132
|
A A D I T YA M AT T O O A N D A R V I N D S U B R A M A N I A N
FIGURE 12. Cross-Section Regression: Scatter Plot of Growth on Uphill Flows, Selected Economies 0.03 KOR IND
0.02
GRC
LKA
growth residuals
MUS PAN
ROM MYS IDN TUR ESP THA FIN PRT CAN SEN PHL ISL GBR NLD USA SWE MWI NZL ITA BRA DNK FRA SGP ISR AUT HND JAM MAR NIC DEU JPN ARG CRI
0.01
CHL
AUS
0.00
GTM SLV
TGO
COL
⫺0.01
HKG PRY
KEN JOR
⫺0.02
MLT PER
CHE
MEX
PAK
MDG
URY
⫺1.5
⫺1.0
⫺0.5
0.0
0.5
1.0
residuals of uphill flows coef = 0.01146931, (robust) se = 0.00344606, t = 3.33 Source: Based on the regression in table 1, column 2. Note: Controls include initial income, institutional quality, primary school enrollment, and capital stock. Coefficient, 0.01146931; (robust) standard error, 0.00344606; t, 3.33. The economies shown are ARG, Argentina; AUS, Australia; AUT, Austria; BRA, Brazil; CAN, Canada; CHE, Switzerland; CHL, Chile; COL, Colombia; CRI, Costa Rica; DEU, Germany; DNK, Denmark; ESP, Spain; FIN, Finland; FRA, France; GBR, United Kingdom; GRC, Greece; GTM, Guatemala; HKG, Hong Kong, China; HND, Honduras; IDN, Indonesia; IND, India; ISL, Iceland; ISR, Israel; ITA, Italy; JAM, Jamaica; JOR, Jordan; JPN, Japan; KEN, Kenya; KOR, Republic of Korea; LKA, Sri Lanka; MAR, Morocco; MDG, Madagascar; MEX, Mexico; MLT, Malta; MUS, Mauritius; MWI, Malawi; MYS, Malaysia; NIC, Nicaragua; NLD, Netherlands; NZL, New Zealand; PAK, Pakistan; PAN, Panama; PER, Peru; PHL, the Philippines; PRT, Portugal; PRY, Paraguay; ROM, Romania; SEN, Senegal; SGP, Singapore; SLV, El Salvador; SWE, Sweden; TGO, Togo; THA, Thailand; TUR, Turkey; URY, Uruguay; and USA, United States of America.
In column 7, to address the potential endogeneity of our uphill measure, we instrument for it with the log of population and log of area, as in Hausmann, Hwang, and Rodrik (2007). The first stage suggests that the instruments are reasonably but not exceptionally strong. In the second stage, the uphill measure has about the same magnitude and remains significant, albeit at the 10 percent confidence level. Of course, there are a number of issues with our estimation method: some of our right-hand-side variables are prone to endogeneity bias despite our use of initial rather than contemporaneous values; we may be omitting other variables; and our variables could be mismeasured. Our results should therefore be interpreted at this stage as conditional associations rather than as full identifications. In table 2, we use the UPHILL2 measure (recognizing that this may well add another layer of endogeneity bias). We introduce these measures in the cross-country regressions instead of their scale-free counterparts that we used earlier. We can either add the total share of sophisticated exports to GDP and the uphill share of that as two