insurance Health care services: out-of-pocket cost of hospital care, doctor services, lab tests, eye, dental, and nursing home care Medical supplies: out-of-pocket costs, not including what’s covered by insurance

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

41

Table 2: Average of Out-of-Pocket Medical Costs from RAND HRS Rank for Total Household Wealth

Wealth—lowest

1

2

3

Wealth— Highest

Retirement Category

Mean

Std

Close

2,027

6,091

Early

2,248

5,895

Mid

3,489

12,417

Std as % of Total Wealth

—

Rank for Nonhouse Household Wealth

Std as % of Annual Consumption

Mean

2,390

6,773

2,797

8,572

3,671

12,310

Late

8,227

20,404

7,957

21,122

Close

2,755

7,486

2,107

6,572

Early

2,975

8,963

Mid

3,031

7,501

—

2,817

9,673

3,154

10,188

Late

8,262

25,365

6,843

21,241

Close

2,552

4,572

2,680

5,119

Early

3,043

8,250

2,649

4,640

Mid

3,376

9,951

3,174

7,339

—

Late

7,091

21,809

Close

3,117

7,919

2.16%

Early

2,758

4,620

Mid

4,085

9,663

Std

Std as % of Non-house Wealth

Std as % of Annual Consumption

—

—

—

6,746

19,366

16.58%

3,073

7,603

4.40%

15.81%

1.26%

9.67%

2,889

4,909

2.84%

10.21%

2.63%

20.23%

3,984

8,668

5.02%

18.02%

Late

7,523

23,368

6.37%

48.93%

7,831

23,567

13.64%

49.01%

Close

3,075

4,999

0.30%

7.65%

3,283

5,589

0.45%

8.78%

Early

3,414

6,298

0.38%

9.64%

3,422

6,523

0.53%

10.24%

Mid

4,473

11,980

0.72%

18.34%

4,555

12,726

1.02%

19.98%

Late

8,878

23,006

1.38%

35.23%

10,210

26,831

2.16%

42.13%

advisors should be concerned about is the variations of outof-pocket health care costs late in retirement. In examining out-of-pocket medical expenses as a portion of total household wealth, we find the standard deviation of out-of-pocket expenditures to be around 2 to 3 percent for households with wealth level in the fourth quintile. It is close to one percent for households ranked in the top quintile. However, in late retirement, the cost jumps dramatically. It is larger when we exclude the house value from the total wealth as shown from the second panel. The impact of out-of-pocket health care cost is bigger when compared with annual consumption. One standard deviation accounts for 42.13 percent of an average household annual consumption for those in the top 20 percent of nonhouse wealth. During retirement when the financial wealth is decreasing, the impact of out-of-pocket medical expenses on the quality of life can be much larger. While multiple research articles discuss the benefits of using an annuity (for instance, Benartzi et al., 2011), our results imply that it might be optimal for many households to maintain a portion of their financial wealth for self-insurance purposes.

Table 3 presents empirical evidence using an ordinary least squares regression analysis. The dependent variable is the time spent on each activity or the consumption in each category.

The key independent variable is retirement category. The group of respondents less than age 65 is used as the reference group.

The average time spent on those activities for this reference group is provided in the first row. For example, respondents younger than age 65 spend on average 2.57 hours on sport and exercise. As individuals get older, we observe a decrease of 0.62 hours in the amount of time dedicated to sports and exercise in early retirement. No significant differences are observed among those in mid and late retirement. For hours spent visiting others, no significant differences can be seen across the different retirement stages. For hours spent using computers, we do see that retirees spend less time using computers as they are getting older, and it is consistent and significant. We observe a decline in the time spent on house cleaning and shopping only in late retirement. It is in line with the previous observation from descriptive statistics that retirees do not have a significant lifestyle change until late retirement. Again, late

Journal of Personal Finance

42

Table 3: Ordinary Least Square Regression Using CAMS 2015 Time use

Reference (Age<65) group mean

Money ($)

Sports/ exercise

Visit in person

Use computer

House cleaning

Shop

Trip and vacation expenses

Rand Total Med

2.57

6.58

12.85

5.12

3.95

$1,690

$2,665

Early

–0.62**

0.13

–4.70***

–0.29

Mid

–0.22

–0.27

–7.10***

–0.98***

–0.07

0.29

–393***

139

Late

–3.00***

293 624

–0.64

–0.47

–9.73***

–1.17***

–987***

Married

+

+

–**

–

–***

–***

1588*** +

Male

–***

+

+*

+***

+***

–

+

Education Rank

+***

+**

+***

–***

+

+***

HH wealth ($1000)

+*

+**

+

–

+**

+***

–

HH consumption ($1000)

+

–

+***

–

–

+***

+

Note: + indicates positive coefficient, – indicates negative coefficient. ***, **, * indicates significant level of <0.01, <0.05, <0.1

retirement is defined as respondents who are at least 85 years old in this study. Even with a big and statistically significant decline in time use, given the average time spent on those activities for the reference group, retirees in late retirement still do some house cleaning and shopping, but not as much as it was in early and middle retirement. As confirmed from the last two columns of regression on money spent, retirees do spend significantly less in traveling but more in out-of-pocket medical costs in late retirement.

There are some significances in control variables. With higher education and more wealth, individuals are more likely to spend more time on sports and exercise in addition to visiting others. Education and consumption can also help explain some variations of time use in computer and house cleaning. Gender is also significant in explaining time spending on multiple activities.

Cross-wave Analysis of Survivor’s Time Use and Retirement Consumption Thus far, we have looked at the time use and the consumption of retirees who fall into different age groups in 2015. To avoid cohort differences, we now look at the aggregate changes in time use and consumption by the same households. This section reduces the noise generated from those respondents who passed away during those survey years. It focuses only on survivors, who entered the survey in the year 2001 and are alive in the year 2015. Starting in September 2001, CAMS surveyed 5,000 households by mail. Those same households were later

interviewed in 2003, 2005, 2007, 2009, 2011, 2013 and 2015. In this section of the analysis, we focus on those respondents who were selected for the CAMS survey in the year 2001 and were alive in the year 2015. It is important to note that each individual may have a different lifestyle, and thus, have different patterns of time use. Hence, we examine the changes in time use relative to the initial report in 2001 for each individual. Age groups are created following the same method in the previous section based on respondents’ age in 2015. Figure 1A shows the changes in hours spent on playing sports or exercising, shopping, visiting others, and using computers. There is a relatively large decline in the time spent playing sports or exercising for those respondents who are in late retirement. Figure 1B shows the changes in the time spent doing housecleaning and the amount of money spent on annual trips and vacations.

The results indicate a gradual decline in the time spent doing household chores throughout retirement. We also see that spending on trips and vacations begins to decrease among individuals in middle retirement. Figure 2 shows the average and standard deviation of total out-of-pocket medical expenditures using imputed RAND HRS datasets. There is an increasing trend in both the average and the volatility of out-of-pocket medical expenditures across retirement groups.

Trace Back from Death In our previous analysis, we focus on individuals who are alive in 2015. Our results reveal that the time use and the consumption do not change until late retirement. However, it is possible

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Figure 1A: Time use (hours last week adjusted) from cross wave analysis for survivors

Figure 1B: Time use (hours last week adjusted) from cross-wave analysis for survivors

Figure 2: Average and standard deviation of total out-of-pocket medical expenditures from cross-wave analysis for survivors

43

44

Journal of Personal Finance

Figure 3: Time use and out-of-pocket medical expenditure by years before death

that time use and consumption could be different for people with the same age who die during different years. To answer this question, we explore the time use and the consumption patterns of individuals before death. Figure 3 shows the time-use statistics. The horizontal axis illustrates the number of years before death. We rescale time uses for each year as percentages of the same time-use activities 12 years before death. Figure 3 shows the time spent using computers, visiting others, shopping, house cleaning, playing sports and exercising from top to bottom. The graphs show that 2 years before death the time spent shopping decreases by 40 percent compared to time shopping by those respondents 12 years before death.

For respondents who are 2 years away from death, they devote 60 percent less time to sport and exercising compared to those who are 12 years away from death. Similar results can be seen in the time dedicated to house cleaning. As mentioned before, the increase in the amount of time spent using computers and technology can be attributed to overall trend of computer usage among the U.S. population. Visiting others does not decline, which indicates that retirees are still socially active even within 2 years before death. However, the opposite is observed for medical expenses. Both the average and the standard deviation of annual out-of-pocket medical expenses increase as individuals approach death. Figure 3 shows that one

standard deviation is around $20,000 for respondents that are 2 years away from death. The cost is much higher for households with more wealth as shown in Table 2.

Sustainable Spending Simulation In this section, we use simulation to illustrate how financial advisors can incorporate health costs into their planning. While ignoring health care cost presents additional risk during retirement, just adding it to current consumption levels can overestimate the financial impact. Retirees who experience high out-ofpocket medical expenses may undergo some lifestyle changes.

Empirical evidence examining the relationship between nonmedical expenditures and out-of-pocket health costs are explored. Table 4 shows the coefficient estimates for the ordinary least square regression analysis of health care costs on nonmedical expenses. Total household consumption from HRS RAND CAMS is obtained along with out-of-pocket medicals costs from RAND HRS Core dataset. Our sample includes households with the top 40 percent wealth level and older than age 65. HRS survey waves 6 to 12 are used for estimation. Fixed effects at the individual level are used to eliminate any potential bias due to omitted variables. Net worth and income levels for each household are used as control variables. When retirees

Š2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

45

Figure 4: A simulation framework for health care cost risks Panel A: Multivariate Lognormal Asset Return Assumptions Correlation Coefficient Arithmetic Mean

Standard Deviation

Equity

Bond

Equity

5.10%

20%

1

0.1

Bond

0.30%

7%

0.1

1

Panel B: Distribution of Health Cost Sample by Age

Frequency and average of out-of-pocket cost by age

Panel C: Probability of Success with Annual Consumption

Journal of Personal Finance

46

Table 4:The Impact of Medical Costs on Nonmedical Consumption Coefficient P- value

>$5,000

>$10,000

>$15,000

>$20,000

â&#x20AC;&#x201C;0.78

â&#x20AC;&#x201C;0.59

0.167

0.381

***

Not significant

Not significant

Not significant

Net worth

Controlled

Income

Controlled

Fixed Effect

Yes

Yes

Yes

Yes

N

814

221

95

53

Note: ***,**,* indicates significant level of <0.01, <0.05, <0.1. Table 4 shows the coefficient estimates of Ordinary Least Square Regression analysis of medical costs on nonmedical consumptions. Total household consumption from HRS RAND CAMS are obtained together with out of pocket expenses from RAND HRS Core dataset. Wave 6 to wave 12 are used for estimation. Fixed effects at the individual level are controlled to estimated biases due to omitted variables. Net worth and income levels for each household are controlled.

have to spend more money and time pursuing health care, the consumption of other activities decreases. Regardless, some basic consumption needs must be met. Thus, health care costs will only reduce nonmedical consumption to a certain level, after which the impact is marginal. Multiple thresholds of outof-pocket medical expenses are used to identify the nonlinear relationship. As shown in Table 4, when annual out-of-pocket medical expenses are above $10,000, the negative impact on nonmedical consumption disappears and is no longer statistically significant. For this reason, we only incorporate annual health cost in our simulation when it is over $10,000.

Using data from RAND HRS, households whose total wealth level (not including house) in the first quintile in survey year 2001 (wave 5) to 2015 (wave 12) are included into one health cost database. This newly created database contains annual out-of-pocket medical expenses at various ages. All medical costs in each wave are adjusted to the nominal dollar of survey year 2015 based on consumer price index of medical expenses reported by the Bureaus of Labor Statistics. We implement a simulation strategy that takes both financial risk and medical cost risk into consideration. For financial risk, we simulate one thousand portfolio-risk scenarios over 30 years using multiple asset allocations of stocks and bonds. The parameters are shown in Panel A of Figure 4. We follow the simulation strategy used by Pfau (2013) by assuming return, standard deviation, and correlation of both bonds and stocks. Health cost risks are simulated by randomly drawing one thousand sequences of out-of-pocket medical expenses at each age between 66 and 95. Panel B of Figure 4 shows the pooled sample where we draw health costs. There are over 1,000 respondents at age 65, and it declines to around 200 for age 90. We find that both the average and standard deviation of out-of-pocket medical

expenses goes up when age increases. We use the average total wealth ($1,546,727, not including house) as the initial wealth for our simulation in calculating the sustainable consumption amount.

Panel C shows the probability of success at the end of year 30 by varying consumption amount. Here we look at how the cost of an out-of-pocket health care shock affects a portfolio with different asset allocation. The results of our simulation show that the health cost risk is significant for portfolio sustainability if advisors want to ensure the probability of success to be over 90 percent.

Conclusions and Discussions A large body of retirement-planning literature study shows how retirees should spend their time and savings to maximize their lifetime utility. Time use and consumption of individual households are closely related and are fundamental aspects of the well-being of individuals. We observe, on average, that the time retirees spend on daily activities is consistent among different retirement stages until a very short period before death. Assuming life quality can be observed from daily activities and consumptions, it suggests that many retirees can maintain a consistent quality of life throughout most of their retirement. It is important for financial advisors to have a good understanding of how their clients spend their time in retirement to help inform the planning process. While many studies tend to focus on the quantitative factors that affect retirement, emphasis must also be placed on qualitative factors. Multiple studies in financial planning discuss the different stages of retirement, and how retirees should spend their money. For example, Stein (1998) points that retirees are in

Š2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

the active phase of retirement before age 75 and that advisors should expect them to travel frequently. Based on empirical evidence provided in this study, we find that individuals on average spend five times more on traveling early in retirement compared to late retirement. Interestingly, retirees have a very smooth lifestyle pattern regarding their daily activities. As more baby boomers enter retirement, understanding how they spend their time is critical. Furthermore, with the advancement in medicine, financial planners should expect their clients to live longer. Many clients will look to their financial planner for more than just financial advice, but also for suggestions and tips on how they should spend their time in retirement.

Although many retirees can sustain a consistent quality of life throughout retirement, we see that there is a spike in health care costs late before death. With simulations, we explore the effects of extreme variations in out-of-pocket medical expenditures and its impact on portfolio sustainability. Financial advisors should implement strategies to reduce the risk of portfolio depletion, in particular for those clients who might have a bequest motive.

References Becker, G. S. (1965). “A Theory of the Allocation of Time.” The Economic Journal, 493–517.

47

Feenberg, D., & Skinner, J. (1992). “The risk and duration of catastrophic health care expenditures” (No. w4147). National Bureau of Economic Research.

Finke, M. S., Howe, J. S., & Huston, S. J. (2016). “Old age and the decline in financial literacy.” Management Science.

Finke, M., and Pfau, W., 2015. “Reduce Retirement Costs with Deferred Income Annuities Purchased before Retirement.”

Journal of Financial Planning 28 (7): 40–49.

Goda, G. S., Shoven, J. B., & Slavov, S. N. (2011). “How Well Are Social Security Recipients Protected from Inflation?” In Investigations in the Economics of Aging (pp. 119–139). The University of Chicago Press.

Hurd, M., & Rohwedder, S. (2003). “The retirement-consumption puzzle: Anticipated and actual declines in spending at retirement” (No. w9586). National Bureau of Economic Research.

Kalenkoski, C. M., & Oumtrakool, E. (2014). “How Retirees Spend Their Time: Helping Clients Set Realistic Income Goals.” Journal of Financial Planning, 27(10).

Kitces, M. (2013). “20 Years of Safe Withdrawal Rate Research – A Literature Review and Practical Applications.” The Retirement Management Journal, 4(2), 25–42.

Benartzi, S., Previtero, A., & Thaler, R. H. (2011). “Annuitization Puzzles.” The Journal of Economic Perspectives, 143–164.

Moore, J. F., & Mitchell, O. S. (1997). Projected Retirement Wealth and Savings Adequacy in the Health and Retirement Study (No. w6240). National Bureau of Economic Research.

Browning, C., Guo, T., Cheng, Y., & Finke, M. S. (2016). “Spending in Retirement: Determining the Consumption Gap.” Journal of Financial Planning, 29(2), 42–53.

Pfau, W. D. (2013). “A Broader Framework for Determining an Efficient Frontier for Retirement Income.” Journal of Financial Planning, 26(2).

Clark, R. L., & Mitchell, O. S. (2014). “How does retiree health insurance influence public sector employee saving?” Journal of Health Economics, 38, 109–118.

Stein, M. K. (1998). The Prosperous Retirement: Guide to the New Reality. Emstco Press.

Davison, T., & Turner, K. (2015). “The Reverse Mortgage: A Strategic Lifetime Income Planning Resource.” The Journal of Retirement, 3(2), 61–79.

Tomlinson, J., S. Pfeiffer, & J. Salter. “Reverse Mortgages, Annuities, and Investments: Sorting Out the Options to Generate Sustainable Retirement Income.” Journal of Personal Finance 15.1 (2016).

Drew, M. E., Walk, A. N., & West, J. M. (2016). “Withdrawal capacity in the face of expected and unexpected health and aged-care expenses during retirement.” The Journal of Retirement, 3(3), 77–94.

Wise, D. (2014). “The Economics of Aging.” NBER Reporter, 2014(2), 1.

48

Journal of Personal Finance

The Impact of Product Knowledge and Quality of Care on Long-term Care Insurance Demand: Evidence from the HRS

Jacob Lumby, Texas Tech University, Lubbock, Texas Christopher Browning, PhD, BS Program Co-Director, Assistant Professor of Personal Financial Planning, Texas Tech University, Lubbock, Texas Michael S. Finke, PhD, Dean and Chief Academic Officer, The American College of Financial Services, Bryn Mawr, PA

Abstract Using a unique module in the Health and Retirement Study (HRS), this paper considers three important factors that may influence consumer demand for long-term care insurance (LTCI): preference for high quality care, potential costs, and knowledge. In addition, this paper proposes a new method for examining insurance demand. Only those individuals who are considering purchasing LTCI in the near future (who donâ&#x20AC;&#x2122;t currently own a LTCI policy) are included in the analysis. By focusing on this group, this paper attempts to determine the factors that are most relevant to the LTCI purchase decision when the consumer is most heavily considering it. Our findings imply that consumers deeply care about the provision for high quality long-term care, and suggest that widespread informational deficiencies currently suppress the demand for private long-term care insurance.

Š2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Introduction Long-term care (LTC) related expenditures represent a growing and substantial financial risk for elderly Americans. It is currently estimated that 70% of Americans over the age of 65 will require some type of long-term care support (Freundlich, 2014), and this segment of the population is expected to double by the year 2050 (Houser, Fox-Grage, & Ujvari, 2012). While the average nursing home stay lasts one year or less, there is a sizeable risk exposure for those who require many years of care (Brown & Finkelstein, 2009).

The cost of a private nursing home room averaged more than $92,000 in 2016 (Genworth Financial, 2016). Norton, Wang and Stearns (2006) find that average monthly out-of-pocket (OOP) health care expenditures rise steadily with age, from $85 per month at age 66 to $485 per month at age 95. This increase is almost entirely due to LTC-related expenditures.

Marshall, McGarry and Skinner (2011) find that toward the end of life, OOP LTC expenditures are large and highly variable. On average, wealthy individuals spend an increasing amount on quality nursing care and LTC support. Ameriks, Briggs, Caplin, Shapiro, & Tonetti (2015b) find that retirees hold a significant amount of wealth into old age as a hedge against unknown LTC expenditures.

One way to limit OOP expenditures is through insurance. Unfortunately, for elderly Americans, LTC services are not covered under most existing insurance plans. Private health insurance rarely covers any type of LTC-related expenses. Medicare does not pay for custodial care or assistance with activities of daily living (ADLs), which is the most commonly needed type of LTC support (Medicare, 2016). Medicare will pay for some intermittent skilled nursing care in the recipient’s home, assuming it is part of a physician-certified care plan. Medicare also covers up to 100 days of care at approved skilled nursing facilities, but only after a qualified hospital stay of at least three days.

Even then, only the first 20 days of care are covered by Medicare, with days 21–100 requiring a significant copayment in excess of $160 per day. As a result, the means-tested Medicaid program is the primary payer of LTC expenditures in America (Reaves & Musumeci, 2015). Medicaid eligibility requirements impose strict income and asset limitations. As a result, a large percentage of middle-class Americans have difficulty qualifying without a significant depletion of wealth. To qualify for coverage, these individuals are forced to spend-down their accumulated life-savings until they meet their states’ asset eligibility levels (Brown & Finkelstein, 2011). This constraint levies

49

an implicit tax rate of 100 percent on all savings that exceed the asset threshold, making Medicaid a highly imperfect form of LTC insurance for many Americans. The presence of public care aversion (Ameriks et al., 2015b; Ameriks, Caplin, Laufer, & Nieuwerburgh, 2011) further decreases the attractiveness of Medicaid.

Funding LTC-related expenditures without the assistance of Medicaid leaves individuals with two options. They can self-insure or purchase a LTC insurance (LTCI) policy. The optimal choice is primarily a function of wealth and individual preference for high quality care. The very wealthy have the ability to self-insure without depleting their assets (Cramer & Jensen, 2006), giving them more flexibility to choose whether to retain or transfer risk depending on risk and bequest preferences. It is generally the middle and upper-middle class who should consider purchasing LTC insurance. These individuals typically have several hundred thousand dollars in assets, along with Social Security and/or pension income. They do not have the capacity to self-insure without risking the entirety of their wealth in an adverse health care event. At the same time, they will not qualify for Medicaid coverage without a significant and immediate depletion of wealth.

Long-term care insurance thus provides protection against a significant source of idiosyncratic spending risk in retirement.

By purchasing private LTC insurance, individuals are able to better predict OOP medical expenditures throughout retirement and potentially eliminate the risk of depleting wealth and relying on Medicaid to fund a LTC event (Bajtelsmit & Rappaport, 2014). In the absence of LTCI, retirees must either accept the possibility of an extreme low probability loss or set aside a significant amount of wealth as a hedge against potential future health shocks, which will result in a significant decrease in expected retirement welfare—especially among those who do not have a strong bequest motive.

Despite these compelling reasons to own LTCI, less than 10 percent of the American population is covered by private long-term care insurance (Gottlieb & Mitchell, 2015), and those private policies pay less than 10 percent of total LTC-related expenditures (Reaves & Musumeci, 2015). Previous research has suggested that multiple demand-side factors currently limit the size of the LTCI market. Explanations include Medicaid’s implicit tax (Brown, Coe, & Finkelstein, 2007; Brown & Finkelstein, 2011), unattractive policy terms and pricing (Cramer & Jensen, 2006), competing bequest motives (Lockwood, 2014), intra-family moral hazards (Coe, Goda, & Van Houtven, 2015),

50

Journal of Personal Finance

low risk perception (Zhou-Richter, Browne, & Gründl, 2010), limited consumer knowledge (AARP, 2006) and narrow framing (Gottlieb & Mitchell, 2015). There also exists a rich literature detailing the impact of other behavioral traits on insurance uptake that broadly applies to the long-term care insurance purchase decision (Kunreuther & Pauly, 2006; Kunreuther, Pauly, & McMorrow, 2013).

This paper adds to the existing literature in a number of ways.

Instead of considering only one variable theorized to impact the demand for LTCI, this paper considers three important sets of factors that may influence consumer demand: quality of care, potential costs, and knowledge. In addition, we propose a new method for examining the demand for LTC insurance.

Only individuals who responded to a survey module on longterm care demand (who don’t currently own a LTCI policy) are included in the analysis. By focusing on this group, this paper attempts to determine the factors that are most relevant to the LTCI purchase decision when the consumer is most heavily considering it.

Another key consideration is the importance of health-dependent utility. Long-term care insurance is designed to transfer wealth from a healthy state to a state of illness. Standard lifecycle theory suggests that the marginal utility of consumption will be the same in both states. If, instead, the marginal utility of consumption differs between the two states, demand for the LTCI will be impacted. Brown et al. (2012) find that individual preferences vary with respect to health-dependent utility. In their full sample, respondents were evenly divided between preferring financial resources in healthy or sick states. However, individuals who preferred resources in the sick state were more likely to own LTCI than those who preferred resources when healthy. Hong et al. (2013) find that the marginal utility of consumption expenditure grows when health declines, consistent with the idea that wealth can improve health and well-being at older ages (Marshall et al., 2011). Ameriks et al. (2015a) find that most individuals assign a high valuation to wealth in the LTC-dependent state. They also find that a majority of individuals would like to insure wealth in this state, if offered suitable insurance products.

Adding to these findings, this paper directly tests the relationship between preference for high quality care and willingness The existing literature has shown that cost is an important con- to purchase LTC insurance. The findings in this paper suggest sideration for those considering LTCI (Brown, Goda, & McGarry, that the desire for high quality nursing care is a stronger LTC 2012; Cramer & Jensen, 2006). Other related work has explained purchase motivator than potential premium costs. why Medicaid’s “implicit tax” essentially makes LTC insurance more costly for a large percentage of the wealth distribution The Importance of Knowledge (Brown et al., 2007; Brown & Finkelstein, 2008). But within the LTC market, any discussion of cost should also consider the The second important contribution of this paper is testing the importance of quality. relationship between knowledge and willingness to purchase LTCI. Theory suggests that consumers will be unwilling to A number of previous findings have documented “public care purchase any insurance product if they are not fully informed aversion,” which is the preference for the quality of private about the probability or severity of the underlying risk. For care over government-provided long-term care (Ameriks et al., example, flood and earthquake insurance is rarely purchased 2015b, 2011; Norton, 2005). This aversion to public care should by individuals living in a high-risk area until after a major not be surprising given the ongoing concerns about the quality disaster occurs (Browne, Knoller, & Richter, 2015; Kunreuther & of nursing care in America. Using data from Pennsylvania, Pauly, 2006). Similarly, if individuals underestimate the cost and Hackmann (2015) finds that low Medicaid reimbursement likelihood of needing LTC support, they are unlikely to purchase rates result in understaffed nursing facilities and lower quality insurance. AARP (2006) finds that nearly two-thirds of responnursing care. These problems are not observed in the nursing dents aged 45-plus severely underestimate nursing home facilities that refuse Medicaid as a method of payment. Given costs. In addition, more than half of respondents mistakenly the ongoing Federal and State funding constraints and the believe that Medicaid, Medicare, or Medigap coverage will pay increasing numbers of elderly Americans, it seems unlikely that for nursing home stays and LTC expenses. These findings imply Medicaid will increase reimbursement rates in the future. As that major informational deficiencies likely reduce demand for such, individuals will have to consider private LTC market soluLTC insurance. tions if they want to avoid Medicaid-funded nursing care.

The Importance of Quality

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

51

Zhou-Richter, Browne, and Gründl (2010) study the importance of LTC knowledge in Germany, where adult children are financially responsible for their parents when the parent is unable to pay for LTC expenses. They find that demand for LTC insurance is low due to low risk perception. As children are provided information about the costs and likelihood of needing LTC services, many become willing to purchase insurance coverage for themselves or an adult parent. Twenty-nine percent of respondents without an initial willingness to buy indicated that they were interested in acquiring LTC insurance after being exposed to the information provided in their survey.

respondents after the main survey is completed. Respondents are assigned to a maximum of one module per survey wave and can refuse to participate at any time. In 2012, there were eleven experimental modules administered, with one being specific to long-term care insurance. The LTC module contains the dependent variable and three explanatory variables of interest in this study. This paper is the first to consider these variables in the HRS, allowing us to test the impact of important demand factors on the LTCI purchase decision for older Americans considering LTC coverage.

Adding to the work of Zhou-Richter et al., this paper directly tests the relationship between LTC knowledge and willingness to purchase LTC insurance in older American adults. By evaluating the roles of consumer perception (cost, quality, and value of LTCI) and product knowledge using a special module in the HRS, we can improve our understanding of what motivates the purchase decision for this group. This insight can lead to consumer education, product structuring, defaults, and framing that guide consumers to improved LTCI purchase decisions.

Model

Methods Data This study examines data from the 2012 wave of Health and Retirement Study (HRS), an on-going biannual study conducted by the University of Michigan and funded by the National Institute on Aging. The HRS study is a nationally representative survey of Americans over the age of 50. The initial 1992 HRS survey was conducted face-to-face in the homes of respondents, with a sample size of 12,654 respondents from 7,608 households. The HRS observational unit is an eligible household financial unit including at least one individual born between the years 1931 and 1941 in the contiguous United States.

The study is designed to collect information on individuals nearing or entering retirement, and to observe their transition into retirement. Respondents are asked a number of questions related to current health, cognitive ability, labor market participation, socioeconomic status, financial status, insurance uptake, and retirement planning. In addition to the core survey, several additional questionnaires (called experimental modules) are used to gather the HRS data. These modules are administered to randomly-selected subsamples of HRS

The dependent variable in this study is willingness to purchase LTC insurance. The question used to measure each individual’s willingness to purchase LTC insurance was: “On a scale of 0 to 100, where 0 means absolutely no chance and 100 means absolutely certain, what are the chances that within the next 5 years (10 years if respondent was under age 60) you will purchase insurance that will pay for some or all of nursing home costs?”

Individuals who currently own an LTCI policy are excluded from the study. Individuals currently in a nursing home are also excluded because they are unlikely to qualify for coverage.

After dropping nonqualifying respondents, the final sample consists of 1,396 individuals from the 2012 HRS wave.

The distribution of responses is non-normal (Figure 1) for the dependent variable. Roughly 45 percent of respondents reported a 0 percent probability of purchasing an LTC policy Figure 1: Distribution of Responses for the Dependent Variable (Probability of Purchasing LTCI)

Journal of Personal Finance

52

in the future. In light of this, we specify an ordered probit model with three ordinal categories—no chance of purchase (0 percent probability), low chance of purchase (1–49 percent probability), and high chance of purchase (50–100 percent probability). The ordered probit model estimated in this paper can be written as: yi*= f (β1 Ki + β2 Pi + β3 Qi + β4 Di + β5 Wi + β6 Hi + εi) yi =1 if yi*= 0 yi = 2 if 0 < yi* < 50 yi = 3 if yi* ≥ 50 Where yi* is the true probability of purchasing a LTC insurance policy for each individual in the sample, Ki is each respondent’s self-reported LTC knowledge. Pi is each respondent’s concern about future potential LTCI premium increases (cost). Qi is each respondent’s preference for quality nursing care that is significantly better than the minimum government standard (Medicaid). Di is a vector of demographic characteristics. Wi is a vector of wealth characteristics previously theorized to influence the decision to purchase LTCI. Hi is a vector of health characteristics, included to capture risk factors that might impact the ability to qualify for LTCI. εi is the error term, assumed to be normally distributed.

Explanatory Variables Three explanatory variables of interest are included in the analysis: LTC specific knowledge, preference for high quality care, and concern about the cost of premiums. To assess the importance of knowledge, the following question is used from the HRS: “How would you rate your knowledge of long-term care insurance? Would you say that you have no knowledge at all, a little knowledge, some knowledge, or a lot of knowledge about long-term care insurance?” Due to sample size restrictions, responses have been converted into a dichotomous variable equal to one if the respondent indicated having “some knowledge” or “a lot of knowledge,” and equal to zero if they indicated “no knowledge” or “a little knowledge.” Theory and previous empirical work suggest a positive association between LTC knowledge and willingness to purchase LTCI. Preference for quality nursing care is tested through the following question asked in the HRS: “The insurance policy covers high quality nursing homes that offer services that are substantially above the minimum government standards. Would

you say this is very important, somewhat important or not at all important to you?” The response is coded as one when the respondent indicated that quality is “very important,” and zero otherwise. Consistent with the documented effects of public care aversion and increasing marginal utility in the LTC-dependent state, the model predicts a positive association between preference for high quality care and willingness to purchase LTCI. Concerns about potential costs are captured by the following question: “In thinking about possibly purchasing long-term care insurance, how important would the following concerns be to you? The insurance company may increase your premiums sometime in the future. Would you say this concern is very important, somewhat important or not at all important to you?” The response is coded as one when the respondent indicated that quality is “very important,” and zero otherwise. Theoretically, the association between potential premium increases and willingness to purchase LTCI is ambiguous. On one hand, this is a question about affordability. Those individuals who are extremely concerned about future premium increases might not have sufficient wealth or income to continue paying premiums, making them less likely to purchase LTCI. Such a story is consistent with the findings of Brown et al. (2012). On the other hand, this question might be capturing the salience of LTCI features as presented in the HRS module. Individuals who have absolutely no desire to own LTCI are unlikely to describe possible premium increases as being “very important.” Thus, as interest in the product increases, so too should the relative importance of premium increases, resulting in a positive association with the dependent variable.

Control Variables The specified model also accounts for a number of factors that economic theory and/or previous empirical work suggests are important predictors of LTCI demand. This includes demographic variables such as age, gender, marital status, number of living children, race, and education. The age variable is separated into four categories: under age 55, age 55–65, age 65–75, and over age 75. This is beneficial because the existing literature has shown that willingness to purchase declines past age 65 due to escalating premium costs and difficulties in the medical underwriting process (Cramer & Jensen, 2006). Gender is included as a dummy variable because women might be more interested in LTCI than men, given their longer average life expectancies. Marital status is coded as a dummy variable,

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

and is important because many married couples fear impoverishing their spouse (Pauly, 1990). Informal care provided by a child has been often considered a substitute for LTCI (Courbage & Eeckhoudt, 2012), making the number of living children an important control variable. Race is coded as white, black, and other due to sample size restrictions, with white being the comparison group. Race is included to capture any cultural differences that might influence willingness to purchase LTCI. Education has been shown to be highly influential in financial decision making (Lusardi & Mitchelli, 2007). To control for education, the sample is divided into categories representing various educational achievements. Individuals who failed to finish high school are the comparison group, with additional categories for completing a GED, a high school diploma, some college, or a bachelor’s degree.

In addition, we control for household financial characteristics including income, assets, completion of estate planning documents, and bequest motive. Income has previously been found to be a significant factor in the demand for LTCI. This is easily understood because low-income individuals will have difficulty affording the cost of LTCI premiums (Brown et al., 2012). In our empirical model, we chose to categorize income into three groups: low (less than $50,000/year), medium ($50,000–$100,000/year), and high (greater than $100,000/ year). Wealth is another important determinant of LTCI demand. Without insurance, an extended stay in a nursing facility can destroy all accumulated savings. Assets are also a crucial part of determining Medicaid eligibility and the ability to self-insure. Because housing wealth is largely exempt from Medicaid rules, we consider nonhousing assets only in this study. Building upon previous research and practitioner advice (Cramer & Jensen, 2006), we specify three asset categories: those likely to default to Medicaid coverage (less than $200,000 in assets), those who are most likely to benefit from LTCI ($200,000–$1,500,000), and those who might consider self-insuring (greater than $1,500,000). The HRS data on estate planning is not particularly robust, but there are two variables used in this analysis: a dummy variable indicating the completion of a valid will, and the respondent’s self-assessed probability of leaving a bequest. To model bequest motive, we create a dummy variable equal to one if the respondent indicated at least a 75 percent probability of leaving $100,000 or more as a bequest, and zero otherwise. Lockwood’s (2014) empirical findings suggest that a bequest motive reduces the value of

53

insurance by reducing the opportunity cost of precautionary savings. An increasing number of studies have found the opposite, suggesting that LTCI protects the resources that can be bequeathed (Ameriks et al., 2015b; Brown et al., 2012; Cramer & Jensen, 2006).

Health status has been shown to be an important predictor of LTCI demand, and an important part of the underwriting process. We control for self-reported health status using a dummy variable that is equal to one if the respondent indicated having good health or better, and zero otherwise. We include a dummy variable equal to one if the respondent is currently a smoker, and zero otherwise. Similarly, we create a dummy variable indicating current depression, and another indicating heavy alcohol use. We control for recent nursing home use with a dummy variable that is equal to one if the respondent used any nursing facility in the previous 2 years, and zero otherwise.

Finally, we control for any current difficulties in completing activities of daily living (ADLs) or instrumental activities of daily living (IADLs) using an ordinal scale.

Results Descriptive Statistics We provide descriptive statistics for the entire sample, broken down by the probability of LTCI purchase (Tables 1 and 2). More than 70 percent of respondents are white, with an average age of 65. Two-thirds are married, with an average of three living children. About 43 percent of respondents are male, perhaps indicative of lower average life expectancies when compared to females. Nearly one-quarter of respondents have a bachelor’s degree or higher, with a similar percentage completing high school and some college. The average total household income is nearly $67,000 per year, but about 60 percent of respondents earn less than $50,000 per year. The average total nonhousing wealth is slightly greater than $250,000, with nearly 75 percent of households having less than $200,000. Twenty-two percent of respondents fall within the middle asset range, with roughly 4 percent of respondents having more than $1.5 million in nonhousing wealth. About a third indicate a high probability of leaving a $100,000 or larger bequest, with nearly half completing a valid will. About 75 percent indicated being in good health or better, and only 2 percent have used a nursing facility in the past 2 years.

54

Journal of Personal Finance

Table 1: Descriptive Statistics for Entire Sample (n=1396) Explanatory Variables of Interest Mean “Some” or “A Lot” of LTC Knowledge 0.26 Potential Premium Increase is “Very Important” 0.59 High Quality Care is “Very Important” 0.48 Demographic Control Variables Mean Average Age 65.02 Under Age 55 0.15 Age 55–65 0.39 Age 65–75 0.23 Over Age 75 0.23 Male 0.43 Married 0.67 Average # of Living Children 2.98 Race White 0.73 Black 0.20 Other 0.07 Education 12.86 Less Than High School 0.17 GED 0.06 High School Graduate 0.27 Some College 0.27 Bachelor’s or Higher 0.23 Wealth Control Variables Mean Total Household Income $66,730 Income < $50,000 0.59 Income $50,000–$100,000 0.22 Income > $100,000 0.19 Total NonHousing Wealth $258,540 Assets < $200,000 0.74 Assets $200,000–$1,500,000 0.22 Assets > $1,500,000 0.04 Bequest Motive Likely to Leave $10,000 bequest 0.56 Likely to Leave $100,000 bequest 0.33 Likely to Leave $500,000 bequest 0.09 Has Valid Will 0.44 Health Control Variables Mean Good Health or Better 0.75 Current Smoker 0.16 Depressed in Last Year 0.15 Drinking Problem 0.09 Average # of IADLs 0.09 Average # of ADLs 0.26 Nursing Home Stay in Last 2 Years 0.02

When comparing the sample by the probability of LTCI purchase, about one-quarter of respondents (n=359) in the full sample (n=1,396) indicated a high probability of purchasing LTC insurance in the future (50 percent probability or greater). A

similar number (n=409) reported a low probability of purchase (1–49 percent), with the remainder (n=629) indicating no willingness to purchase LTC insurance.

The probability of purchasing LTCI increases monotonically with all three variables of interest in this study. Only 22 percent of respondents who indicated no probability of purchase had “some” or “a lot” of LTC knowledge. That number increases by 50 percent within the high probability group, where 33 percent of respondents indicate having knowledge. Individuals who describe potential premium increases as being “very important” are more likely to purchase LTCI in the future, as are those individuals who describe high quality nursing care as being “very important.”

Interestingly, marriage did have some correlation with willingness to purchase. About 62 percent of individuals who indicated no chance of purchase were married, while almost 75 percent of individuals who indicated a high probability of purchase were married. This suggests that individuals might view LTC insurance as more useful when a spouse is present.

Individuals who indicated a 0 percent probability of purchasing a policy were, on average, much less educated than individuals who indicated any probability of purchase (1–100 percent). This effect is strong and distinct from LTC-specific knowledge. That pattern holds true for income and assets as well. Individuals who indicated a zero probability of purchase have much lower annual incomes and fewer nonhousing assets, on average.

Multivariate Analysis The main predication of the model is that, after controlling for other variables, LTC knowledge and preference for high quality care will significantly increase the probability of purchasing LTC insurance. Because straightforward interpretation of coefficients is difficult in the original ordered probit regression, marginal effects are calculated and shown in Table 3. For ease of viewing, statistically significant results are displayed.

All three variables of interest are positively and significantly associated with the probability of purchasing LTCI. The statistically significant coefficients (p-values less than 0.01) imply a large economic effect that persists after controlling for a number of variables. Respondents who indicate that high quality care is very important are 7 percent more likely to indicate a high willingness to purchase LTCI, and 8.5 percent less likely to indicate no willingness to purchase when compared to the baseline group. The magnitude of the calculated marginal

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

55

Table 2: Descriptive Statistics by Probability of LTCI Purchase (n=1396) 0% Probability (n=629)

1-49% Probability (n=409)

50–100% Probability (n=359)

Explanatory Variables of Interest

Mean

Mean

Mean

"Some" or "A Lot" of LTC Knowledge

22%

25%

33%

Potential Premium Increase is "Very Important"

52%

64%

68%

High Quality Care is "Very Important"

40%

47%

62%

67

64

62

Under Age 55

12%

16%

20%

Age 55–65

31%

41%

49%

Age 65–75

24%

27%

19%

Over Age 75

Demographic Control Variables Average Age

33%

16%

12%

Percent Male

42%

46%

40%

Percent Married

62%

69%

74%

Number of Children

3.15

2.80

2.87

Average Educational Years

12.2

13.4

13.3

Less Than High School

22%

12%

13%

GED

6%

5%

7%

High School Graduate

31%

24%

25%

Some College

25%

29%

27%

Bachelor's or Higher

16%

30%

28%

White

74%

77%

66%

Black

18%

16%

27%

Other

7%

7%

7%

$47,539

$84,661

$79,927

69%

55%

48%

Wealth Control Variables Total Income Income < $50,000 Income $50,000–$100,000

19%

22%

27%

Income > $100,000

12%

23%

25%

$210,913

$356,744

$230,105

Assets < $200,000

79%

68%

73%

Assets $200,000–$1,500,000

18%

27%

25%

Assets > $1,500,000

3%

5%

2%

75% Chance of Leaving $100,000+ Bequest

26%

38%

38%

Has Valid Will

47%

47%

37%

In Good Health or Better

66%

82%

82%

Currently Smoking

20%

12%

14%

Heavy Alcohol Use

9%

9%

10%

Average # of IADLs

0.14

0.07

0.06

Total NonHousing Assets

Health Control Variables

Average # of ADLs

0.39

0.16

0.19

Depressed in the Last Year

17%

11%

14%

Used Nursing Home in Previous 2 Years

3%

1%

1%

Journal of Personal Finance

56

Table 3: Marginal Effects of the Ordered Probit Regression on the Probability of Purchasing LTCI (n=1396) Ordered categories include: 0% Probability, 1–49% Probability, 50–100% Probability (Standard errors are included under the marginal effects for each variable.)

Explanatory Variables of Interest “Some” or “A Lot” of LTC Knowledge Potential Premium Increase is “Very Important” High Quality Care is “Very Important”

0% Probability –0.0753 *** (0.0268) –0.0656 *** (0.0254) –0.0852 *** (0.0246)

1–49% Probability 0.0137 *** (0.0051) 0.0119 ** (0.0048) 0.0155 *** (0.0047)

50–100% Probability 0.0616 *** (0.022) 0.0537 *** (0.0208) 0.0697 *** (0.0203)

Demographic Control Variables Under Age 55 (Baseline Comparison Group) Age 55–65 Age 65–75 Over Age 75 Less than High School (Baseline Comparison Group) Bachelor’s or Higher White (Baseline Comparison Group) Black

0.0023 (0.0332) 0.08 ** (0.0395) 0.1945 *** (0.0457)

–0.0002 (0.003) –0.0128 ** (0.0063) –0.0483 *** (0.0132)

–0.0021 (0.0302) –0.0672 ** (0.0339) –0.1462 *** (0.0347)

–0.0907 ** (0.0424)

0.0155 * (0.0086)

0.0752 ** (0.0343)

–0.0892 *** (0.0314)

0.0121 *** (0.0034)

0.0771 *** (0.0286)

–0.0468 (0.0317) –0.0636 * (0.0367)

0.0086 (0.0056) 0.0108 * (0.0057)

0.0382 (0.0263) 0.0528 * (0.0313)

–0.0245 (0.0303) 0.1099 * (0.0592) –0.0495 * (0.0277) 0.0458 * (0.0269)

0.004 (0.0047) –0.0293 (0.0197) 0.009 * (0.0052) –0.0083 * (0.0049)

0.0204 (0.0256) –0.0806 ** (0.0397) 0.0405 * (0.0227) –0.0375 * (0.0221)

–0.0794 ** (0.0324) 0.0787 ** (0.0346)

0.0144 ** (0.0061) –0.0143 ** (0.0065)

0.065 ** (0.0266) –0.0644 ** (0.0283)

Financial Control Variables Income < $50,000 (Baseline Comparison Group) Income $50,000–$100,000 Income > $100,000 Assets < $200,000 (Baseline Comparison Group) Assets $200,000–$1,500,000 Assets > $1,500,000 75% Chance of Leaving $100,000+ Bequest Valid Will Health Control Variables In Good Health or Better Currently Smoking Pseudo R-Squared = 0.0916 *** Indicates significance at the 1% level ** Indicates significance at the 5% level * Indicates significance at the 10% level

effects is larger than the other explanatory variables of interest, implying that individuals deeply care about the quality of care offered in any LTC setting. Respondents who have at least some

LTC knowledge are 6.2 percent more likely to indicate a high willingness to purchase, and 7.5 percent less likely to indicate no willingness to purchase when compared to the baseline

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

group. This finding is separate from the effects of general education, which is controlled for elsewhere. Consistent with Zhou-Richter et al. (2010), this finding highlights the importance of product knowledge in the LTC insurance purchase decision. Respondents who indicate that a potential premium increase is very important are 5.4 percent more likely to indicate a high willingness to purchase LTCI, and 6.5 percent less likely to indicate no willingness to purchase. As described earlier, we believe this finding highlights the underlying differences among the three categorical-dependent variables.

Individuals with no interest in LTCI are unlikely to care about future premium increases. On the other hand, individuals who are highly interested in purchasing LTCI might care a great deal about future premium increases.

Other statistically significant findings are in line with the existing literature on LTC insurance. Willingness to purchase LTCI decreases with age. There is no statistical difference between respondents age 55–65 and respondents younger than 55, but individuals over the age of 65 (and 75) are much less willing to purchase LTCI. Most of the educational categories are insignificant with the exception of those who have a bachelor’s degree or higher. These individuals are 7.5 percent more likely to indicate a high willingness to purchase LTCI, and 9 percent less likely to indicate no willingness to purchase. This is likely due to a number of factors including being more knowledgeable about health concerns and more future oriented (having a low discount rate). Those with more formal education might also have better financial literacy and numeracy skills, both of which have been shown to be important factors in stimulating the demand for insurance (Lusardi & Mitchelli, 2007; McGarry et al, 2016).

There are numerous interesting findings within the financial control variables. Having annual income in excess of $100,000 per year is positively associated with willingness to purchase LTC insurance, as compared to the baseline income group (those earning less than $50,000 per year).

Because LTCI premiums can be expensive, individuals need to have discretionary income to afford the monthly payments.

Individuals without significant income will likely struggle to afford the monthly premiums, reducing their willingness to purchase LTCI. Perhaps most interesting is the relationship between nonhousing assets and willingness to purchase LTCI. The “target audience” for LTCI is the group with modest assets ($200,000–$1,500,000), but this group does not appear any more likely to purchase LTCI than the baseline group with

57

less than $200,000 in assets. However, our findings suggest that high net worth individuals are often rational consumers of LTCI. Being in the highest asset category is associated with a strong and significant decrease in the probability of purchasing LTCI. This might suggest that this group of individuals has decided to self-insure against potential LTC expenditures. Having a high probability of leaving a $100,000 bequest is positively associated with willingness to purchase LTCI. This finding is in line with several other recent findings (Brown et al., 2012; Cramer & Jensen, 2006), and suggests a complimentary relationship between LTC insurance and bequest motive. Having a valid will is negatively associated with willingness to purchase LTCI. One possible explanation is the difference in average age across the dependent variable cohorts. The average age of the group that is completely unwilling to purchase LTCI is 67. The average age of the high probability group is 62.

Of the various health control variables, only current health status and smoking status are significant. Being in good health or better is associated with a higher probability of purchasing LTCI. The significant and positive association suggests that healthy individuals may expect to live longer and eventually need some type of LTC support. Being a current smoker is associated with a decreased probability of purchasing LTCI, which is likely a combination of health concerns (likely resulting in higher premiums) and a high discount rate (being present oriented).

One unexpected finding appears in the multivariate analysis.

Black individuals were much more likely to indicate a high probability of purchase when compared to whites and other races. In our sample, black individuals are 7.7 percent more likely to indicate a high willingness to purchase LTCI, and 8.9 percent less likely to indicate no willingness to purchase when compared to whites. We’ve seen very little in the existing literature to explain this finding, and future research might explore how the long-term care purchase decision varies by race.

Sensitivity Analysis Due to the nonnormal distribution of responses in the dependent variable, we estimate several additional multivariate models (Probit and OLS). The simple probit model is specified as equal to one if the respondent indicates a high probability of purchase (50–100 percent likely), and equal to zero if the

Journal of Personal Finance

58

respondent indicates no or low probability of purchase (0–49 percent likely). To obtain the estimated effects of each explanatory variable on the observed probability of purchasing private LTCI, marginal effects are calculated and displayed in Table 4 below.

The findings are consistent with the ordered probit model. The biggest change is that the premium variable becomes statistically insignificant1, which is likely caused by collapsing the 1.

Alternatively, the simple probit could be specified as 0=0% probability of purchase, and 1=1–100% probability of purchase. This change results in the premium, health, and smoking variables becoming highly significant, with many wealth variables becoming insignificant.

Table 4: Marginal Effects of the Probit Regression on the Probability of Purchasing LTCI (n=1396) Explanatory Variables of Interest “Some” or “A Lot” of LTC Knowledge Potential Premium Increase is “Very Important” High Quality Care is “Very Important”

Coefficient 0.085 *** 0.022 0.104 ***

Std. Err. 0.026 0.026 0.025

Demographic Control Variables Under Age 55 Age 55–65 Age 65–75 Over Age 75 Male Married Number of Children Less Than High School GED High School Graduate Some College Bachelor’s or Higher White Black Other

0.010 –0.067 * –0.098 ** –0.037 0.041 –0.005

0.034 0.040 0.043 0.024 0.028 0.006

0.032 0.019 –0.006 0.037

0.054 0.037 0.037 0.041

0.120 *** –0.009

0.033 0.041

0.071 ** 0.068 *

0.032 0.038

0.011 –0.126 ** 0.039 –0.055 **

0.032 0.049 0.028 0.027

Wealth Control Variables Income < $50,000 Income $50,000–$100,000 Income > $100,000 Assets < $200,000 Assets $200,000–$1,500,000 Assets > $1,500,000 75% Chance of Leaving $100,000+ Bequest Valid Will Health Control Variables In Good Health or Better Currently Smoking Heavy Alcohol Use Average # of IADLs Average # of ADLs Depressed in the Last Year Used Nursing Home in Previous 2 Years Pseudo R-Squared = 0.096 *** Indicates significance at the 1% level ** Indicates significance at the 5% level * Indicates significance at the 10% level

0.046 –0.044 0.022 –0.028 –-0.002 0.016 –0.033

©2017, IARFC. All rights of reproduction in any form reserved.

0.032 0.033 0.039 0.036 0.018 0.034 0.094

Volume 16, Issue 2

dependent variable and losing some of the information available in the ordered probit. High quality care remains highly significant, as does LTC knowledge. Health and current smoking status become insignificant, but retain the same expected sign.

All of the wealth control variables retain the same sign and interpretation, and some become more statistically significant when compared to the ordered probit. Overall, interpretation remains consistent with both the simple and ordered probit models.

59

Table 5: Ordinary Least Squares Regression of the Probability of Purchasing LTCI (n=1396) Independent Variables of Interest

Coefficent

Std. Err.

“Some” or “A Lot” of LTC Knowledge

5.769 ***

1.746

Potential Premium Increase is “Very Important”

1.572

1.652

High Quality Care is “Very Important”

7.396 ***

1.641

Demographic Control Variables Under Age 55

Some readers might object to collapsing the dependent variable as shown in the probit and ordered probit models. To consider this possibility, the dependent variable can be treated as continuous and ordinary least squares (OLS) regression can be considered. OLS results are presented in Table 5, and the overall findings are consistent with previously discussed models.

Discussion The current study investigates the factors related to the willingness to purchase long-term care insurance (LTCI) among elderly American adults. Individuals who are considering purchasing LTCI in the near future (and who don’t currently own an LTCI policy) are included in the analysis. By excluding current insurance owners, this paper attempts to determine the traits and characteristics that make some consumers more likely to purchase LTCI than others.

We find evidence that individuals deeply care about the quality of care covered by LTCI policies. Describing high quality care coverage as being “very important” is associated with a significantly higher probability of purchasing LTC insurance at some point in the future. This finding is related to the growing literature on public care aversion that is observed in many elderly adults (Ameriks et al., 2015b, 2011; Norton, 2005), and suggests that perhaps the most important feature of LTCI is the provision for high quality care and support.

In addition, we find evidence supporting the importance of long-term care knowledge. Having “some” or “a lot” of LTC specific knowledge is associated with a significantly higher probability of purchasing LTC insurance at some point in the future. Consistent with several other research findings (AARP, 2006; Zhou-Richter et al., 2010), this finding suggests that few individuals are currently well-informed about the risks and costs associated with LTC expenditures, and these information deficiencies greatly suppress the demand for LTC insurance.

Age 55–65

–0.987

2.303

Age 65–75

–6.181 **

2.618

Over Age 75 Male Married Number of Children

–11.226 ***

2.731

–1.282

1.471

1.884

1.672

–0.106

0.391

1.359

3.456

Less Than High School GED High School Graduate

–0.386

2.309

Some College

–1.593

2.393

3.070

2.640

7.505 ***

2.177

Bachelor’s or Higher White Black Other

–1.387

2.598

Income $50,000–$100,000

3.063

2.070

Income > $100,000

4.736 *

2.437

0.563

1.904

Wealth Control Variables Income < $50,000

Assets < $200,000 Assets $200,000–$1,500,000 Assets > $1,500,000 75% Chance of Leaving $100,000+ Bequest Valid Will

–7.360 * 3.550 **

4.020 1.790

–3.222 *

1.677

2.681

1.864

–3.494

2.132

Health Control Variables In Good Health or Better Currently Smoking Heavy Alcohol Use

0.771

2.559

Average # of IADLs

–1.441

1.995

Average # of ADLs

–0.958

0.996

Depressed in the Last Year

1.412

2.208

Used Nursing Home in Previous 2 Years

–1.719

3.653

Constant

15.296 ***

3.660

R-Squared = 0.1331 *** Indicates significance at the 1% level ** Indicates significance at the 5% level * Indicates significance at the 10% level

60

Journal of Personal Finance

Individuals who are better informed display a much greater willingness to purchase coverage.

Obviously, intending to purchase LTC insurance and buying LTC insurance are two different things and a limitation of this analysis. However, these data reflect the receptiveness of older American adults to purchase LTCI, which may be of greatest importance to those interested in private sector or public policy solutions that can grow the LTC insurance market.

Our findings can help guide financial advisors who are navigating potential LTC solutions with clients. The biggest take away is that many individuals value high quality long-term care. Advisors need to help educate clients about the tradeoff between cost and quality. If clients have substantial assets, they are unlikely to qualify for Medicaid and, perhaps more importantly, they are probably unwilling to spend time in a low-cost nursing facility. Our results also highlight the trade-off between wealth and willingness to purchase LTCI. High net worth clients might choose to forego LTCI coverage and self-insure. If, however, these clients also have a bequest motive, LTCI can serve as a complimentary solution, helping to secure the bequest and maintain a reasonable level of consumption throughout retirement. Educating clients about these trade-offs and discussing possible LTC solutions before the need arises will lead to better informed LTCI decisions and stronger advisor-client relationships.

In addition, our results have important implications for public policy. As the American population continues to age, the need for LTC services will grow. Medicaid currently pays for most LTC expenditures in the U.S.—a trend that appears unsustainable given the projected Federal deficit. If policymakers want to shift Medicaid’s future LTC responsibilities to the private sector, our results provide one very simple solution: better educate Americans about the risks, costs, and needs associated with long-term care. Part of that educational material could explain differences in the quality of LTC support, differentiating between Medicaid and other, more expensive facilities. Additionally, any public policies designed to encourage private LTCI purchase should consider the savings associated with reduced Medicaid enrollment and increased personal savings.

References AARP. (2006). The Costs of Long-Term Care: Public Perceptions Versus Reality. Retrieved from http://assets.aarp.org/rgcenter/ health/ltc_costs_2006.pdf Ameriks, J., Briggs, J., Caplin, A., Shapiro, M. D., & Tonetti, C. (2015a). Late-in-Life Risks and the Under-Insurance Puzzle (NBER Working Paper No. 22726). Cambridge, MA. Ameriks, J., Briggs, J., Caplin, A., Shapiro, M. D., & Tonetti, C. (2015b). Long-Term Care Utility and Late in Life Saving (NBER Working Paper No. w20973). Cambridge, MA. Ameriks, J., Caplin, A., Laufer, S., & Nieuwerburgh, S. Van. (2011).

The Joy of Giving or Assisted Living? Using Strategic Surveys to Separate Public Care Aversion from Bequest Motives. Journal of Finance, 66(2), 519–561.

Bajtelsmit, V., & Rappaport, A. (2014). The Impact of Long-Term Care Costs on Retirement Wealth Needs (Society of Actuaries).

Brown, J. R., Coe, N. B., & Finkelstein, A. (2007). Medicaid Crowd-Out of Private Long-Term Care Insurance Demand: Evidence from the Health and Retirement Survey. Tax Policy and the Economy, 21, 1–34.

Brown, J. R., & Finkelstein, A. (2008). The Interaction of Public and Private Insurance: Medicaid and the Long-Term Care Insurance Market. American Economic Review, 98(3), 1083–1102.

Brown, J. R., & Finkelstein, A. (2009). The Private Market for Long-Term Care Insurance in the U.S.: A Review of the Evidence.

The Journal of Risk and Insurance, 76(1), 5–29.

Brown, J. R., & Finkelstein, A. (2011). Insuring Long-Term Care in the United States. Journal of Economic Perspectives, 25(4), 119–142.

Brown, J. R., Goda, G. S., & McGarry, K. (2012). Long-Term Care Insurance Demand Limited by Beliefs About Needs, Concerns about Insurers, and Care Available from Family. Health Affairs (Project Hope), 31(6), 1294–302.

Browne, M. J., Knoller, C., & Richter, A. (2015). Behavioral Bias and the Demand for Bicycle and Flood Insurance. Journal of Risk and Uncertainty, 50(2), 141–160.

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

61

Coe, N. B., Goda, G. S., & Van Houtven, C. H. (2015). Family Spillovers of Long-Term Care Insurance (No. NBER Working Paper No. 21483). Cambridge, MA.

Lockwood, L. M. (2014). Incidental Bequests: Bequest Motives and the Choice to Self-Insure Late-Life Risks (NBER Working Paper No. 20745). Cambridge, MA.

Courbage, C., & Eeckhoudt, L. (2012). On Insuring and Caring for Parents’ Long-Term Care Needs. Journal of Health Economics, 31(6), 842–50.

Lusardi, A., & Mitchelli, O. (2007). Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education. Business Economics.

Cramer, A. T., & Jensen, G. A. (2006). Why Don’t People Buy Long-Term-Care Insurance? Journal of Gerontology, 61(4), 185–193.

Marshall, S., McGarry, K. M., & Skinner, J. S. (2011). The Risk of Out-of-Pocket Health Care Expenditure at End of Life. Explorations in the Economics of Aging, 101–128.

Freundlich, N. (2014). Long-Term Care: What Are the Issues?

Retrieved from http://www.rwjf.org/content/dam/farm/ reports/issue_briefs/2014/rwjf410654

McGarry, B., Temkin-Greener, H., Chapman, B., Grabowski, D., & Li, Y. (2016). The Impact of Consumer Numeracy on the Purchase of Long-Term Care Insurance. Health Services Research.

Genworth Financial. (2016). The Cost of Long Term Care in 2016. Retrieved October 19, 2016, from https://www.genworth.com/about-us/industry-expertise/cost-of-care.html

Medicare. (2016). What is Long-Term Care? Retrieved October 28, 2016, from https://www.medicare.gov/coverage/long-termcare.html

Gottlieb, D., & Mitchell, O. S. (2015). Narrow Framing and LongTerm Care Insurance (NBER Working Paper No. 21048). Cambridge, MA.

Norton, E. (2005). Elderly Assets, Medicaid Policy, and Spend Down in Nursing Homes. Review of Income and Wealth, 41(3), 309–329.

Hackmann, M. (2015). Incentivizing Better Quality of Care: The Role of Medicaid and Competition in the Nursing Home Industry (Working Paper). Retrieved from https://martinhackmann.files. wordpress.com/2015/10/nh_comp_reimb_latest.pdf

Norton, E., Wang, H., & Stearns, S. (2006). Behavioral Implications of Out-of-Pocket Health Care Expenditures. Swiss Journal of Economics and Statistics, 142(S (special issue)), 3–11.

Hong, J., Pijoan-Mas, J., & Rios-Rull, J. (2013). Health Heterogeneity and Preferences for Consumption Growth (Working Paper). Retrieved from http://econpapers.repec.org/paper/ redsed013/1.htm Houser, A., Fox-Grage, W., & Ujvari, K. (2012). Across the states: Profiles of Long-Term Services and Supports (AARP Public Policy Institute). Retrieved from http://www.aarp.org/content/dam/ aarp/research/public_policy_institute/ltc/2012/across-thestates-2012-full-report-AARP-ppi-ltc.pdf Kunreuther, H., & Pauly, M. (2006). Insurance Decision-making and Market Behavior. Now Publishers Inc. Kunreuther, H., Pauly, M., & McMorrow, S. (2013). Insurance and Behavioral Economics: Improving Decisions in the Most Misunderstood Industry. Cambridge University Press.

Pauly, M. (1990). The Rational Nonpurchase of Long-Term-Care Insurance. Journal of Political Economy, 98(1), 153–168.

Reaves, E., & Musumeci, M. (2015). Medicaid and LongTerm Services and Supports: A Primer. Washington D.C. Retrieved from http://kff.org/medicaid/report/ medicaid-and-long-term-services-and-supports-a-primer/ Zhou-Richter, T., Browne, M. J., & Gründl, H. (2010). Don’t They Care? Or, Are They Just Unaware? Risk Perception and the Demand for Long-Term Care Insurance. Journal of Risk and Insurance, 77(4), 715–747.

62

Journal of Personal Finance

Why A QLAC in an IRA Is a Terrible Way to Defer the Required Minimum Distribution (RMD) Obligation

Michael E. Kitces, MSFS, MTAX, CFP®, CLU, ChFC, RHU, REBC, CASL, Partner and the Director of Wealth Management for Pinnacle Advisory Group, Columbia, Maryland

Abstract The longevity annuity has become increasingly popular in recent years as a potential new vehicle for retirement income, as its ability to delay payments to an advanced age like 85 allows for a significant accumulation of mortality credits. And since the introduction of Treasury Regulations in 2014, a so-called “qualified longevity annuity contract” (QLAC) can even be purchased inside of an IRA or other retirement account, allowing a portion of a retiree’s RMDs to be deferred from 70½ to as late as age 85. However, as it turns out the unique nature of a longevity annuity’s payment structure is not very hospitable as an RMD deferral strategy.

The fact that it can take until a retiree’s late 80s just to break even and recover principal means the retiree risks significant forgone growth by trying to merely defer RMDs through the use of a QLAC. And of course, the RMDs will still eventually happen anyway, as the QLAC merely defers when payments begin. In fact, ironically, if the retiree does live, the accelerated payments of a QLAC in the later years can actually deplete an IRA even faster than normal IRA RMDs. Ultimately, this doesn’t mean that the longevity annuity (or a QLAC inside an IRA) is a bad deal. The ability to accumulate mortality credits still means it can be very effective as a fixed income alternative for those who fear they may not have enough money to fund a retirement well beyond their life expectancy.

And if a retiree intends to spend all of his/her assets anyway, and the only available dollars for retirement are held in an IRA or other retirement account, the QLAC is an effective means to engage in such a strategy. Nonetheless, the bottom line is that while a QLAC may be a valid way to use a retirement account to hedge against longevity—and defer RMDs along the way—it’s still not very effective as an RMD avoidance or deferral strategy. Just because you can buy a longevity annuity inside a retirement account as a QLAC doesn’t mean you should.

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

Introduction The basic concept of a longevity annuity is that, like an immediate annuity, a lump-sum payment is made in exchange for guaranteed payments (typically for life) in the future. The difference is that while the lifetime payments from an immediate annuity start immediately—as the name implies—with a longevity annuity the onset of those payments is deferred until some point in the future (and thus is also known as a “deferred income annuity” [DIA]).

For instance, a 65-year-old couple today could put $100,000 into a single premium immediate annuity and get (level) payments of almost $6,000/year for life, but if the couple was willing to wait until age 85 to get the first payment, the subsequent payments would be nearly $32,000/year for life instead.

The good news of this approach is that the payments in the later years are dramatically larger with a longevity annuity than with an immediate annuity. The bad news, of course, is that you have to wait 20 years to get the first check.

Even when adjusting for the waiting period and the time value of money, though, the reality is that in the long run (for those who actually do live a long time), a longevity annuity’s payments provide a better internal rate of return (IRR) than an immediate annuity, as shown in Figure 1. In other words, while the longevity annuity buyer is at risk for a greater loss in the early years (since it may take 20 years just to get the

Figure 1

63

first payment), those greater payments pay off in the long run with a superior implied return (for those who actually live long enough to see them).

Longevity Annuities, the Required Minimum Distribution (RMD) Rules, and the QLAC While the trade-off of a longevity annuity may be appealing for retirees, the situation gets more complicated if the only dollars available to purchase such an annuity are inside retirement accounts, like an IRA or a 401(k) plan. The problem is that by its nature, a longevity annuity doesn’t begin payments until a distant point in the future—such as at the advanced age of 85—and this presents a serious conflict when the standard rules for retirement accounts require that minimum distributions begin beyond age 70½. In other words, it’s hard to start taking RMDs at 70½ when the longevity annuity payments aren’t even scheduled to begin until nearly 15 years beyond that point.

Technically, it would be possible to still calculate a present value of the expected future longevity annuity payments, in order to calculate what an RMD would be—and then take that RMD from other available retirement account assets, since the IRA aggregation rule does allow RMDs to be taken from any account. But given that longevity annuities typically have

Journal of Personal Finance

64

no liquidity until the payments begin, if too much is invested into the contract, and/or the other retirement account assets are spent down too quickly, the retiree could still be stuck in a liquidity squeeze where there are no retirement dollars available outside the longevity annuity to cover the next RMD when it is due.

To resolve the issue, in 2014 the Treasury issued new regulations under 1.401(a)(9)-6, declaring that a “qualified longevity annuity contract” (or QLAC for short) could be owned inside of a retirement account, and automatically have its payments (which still might not begin until after age 70½) be deemed to satisfy the RMD rules. In order to be a “qualified” longevity annuity eligible to be held inside a retirement account, though, the new rules required that •

Only 25 percent of retirement accounts can be invested into a QLAC.

•

The cumulative dollar amount invested into QLACs cannot exceed $125,000.

•

The QLAC still cannot defer payments beyond age 85 (i.e., age 85 is the latest possible start date).

•

The QLAC cannot have a liquid cash surrender value (i.e., it must be irrevocable and illiquid, although it can still have a return-of-premium death benefit payable to heirs).

The establishment of the QLAC rules relieves the need for a retiree to keep other retirement dollars available and liquid to meet any RMD requirements associated with the longevity annuity. Instead, the longevity annuity payments themselves— even if not beginning until age 85—are automatically deemed to satisfy the retiree’s RMD obligations for the funds in the longevity annuity (so only the other remaining retirement accounts must deal with RMDs, which would come from those accounts). And to limit the potential tax advantages—since the QLAC rules essentially allow someone to defer RMDs from age 70½ until age 85—the Treasury limited both the maximum dollar amount and the percentage of retirement accounts that can be allocated from such retirement accounts into a QLAC. On the one hand, the issuance of the QLAC regulations meant that retirees who wanted to own a longevity annuity but only had retirement account dollars available now had a means to do so (without worrying about liquidity issues for satisfying RMDs).

On the other hand, the QLAC regulations also introduce the potential of using a QLAC specifically as an RMD avoidance (or at

least, deferral) strategy. But is a QLAC actually a good way to delay the onset of RMDs?

The Problem with Using a QLAC to Avoid RMD Obligations While it may appear intuitively appealing to use a QLAC to defer RMDs from age 70½ out as late as age 85, there is an important caveat to consider; delaying payments until age 85 also means that the retiree doesn’t get the money back until that point, either. And in fact, even when the payments begin, it takes several years for the payments just to add up to the original principal. Which means the retiree may be deferring RMDs along the way, but in a pure economic sense, must also live until his/her late 80s just to break even and recover the original principal.

Unfortunately, the problem is that’s not even an odds-on bet. For instance, Figure 2 shows the survival rates (using the Health Annuitant RP-2014 Mortality Tables from the Society of Actuaries) for a 69-year-old single male or single female (i.e., as QLAC purchases from a single person’s IRA will typically be single life, and the buyer would ostensibly make the purchase at age 69 to avoid the onset of RMDs at age 70½).

Given that the longevity annuity payment beginning at age 85 would be $36,920/year or $32,606 (for males and females, respectively), such that the buyer must survive to roughly age 88 merely to break even, these results reveal that there’s a barely 50 percent chance that the 69-year-old QLAC buyer even lives long enough to recover his/her principal. In other words, the retiree has to live to or beyond life expectancy just to get enough QLAC payments to get his/her principal back.

And of course, the reality is that even if a QLAC buyer does live long enough to recover principal, that’s still a significant “loss” in opportunity cost; at an 8 percent growth rate, an investment would quadruple in value over a comparable roughly-18-year time horizon. So merely getting the principal back after 18 years is not exactly an accomplishment, even if it did allow for RMDs to be deferred along the way.

In fact, Figure 3 shows the value of a $100,000 QLAC purchase (assuming it has a return-of-principal death benefit guarantee, so its “minimum” value will always be the original starting amount, less taxes), versus the value of a $100,000

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

65

Figure 2

Figure 3

retirement account (e.g., a traditional IRA) that simply stays invested at 8 percent, takes the RMDs as compelled to do so, and reinvests the proceeds in a taxable account. The RMDs are assumed to be taxed at 25 percent (and in order to evaluate comparable â&#x20AC;&#x153;after-taxâ&#x20AC;? spendable values, any remaining pre-tax value of the QLAC or IRA is also haircut at 25 percent to ensure

an apples-to-apples after-tax-wealth comparison). Growth on the funds already pushed into the taxable account (from RMDs that have occurred, or QLAC payments once they have occurred) are assumed to be taxed at 20 percent annually, a combination of taxing ordinary income interest and preferential long-term capital gains and qualified dividends.

66

Journal of Personal Finance

As Figure 3 reveals, using a QLAC to avoid RMDs is indeed a losing proposition. While the RMD payments may be delayed, the reality is that the QLAC still forces money out of the IRA as well, and in the later years does so even faster than RMDs would have. And in the meantime, the IRR on a QLAC is still “only” about 6 percent to age 100, while a balanced account can potentially grow closer to 8 percent, which means the QLAC just lags further and further behind in the initial years. Accordingly, as the chart shows, the retiree must live to age 105 for the QLAC to generate as much wealth as simply keeping the IRA, taking the RMDs, paying the taxes, and reinvesting the proceeds.

On the other hand, for the retiree who “merely” lives to life expectancy, the loss of using a QLAC to avoid RMDs is dramatic. Even with a return-of-premium death benefit guarantee, at even a healthy retiree life expectancy around age 88 the retiree has merely recovered principal, Figure 4 reveals that while the IRA-plus-taxable-account-funded-with-RMDs would be up to more than $300,000 even after taking a haircut for IRA taxes. And in fact, the IRA would still have over $160,000 in the account on a gross basis by then—more than the original $100,000—which would be available for beneficiaries to stretch, which means the IRA-plus-taxable-account would actually fare even better than the figure implies (unless Congress eliminates the stretch IRA, of course). By contrast, the QLAC buyer will only be leaving behind non-tax-deferred investments in a brokerage account (as any remaining pre-tax payments cease when the QLAC annuitant passes away).

Which means ultimately, while a QLAC does avoid some RMDs in the retiree’s 70s and 80s that the IRA would have triggered, it also forgoes significant growth during that time period, and in the later years the QLAC effectively liquidates the IRA even faster during the retiree’s 90s. In fact, as shown below, the QLAC always leaves behind less of an IRA to stretch than just keeping the IRA and taking the RMDs. Notably, the inferior result of the QLAC is ultimately driven both by the fact that the level of tax deferral for a QLAC isn’t actually that significant (after all, it’s not tax avoidance, it’s just tax deferral), and that a QLAC simply doesn’t have that great of an internal rate of return for those who live to life expectancy, especially when compared to an IRA invested in a balanced portfolio. On the other hand, if the IRA is invested more conservatively as well—such that the QLAC is at least somewhat more competitive on a head-to-head investment basis, thanks to mortality credits—the QLAC fares at least a little better,

as seen in Figure 5. Still, though, even when the IRA returns are “just” 5 percent, the QLAC still doesn’t pull ahead until the retiree reaches age 93, an age that even amongst healthy retiree annuitants only 20 percent of 69-year-old males and 29 percent of 69-year-old females are expected to reach (i.e., it’s still not an odds-on bet.).

When a Longevity Annuity Still Makes Sense as a QLAC in a Retirement Account Notwithstanding the issues with using a QLAC as a means to defer RMDs, it’s not necessarily a bad deal altogether to buy a longevity annuity inside of an IRA (or other retirement account) as a QLAC. If the purpose of the QLAC is specifically for retirement income to spend, and makes sense as a part of the entire retirement income picture, a QLAC is still a reasonable approach (though notably, still inferior to delaying Social Security).

For instance, lifetime annuities in general will provide more retirement income than a bond alone over a comparable time horizon, thanks to mortality credits. And as shown earlier, if the goal is to generate strong lifetime payments with big mortality credits, longevity annuities provide an even better internal rate of return in the long run than a traditional single premium immediate annuity. In other words, if the retiree is going to hold fixed-income investments anyway, and wants to hedge against a long life, lifetime immediate annuities are better than bonds, and longevity annuities are better than immediate annuities.

And if the available dollars for such a transaction happen to be inside of an IRA, so be it; the longevity annuity will simply be a QLAC. (Notably, though, longevity insurance is far less compelling as an alternative to stocks.)

Similarly, if the reality is that the retiree primarily intends to spend all the available dollars during life, and doesn’t necessarily have legacy goals, there may not be any intention to leave a remaining IRA balance behind. In that case, the fact that a QLAC will accelerate distributions out of the IRA in the retiree’s 90s isn’t a negative but simply a reality of getting the needed cash flows for retirement. The loss of a stretch IRA for the next generation is a moot point because there isn’t any plan to leave an inherited IRA in the first place.

Ultimately, then, the key point is that if a longevity annuity is appealing to buy for retirement income and longevity hedging purposes—especially as a fixed income alternative—and the available dollars to buy it are within an IRA or other retirement

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

67

Figure 4

Figure 5

account, there’s nothing wrong with using those retirement dollars and buying a QLAC, and getting the RMD deferral along the way. But if the goal is to defer RMDs in the first place, the value proposition of the QLAC isn’t very compelling. The retiree takes on a significant risk of losing out on almost two decades’ worth of compounding growth just to defer RMDs, only to

find that if he/she lives, the QLAC distributions in the client’s 90s will be even more severe than the RMDs ever would have been. Economic growth may have been given up along the way as well, if the retiree simply could have invested in a balanced portfolio over that multi-decade time horizon.

68

Journal of Personal Finance

About the Author Michael E. Kitces, MSFS, MTAX, CFP®, CLU, ChFC, RHU, REBC, CASL, is a Partner and the Director of Wealth Management for Pinnacle Advisory Group (www.pinnacleadvisory.com), a private wealth management firm located in Columbia, Maryland. In addition, he is an active writer and speaker, and publishes The Kitces Report and his blog “Nerd’s Eye View” through his website www.kitces. com.

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

69

Is Deferring Social Security the Lowest Cost Option for Adding Guaranteed Income?

David A. Littell, JD, ChFC, Chairholder of the Joseph E. Boettner Chair in Research, Co-Director of the New York Life Center for Retirement Income, The American College of Financial Services, Bryn Mawr, PA Kirk S. Okumura, MSFS, ChFC, Academic Director of the Financial Services Certified Professional® (FSCP®) Program, The American College of Financial Services, Bryn Mawr, PA

Editors’ Note: Although the Journal of Personal Finance is a refereed, research journal that targets the practitioner market, the editors believe there is a role in the literature for complex cases that are instructive or allow practitioners to compare their insights into a planning scenario with how others would approach the same situation. In last year’s Fall Issue, we published the winning case for the IARFC National Financial Plan Competition, which was a multifaceted planning case. In this issue, we are delighted to present an analysis of a retirement planning issue involving Social Security that is extremely important in today’s world where so many people are now having to make decisions about when to start the Social Security retirement benefits. The selection and publication of cases is on an editorial basis rather than a refereed basis. We would like to publish one case with each issue, and welcome submissions of any planning situations our readers have encountered such that they feel the cases would provide a beneficial learning opportunity for others in the profession. We also invite readers to submit comments on any of the case presentations.

Wade Pfau, Co-Editor Walt Woerheide, Co-Editor

Journal of Personal Finance

70

Introduction

perspective, the variability of returns may also be easier to tolerate when there is a guaranteed income floor.3

When to claim Social Security is one of many retirement decisions that should be considered together as part of a comprehensive retirement income plan. A good retirement income plan addresses the retiree’s financial objectives as well as the large number of risks and contingencies faced in retirement.

As some financial goals are more important than others, a common planning approach is to build an income floor with low-risk sources of income to ensure resources are always available to meet essential expenses of food, shelter, clothing, transportation and medical care. Other goals—like meeting discretionary expenses and legacy goals—can be satisfied by accessing a portfolio of riskier assets (stocks and bonds).

This bifurcated strategy creates income security for the most important expenses as well as some upside potential with good market returns to meet these other goals. This approach is based on life-cycle investment theory, and Wade Pfau has referred to it as the safety-first approach to retirement income planning.1 Another argument for this approach is that it efficiently addresses many common retirement risks. Locking in lifetime income streams reduces or eliminates market and other investment risks, insures against longevity risk, helps to budget resources, simplifies financial planning, and even reduces the risk of assets being lost through elder financial abuse. It also can provide peace of mind to the retiree by removing the worry of not being able to pay basic expenses when market returns are negative. Not surprisingly, research shows retirees with more guaranteed income have greater satisfaction.2 At the same time, maintaining a diversified investment portfolio provides needed flexibility to adjust for changing laws, regulations and market conditions as well as to finance discretionary spending. It also can allow for increased spending or a larger legacy for the family due to the possibility of higher-than-expected, albeit variable, returns. From a psychological

1.

2.

Wade Pfau, “What Is a Safety-First Retirement Plan?” Forbes. com, Last updated April 26, 2016. Accessed May 22, 2017, https://www.forbes.com/sites/wadepfau/2016/04/26/ what-is-a-safety-first-retirement-plan/#f83707b34534.

Steve Nyce and Billie Jean Quade, “Annuities and Retirement Happiness,” Willis Towers Watson, last modified September, 2012, accessed May 3, 2017, https://www.towerswatson.com/en-US/Insights/Newsletters/ Americas/insider/2012/Annuities-and-Retirement-Happiness.

The practical question with this approach is, “What is the most cost-effective way to build the income floor?” Social Security is the foundation of the income floor for most households.

In the past, many retirees also had monthly income from a defined-benefit pension. Together with Social Security, the pension created a generous income floor. But defined-benefit pensions have been on the decline. From 1979 to 2013, the percentage of private-sector workers participating in a pension plan plummeted from 38 percent to 13 percent, In contrast the percentages for defined-contribution plans jumped from 17 percent to 44 percent.4 Thus, today’s retirees are usually retiring with Social Security, an employer-sponsored defined-contribution plan or IRA balance (instead of a pension), and a modest taxable account. For them, it is important to find the most cost-effective way to assemble the income floor. Building a large enough income floor may require allocating a percentage of their assets to buying a commercial life annuity. However, an alternative for at least some of that floor is deferring Social Security. So a key question is what is the less expensive option for the first part of the income floor, using a portion of the retirement assets to defer Social Security or purchase an income annuity?

Intuitively, it seems that the most cost-effective strategy for buying additional income should be deferring Social Security.

Unfortunately, it would appear from claiming statistics that many clients still need to be educated about the value of delaying benefits. Although the percentage of people claiming at age 62 has declined noticeably over the past 30 years, it is still the most popular claiming age. According to the Center of Retirement Research, 48 percent of women and 42 percent of men claim at age 62. Only 10 percent of women and 9 percent of men delay past full retirement age.5 Thus, regardless of how 3.

4.

5.

Michael Finke and Wade Pfau, “The Retirement Income Challenge,” NorthwesternMutual.com. Last modified 2015. Accessed May 22, 2017, https://www.northwesternmutual.com/~/media/nmcom/files/about%20 us/retirement-income-challenge-whitepaper.pdf.

Employee Benefit Research Institute, “FAQs About Benefits—Retirement Issues: What are the trends in U.S. retirement plans?” ebri.org.

Accessed May 22, 2017. https://www.ebri.org/publications/benfaq/index. cfm?fa=retfaq14.

Alicia H. Munnell and Anqui Chen, “Trends in Social Security Claiming,” Center for Retirement Research at Boston College. Last updated May, 2015. Accessed May 22, 2017. http://crr.bc.edu/wp-content/uploads/2015/05/IB_15-8.pdf.

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

much the value of deferring Social Security is stressed, many people still do not defer.

One method for demonstrating the value of deferral uses annuities to price the added value of deferral. The idea is that an annuity is a commercial product alternative that the client could buy instead of deferring. Seeing the cost of the commercial option will provide the client the cost of duplicating delayed benefits.

There have been a couple of studies that have used annuities to illustrate the value of deferring Social Security. In the first method, Wade Pfau quantifies the 8 years of lost benefits from claiming at age 70 instead of age 62. He uses the present value as a premium for a deferred income annuity with a monthly benefit equal to the delayed retirement credit. He then calculates an implied payout rate, which was 9.5 percent in his example—a rate that cannot be matched by a commercial annuity at today’s rates. However, Pfau did not consider the difference in rates for a female or a married couple.6 A second method used by Michael Kitces assumes one claims Social Security at age 62 and saves the benefits from age 62 to age 70 at a 2 percent rate of return. He then projects a fixed annuity for a 70-year-old male and female with the accumulated savings. The results: the excess Social Security benefits are $722.00 per month, which tops a fixed annuity (not adjusting for inflation) of $572.20 for a male and $537.11 for a female. The numbers are worse for a CPI-indexed annuity. While Kitces’ approach accounted for the differences between men and women, it did not look at the benefits for a married couple.7 The approach we use looks at the two options, deferring Social Security and purchasing an annuity, and answers the practical question facing a retiree wanting to add guaranteed lifetime income: Which one costs more? In addition, our 6.

7.

Wade Pfau, “Social Security: The Best Annuity Money Can buy,” Forbes.com. Last updated November 17, 2015. Accessed May 22, 2017. https://retirementresearcher.com/ social-security-the-best-annuity-money-can-buy.

Michael Kitces. “How Delaying Social Security Can Trump Long-Term Portfolio Returns or Lifetime Annuity,” Nerd’s Eye View at Kitces.com.

Last updated April 2, 2014. Accessed May 22, 2017. https://www.kitces. com/blog/how-delaying-social-security-can-be-the-best-long-term-investment-or-annuity-money-can-buy.

71

approach considers four different scenarios: single male, single female, married couple, and married couple with a spousal benefit. In doing so, we aim to extend insight into the value of Social Security as well as provide one more way to explain to clients the value of deferring.

As other researchers have done, we approach this practical problem with a case study. We take a hypothetical individual eligible for the maximum allowable Social Security benefit at the current full retirement age of 66 and assume that this person retires at age 65. We compare the costs of two equivalent options the retiree will have: 1) Option 1: defer Social Security to age 70 and buy 5 years of additional income to cover the deferral period with a single premium immediate annuity; 2) Option 2: claim Social Security at age 65 and buy a commercial deferred income annuity that covers the additional income that would have been provided if benefits were deferred to age 70.

The first option represents the real cost of deferral, and the second demonstrates the value of the additional benefit provided by deferring Social Security by showing the cost of buying this additional income. By making this comparison we can also determine which is the better deal, buying more income by deferring Social Security or by purchasing a commercial annuity.

Making the Comparison Table 1 shows the assumptions used in the case-study comparison.

Table 1: Assumptions Full Retirement Age

66

Current Age

65

Social Security at age 65 (6.7% lower)

$2,507/month or $30,084/year

Social Security at age 66 (FRA)

$2,687/month or $32,244/year

Social Security at age 70 (32% higher)

$3,547/month or $42,562/year

Marginal Tax Bracket

28%

Assumed Inflation on Income*

3%

* Note: The annualized average COLA inflation rate from 1975, when Social Security began the COLA program, to 2017 was 3.8%. The argument could be made for a lower rate as over the past 30 years, the average rate was 2.6%. (Social Security Administration, Cost-of-Living Adjustments, accessed May 3, 2017, https://www.ssa.gov/oact/cola/colaseries.html.)

Journal of Personal Finance

72

Option 1: Calculating the Cost of Deferring Social Security Benefits Option 1 calculates the cost of delaying Social Security until age 70. An individual filing for benefits at age 65 would need to replace the lost benefits. To identify the cost of that decision, we price a single premium immediate term certain annuity (SPIA) that pays for a 5-year period to make up for the lost $2,507 per month ($30,084 annually) in Social Security benefits. With the assumption that Social Security benefits increase 3 percent each year for inflation, the annuity priced includes a three percent increase in benefits each year over the 5-year period.

Since it is more common for retirees to have more resources in tax-deferred retirement plans (IRAs and 401(k) plans), we will assume that the purchase is inside of a tax-deferred plan.

For a retiree receiving the maximum Social Security benefit we will assume that 85 percent of the monthly benefit ($2,131) will be taxable and 15 percent ($376) will be tax-free.

The distribution of annuity payments from the qualified plan will be fully taxable. This means that the annuity payout needs to be slightly larger to reflect the 15 percent of the Social Security benefit that is tax free. We will use the 28 percent marginal tax rate to gross this up to a pretax equivalent.

Thus, the annuity’s monthly benefit would need to be $2,131 + ($376/[1 – .28]) or $2,653.

Note that the $2,653 5-year certain benefit will cost the same for a single male, single female, and a couple because it is a non-life-contingent annuity—mortality plays no role in calculating the benefit.

Option 2: Calculating the Value of Deferring Benefits Option 2 determines the value of deferral by assuming that the beneficiary files for Social Security at age 65, receives $2,507 a month ($30,084 annual), and looks to replace the

delayed retirement benefits lost from not waiting by purchasing a life annuity. The lost benefit at age 70 is an additional $1,040 in monthly income ($12,480 annual). To buy this income we price a deferred income annuity (DIA) purchased at age 65 to provide the additional income beginning at age 70.

The assumptions for taxes will be the same as they are in option 1. We assume 85 percent of the Social Security benefit ($884) is taxable and 15 percent of the monthly benefit ($156) is received tax free. To fully replace the $1,040 Social Security benefit, $156 of the annuity benefit will need to be grossed up for taxes (28 percent). Thus, the total of the Social Security benefit at age 65 plus the annuity’s monthly benefit would need to equal $884 + ($156/[1 – .28]) or $1,101.

In this option, we assume the DIA won’t have death benefits, which is equivalent to Social Security benefits.8 Since the income will not begin for 5 years, the DIA’s initial monthly benefit should be increased for 3-percent inflation, a monthly payout of $1,276 in the first year. Furthermore, the DIA will have a 3 percent increase in annual benefits.

Since the equivalent annuities will be payable for life, unlike the 5-year certain annuity, the cost will be higher for a female age 65 and higher still for a couple (male and female, both age 65).

Comparison of Option 1 vs. Option 2 Table 2 below shows the results, comparing the two options for a single male, single female, and married couple. For pricing the annuity with the couple, we assume that the worker with the larger benefit is a 65-year-old male, and his spouse is a 65-year-old female. The larger Social Security benefit for a couple is equivalent to a 100 percent

8.

Social Security does have a $255 death benefit, but we ignore it due to its very limited value.

Table 2: Option 1 vs. Option 2 (SPIA and DIA Indexed to 3% Inflation) Single Male

Option 1: Cost of SPIA

Option 2: Cost of DIA

Option 1 – Option 2

$165,888

$214,399

($48,511)

Single Female

$165,888

$237,090

($71,202)

Married Couple—larger benefit

$165,888

$284,444

($118,556)

(Cannex SPIA/DIA Quote Generator was used for quoting purposes. The lowest quote generated on May 1, 2017 was used.)

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

73

joint-and-survivor benefit as the benefit is paid as long as either spouse survives.

The results corroborate the expectation that deferral is substantially cheaper than purchasing an annuity to provide the additional inflation-adjusting, guaranteed lifetime income. Delay is a good deal for all three categories. In our example, the cost of deferral is $48,511 less than the annuity purchase for a single male, $71,202 less expensive than the annuity for a single female.

But the biggest winner in delaying benefits are married couples who defer the larger Social Security benefit, with an astonishing $118,556 difference between the cost of the two options.

Comparing Social Security Deferral to CPI InflationAdjusted Annuities It’s also worth looking at a second comparison, which considers annuity pricing based on the CPI, instead of a set annual increase of 3 percent. This product is arguably more comparable to Social Security as it provides a larger increase in years of high inflation.

In this comparison in Option 1, we have kept the same SPIA product, as the retiree has a limited exposure to higher-than-expected inflation since the SPIA’s payout period is only 5 years.

In contrast, with Option 2, we have priced an inflation-adjusted annuity as the person’s exposure to inflation begins at age 70 and lasts as long at retirement lasts. Table 3 shows the higher costs of the inflation-adjusted annuities. The reason that we see this comparison as a secondary comparison, however, is that there is only one company that underwrites this type of annuity inside of a qualified plan. Many feel that the limited supply of these types of annuities make them somewhat overpriced. As you can see in Table 3, the takeaway is that the value of deferral is larger when compared to a CPI-indexed annuity.

Married Couples—Age and Gender Differences In our example, both spouses were the same age. Age differences would have an impact on the cost of a deferred income annuity to cover the additional benefit provided by Social Security

deferral. It would not affect the cost of the 5-year SPIA. If the spouse with the larger benefit is older, the cost of the commercial annuity would be higher than the example, making deferral even more valuable. The difference is more noticeable the wider the age gap. For example, if we assume the female spouse had the smaller benefit and was 10 years younger, at age 65 we would be buying a DIA that begins when the husband is age 70 and the wife is age 60. In the Table 2 example, the cost of the joint-and-survivor annuity is now $353,754 and the difference between that amount and the cost for the 5-year SPIA ($165,888) is $187,866. Because of the increase in joint life expectancy, there is even more value in deferring when the younger spouse has the smaller benefit.

If the spouse with the larger benefit is younger, the cost of the annuity would be lower than the example, because of the shorter joint life expectancy. However, even if there is a substantial age difference, the cost of the joint annuity will always be at least slightly more than a single life annuity. This means that there is still value in deferring, but the difference will be closer to the benefit incurred by the single man or single woman (depending on the gender of the younger spouse with the higher benefit).

Gender differences also affect the comparison for married couples. Commercial annuities cost more for women because of their longer life expectancy. So as compared to a married man and woman of the same age, a commercial annuity for two married women would cost more and for two married men would cost less. Again, we are discussing relative value, but in each case there is still a cost advantage to deferral versus a commercial annuity.

Married Couples—Smaller Social Security Benefit In this analysis, we have focused on the larger benefit paid to a married couple. We have assumed that the other spouse is entitled to worker’s benefit that is in excess of the spousal benefit. We have not included a comparison between the

Table 3: Option 1 vs. Option 2 (SPIA and DIA Indexed to CPI-U) Option 1: Cost of SPIA

Option 2: Cost of DIA

Option 1 – Option 2

Single Male

$165,888

$229,966

($64,078)

Single Female

$165,888

255,673

($89,785)

Married Couple—larger benefit

$165,888

295,201

($129,313)

74

Journal of Personal Finance

smaller worker’s benefit paid to a married couple and a commercial annuity. This is primarily because there is no comparable annuity product. This benefit is essentially a joint annuity that stops at the first death. Because of this short payout period, the expectation is that it would have a lower value than a single life annuity—meaning that there is less value to deferring the lower of the two workers benefits.

The more complicated situation is if the spouse is only eligible for a spousal benefit. Since the spousal benefit cannot be paid until the worker’s benefit begins, deferral can also mean the loss of the spousal benefit. This can be illustrated with our case study where both spouses are the same age. Assume that retirement is still at age 65, and both file for Social Security at age 70. For 5 years, the spouse is foregoing the spousal benefit, which is 45.8 percent of the higher-earning spouse’s PIA. Thus, the SPIA benefit must now cover $2,653 + $1,302 or $3,956 of lost income.

Likewise, if the couple opts to file early, they forego a small percentage of the spousal benefit equal to 4.2 percent of the higher earning spouse’s benefit for 5 years, which is $113.

Grossing up the tax-free portion for taxes and applying 3-percent inflation for 5 years, the lost delayed benefit would be $138. Since the spousal benefit is only paid until the death of the first spouse, there is no annuity product to match this situation. The next best approach is to buy a single-life annuity on whichever spouse’s life results in the lowest cost. In this case, that would be the male. Thus, a $138 monthly single-life DIA is purchased at a cost of $23,706 and added to the DIA that replaces the lost delayed-retirement credit, which cost $284,444. Thus, the total cost of Option 2 increases to $308,150. This cuts the advantage of the delay by about half—from $118,556 to $60,788.

be depleted in 2034 at which time Social Security income will only be sufficient to pay 79 percent of benefits. It’s difficult to see the government not resolving the funding problem, as Social Security provides more than half of retirement income for two-thirds of retirees. However, there is no certainty that this will occur.

Observations and Conclusions The most important observation is maybe the most obvious.

These examples show that deferring Social Security is a substantially lower cost option for adding lifetime income than purchasing a life annuity product today for the three categories tested. This means that for most singles and married couples it makes sense to use a portion of the assets earmarked for purchasing guaranteed income to defer Social Security.

In addition, the amount of value provided by deferral is represented by the difference in cost. There is a clear ordering of value as shown below: 1. Married couples—larger benefit—Social Security represents a joint-and-survivor annuity priced using a single life expectancy.

2. Single women—they benefit from being included in the same pool as men who have lower life expectancies.

3. Single men—they benefit from the older annuity mortality tables and higher interest rates used to calculate benefits.

For married couples two additional factors affect the value of deferral.

•

Age difference—the value of deferral increases when the lower-earning spouse is younger than the higher-earning spouse. There is somewhat less value when the higher-earning spouse is younger.

•

Spousal benefit—the value of deferral may decrease if the spouse is not entitled to a worker’s benefit. The impact depends upon the relative ages of the spouses. Also, if the spouse was entitled to a worker’s benefit smaller than the spousal benefit, that would diminish the potential loss in value.

The Reliability Factor When comparing deferring Social Security to buying a commercial annuity it is appropriate to consider the reliability of each source. Annuity payments are dependent upon the insurance company’s ability to pay benefits. However, this is a highly regulated business, with reserve requirements and other safeguards such as the state insurance pools, to help limit potential defaults. At first blush, Social Security seems to have an edge on reliability of payments; however, Social Security has risks too. According to the 2016 Trustee’s report, it is currently projected that trust fund reserves will

If we frame the value of deferral as primarily a way to buy additional income, the question is whether this is the right

©2017, IARFC. All rights of reproduction in any form reserved.

Volume 16, Issue 2

choice for everyone. There are countervailing considerations that would make this option less attractive. First, some retirees may already have sufficient income from company pensions or other sources who may not want to use up limited portfolio assets to support additional deferral. These retirees may have a greater need for liquidity and flexibility and may be better off maintaining portfolio assets than increasing income. Second, health status is a relevant factor, and those with a compromised life expectancy may not want to choose deferral just like they wouldn’t be likely candidates to buy a life annuity. Just remember that with married couples, if either spouse is in good health, deferring the larger of the couple’s benefits will still have value since the benefit is paid as long as one spouse is alive.

We also pointed out that for a couple eligible for two worker’s benefits, there is no comparable annuity product for the smaller benefit, so we can’t use this same approach to quantify the value. However, since the payout period is shorter than other options, we would expect an annuity product to have a lower cost than the other categories, and it’s safe to say that there is less value in deferring the smaller worker’s benefit for a married couple than the other three categories.

This may provide some relief to couples retiring before age 70 who are trying to figure out how to afford deferring the larger of the two Social Security benefits. If the spouse with the lower benefit claims at retirement, it can help support deferral.

In addition, retirees may still be eligible for the grandfathering rule that allows the restricted filing for a spousal benefit for those who were at least age 62 by the end of 2015. To illustrate take our couple who both retire at age 65.

If the spouse with the smaller benefit claims the worker’s benefit, the other spouse can do a restricted filing for the spousal benefit from full retirement age until age 70—and then switch to the higher worker’s benefit.

As we pointed out, the situation for a couple is different if only one worker is eligible for a worker’s benefit. When the ages of the married couple are such that the spousal benefit cannot be received because of the worker’s deferral, the value of deferring the worker’s benefit is reduced. But while the reduction can be substantial, deferral can still make sense. Note the impact is less when the nonworking spouse is younger than the worker. In that case, the worker’s deferral results in fewer years of the loss of the spousal benefit. If in

75

our example the spouse entitled to the spousal benefit was 5 years younger than the worker, then there is no loss of spousal benefits because the spouse can claim at 65 when the worker claims at age 70. Another mitigating factor would be if one spouse had a modest worker’s benefit that was smaller than the spousal benefit. Now the spouse can first claim the small worker’s benefit and later switch to the spousal benefit.

It’s also important to point out that the focus here has been choosing between deferring Social Security and purchasing an annuity today. It is possible that retirees have other options for obtaining income that will have a lower cost than today’s annuity products, which are priced based on today’s low interest rates. A retiree, for example, may have an existing deferred annuity that can be annuitized at a lower cost due to higher guaranteed interest rates and older mortality tables.

Participants in traditional pensions can often choose between a lump sum or annuity payments—and choosing the annuity is often a cost effective way to buy more income. It is worth making the comparison, but it is generally unlikely that these options will be at a lower cost than deferring Social Security.

If available, however, they can be cost-effective options for adding additional income.

Finally, an interesting question is why is the cost for the additional Social Security income so much lower than purchasing that income with a commercial annuity? First, the Social Security benefit structure was built based on older mortality tables and higher interest rates than used to price today’s annuity products. Also, mortality tables used to price annuity products are tied to the limited (and healthier) population that purchases income annuities while Social Security looks at mortality assumptions based on a much larger pool.9 Third, even though Social Security was built to provide actuarially equivalent benefits for those claiming at different ages—this is not entirely true for married couples—because of the calculation of the survivor benefit.10 This shows up our example in the large difference in value when looking at married couples.

9.

10.

Nancy J. Altman and Eric R. Kingston, Social security works!: why social security isn’t going broke and how expanding it will help us all (New York: The New Press, 2015), 30-31.

Steven A. Sass, Wei sun, and Anthony Webb, “Why Do Married Men Claim Social Security Benefits So Early? Ignorance or Caddishness?” Center for Retirement Research at Boston College. Last updated October, 2007.

Accessed May 22, 2017. http://crr.bc.edu/wp-content/uploads/2007/10/ wp_2007-171-508.pdf.

Volume 16, Issue 2

77

CE Exam for Members of the IARFC Members of the IARFC can earn CE credits through the Journal of Personal Finance (JPF). Register to take the IARFC JPF Online CE quizzes and receive both the fall 2017 and spring 2018 quizzes for $20. Two hours of IARFC CE will be awarded to anyone who achieves a score of 13 or higher per quiz. Only one submission per member is allowed; quizzes are available as JPF issues are published. To take the quiz, email editor@IARFC.org to sign up.

1.

In the article “Do Self-Control Measures Affect Saving Behavior?,” having a simple decision rule is viewed as (Kim and Hanna) a. being inconsistent with having a saving goal b. increasing the chance people will postpone saving for retirement c. a rational approach consistent with the normative life cycle model d. a commitment device

2.

Based on the logistic regression, which of saving goals listed below has a significant effect on the likelihood of saving in 2013? (Kim and Hanna) a. retirement b. purchase c. precautionary d. education

3.

Which of the following specific saving rules used by the authors to create their composite saving rule variable was the most common rule? (Kim and Hanna) a. save whatever is left at the end of the month b. save income of one family member, spend the other c. spend regular income, save other income d. save regularly by putting money aside each month

4.

In the study, as men grew older they became more _____ on gains. (Muralidhar and Berlik) a. risk averse b. risk-seeking c. loss averse d. conservative

5.

In the study, in keeping with studies of teen behavior, teen males and females tended to take _____ bets on losses than adult males and females. (Muralidhar and Berlik) a. safer b. same c. riskier d. none of the above

6.

Loss Aversion pertains to which of the following behavioral characteristics? (Muralidhar and Berlik) a. risk-seeking on losses b. risk averse on losses c. neutral on losses d. risk averse on both losses and gains

7.

In “Life Quality and Health Costs in Late Retirement” the authors found that (Cheng, Gibson, and Guo) a. There are no significant differences in hours spent watching television, reading the paper or magazines, reading books, listening to music, and sleeping throughout retirement. b. Found significant differences in the time spent walking, playing sports or exercising, visiting others, and communicating with others for retirees throughout retirement. c. The amount of time allocated to shopping, meal preparation, and pet care appears to be increasing throughout retirement. d. As individuals get older, the number of hours spent helping others, taking care of grandchildren, doing volunteer work, attending religious activities and other meetings slowly increases.

8.

In “Life Quality and Health Costs in Late Retirement” the authors found that (Cheng, Gibson, and Guo) a. Out-of-pocket medical expenses do not change that much throughout retirement. b. Retirees tend to avoid using computers in retirement. c. Retirees overall do not spend their time significantly different throughout retirement. d. The simulation results suggest the optimal withdrawal rate does not change after considering medical costs.

9.

In “Life Quality and Health Costs in Late Retirement” the authors found that all of the following EXCEPT (Cheng, Gibson, and Guo) a. Time spent on shopping decreases by 40 percent for retirees 2 years before death when compared to those 12 years before death. b. For respondents who are 2 years away from death, they devote 60 percent less time to sport and exercising compared to those who are 12 years away from death. c. Time spent visiting others does not decline throughout retirement, indicating that retirees are still socially active even within 2 years before death. d. Both the average and standard deviation of annual out-of-pocket medical expenses are constant as individuals approach death.

78

Journal of Personal Finance

10. According to recent estimates, what percentage of Americans will require long-term care (LTC) support at some point in their life? (Lumby, Browning, and Finke) a. 60% b. 65% c. 70% d. 75% 11. Which of the following insurance programs impose strict income and asset limitations with regard to LTC eligibility?

(Lumby, Browning, and Finke) a. Medicare b. Medicaid c. medigap Policies d. private LTC Insurance 12. Which of the following is the approximate annual cost of a private nursing home room in 2016? (Lumby, Browning, and Finke) a. $60,000 b. $70,000 c. $80,000 d. $90,000 13. The tax implications of qualified longevity annuity contracts have attracted the attention of financial advisors because: (Kitces) a. They provide a way to receive tax-free income from a qualified retirement plan. b. They allow for the deferral of required minimum distributions from the annuitized assets to up to age 85. c. An exclusion ratio allows for a smaller percentage of the annuity income to be taxed in the early years of retirement. d. None of the above.

14. An alternative name commonly used for qualified longevity annuity contracts is: (Kitces) a. single premium immediate annuity b. fixed indexed annuity c. deferred income annuity d. fixed deferred annuity

15. Assuming actuarially fair pricing for a 65-year old, the internal rate of return on the income generated through age 100 will generally be highest for: (Kitces) a. single premium immediate annuity purchased at age 65 b. qualified longevity annuity contract with deferral to age 75 c. qualified longevity annuity contract with deferral to age 80 d. qualified longevity annuity contract with deferral to age 85 16. The safety-first approach to retirement income planning is an approach that (Littell and Okumura) a. invests the entire portfolio in bonds and cash equivalents b. involves designating some assets to buy an income floor and the rest to invest for growth c. sets aside cash to pay for the next yearâ&#x20AC;&#x2122;s expenses d. invests in established, dividend-paying companies 17. The group that generally has the biggest difference in price between deferring Social Security and purchasing an annuity is (Littell and Okumura) a. married couples larger benefit b. single men c. single women d. married couples smaller benefit 18. There is more value in deferring Social Security for the larger benefit for a married couple when (Littell and Okumura) a. the spouse with the smaller benefit is much older b. the spouse with the smaller benefit is about the same age c. the spouse with the smaller benefit is a little bit older d. the spouse with the smaller benefit is much younger

Š2017, IARFC. All rights of reproduction in any form reserved.

As an IARFC member, your average annual investment is $225. What you will get back will be worth many times the small investment in your practice. In return for your membership, the IARFC provides you with a variety of exclusive and free member benefits and savings: industry-related product discounts, professional development opportunities, practice management tools and resources. Your membership and investment in the IARFC quickly pays for itself — if you take advantage of it. Along these lines your membership dues are a drop in the bucket compared to the potential impact of being a member will have on your career. Not only will you recoup your annual membership dues, but you significantly increase your bottom line.

Professional Designations Continuing Education Publications — academic and practice management

MRFC/RFC®/RFA® 2018 IARFC Biltmore Conference Journal of Personal Finance Register Ethics Approved Status

Programs

Group Life/ADD/LTD Advisors Launchpad, Copytalk, Social Security Timing, Tax Clarity, RME, VSA

Visibility

Professional Branding Practice Management Tools Awards

IARFC.org Member Listing Media Releases Social Media: Facebook, LinkedIn, Twitter, Blog Register opportunities: Cover Profile, New Member listing, Members in the News Logos, Binders, brochures, certificates, Code of Ethics plaque, Insignia Items Presentations, Checklists, White Papers, Plan Building Supplies Loren Dunton Memorial Award, Founders Award, Membership Recognition Award

Journal of Personal Finance International Association of Registered Financial Consultants - IARFC 1046 Summit Drive P.O. Box 42506 Middletown, Ohio 45042

www.journalofpersonalfinance.com

sponsor the collegiate 2018 IARFC National Financial Plan Competition!

Participation as a sponsor for the National Plan Competition is mutually beneficial and allows for various levels of interaction with the students and members.

learn more www.iarfc.org/FinancialPlanCompetition 800.532.9060 info@iarfc.og Join us at the IARFC Biltmore Conference, Ashville, NC, April 17-19, 2018 and meet the sponsors and judge the finals of the National Financial Plan Competition