Individual Responsibility and Economic Development: Evidence from Rainfall Data Lewis Davis Department of Economics Union College University of Perugia May 15, 2014 Is a taste for individual responsibility good for economic development? • Yes: Individualism is good for development – Focus on individual motivations and actions – Weber (1930): Individualism frees “the acquisition of goods from the inhibitions of traditional ethics.” – Lewis (1955): “initiative is likely to be stifled if the individual who makes the effort is required to share the reward with many others whose claims he does not recognize.” – Gorodnichenko and Roland (2010): individualism fosters a taste for innovation Is a taste for individual responsibility good for economic development? • No: Collectivism is good for development – Focus on interactions and transactions – Arrow (1972), Dixit (2004), Tabellini (2008b): Trust reduces transaction costs – Landa (1981) and others: Evidence that social ties matter for ethnic trading groups Individual Responsibility and Economic Development 89 countries, WVS 1981-2005 Caveat: • But: This relationship may not be causal. It may reflect – reverse causation: development destroys traditional social structures – Influence of other factors, e.g. democracy , war, history • Challenge: find an instrumental variable, X, that – influences the taste for individual responsibility, – but does not otherwise affect economic development. • Candidate: the variation of monthly rainfall. – In preindustrial societies, the return to collectivist social norms was higher in regions with greater agricultural risk. – These norms persist to the present and continue to influence values and economic outcomes. Claim 1: Risk and Social Ties • In the absence of formal insurance markets, agricultural households rely on informal risk-sharing arrangements to smooth consumption. – Rosenzweig (1988), Rosenzweig and Stark (1989), Foster and Rosenzweig (2001). • Social ties matter for risk- and income-sharing relationships: – Rosenzweig and coauthors, Angelucci et al. (2008) on Progresa, Fafchamps and Lund (2003) on health shocks. • The return to collectivist social norms is higher in societies exposed to greater agricultural risk. – Rosenzweig and Stark (1989): HH transfers are increasing in the number of daughters in law. HH’s in riskier villages marry further away, creating a geographically dispersed risk-sharing family. Claim 2: Persistence of Social Norms • Nunn and Wantchenkon (2009): the transatlantic slave trade affects trust in contemporary Africa. • Tabellini (2010): early despotism and education affect contemporary levels of trust in Europe. • Guiso, Sapienza and Zingales (2006) and Fernandez (2010): cultural norms persist among second and third generation US immigrants. • Voigtlaender and Voth (2011): Medieval anti-Semitism persists to the early twentieth century, • Alesina, Guiliano and Nunn (2011): preindustrial use of the plow affects contemporary gender roles. Related Work on Individualism • Using an instrumental variables from Linguistics – Licht, Goldschmidt, and Schwartz (2007): Individualism increases institutional quality – Tabellini (2008a): Having a general rather than limited morality increases institutional quality. Tabellini links a “general morality” to enlightenment thinking on individual rights • Using instrumental variables from genetics – Gorodnichenko and Roland (2010, 2011): Individualism increases innovation rates and economic development. • Criticism: Hard for economists to judge the validity of instruments from other fields. In contrast, I draw on an economic argument to motivate the use of rainfall variation. Contribution • Develop a formal model of risk-sharing and socialization demonstrating that equilibrium social ties are higher in riskier environments. • Present evidence of statistically and economically significant relationships between rainfall variation, the taste for individual responsibility, and economic development. • Control for current and historical agricultural development, measures of financial development, early institutional quality, and proxies for the influence of climate on institutional development. Related work on Climate and Development • Miguel, Satyanath and Sergenit (2004) and Barrios, Bertenelli and Strobl, (2010): Short run relationship between average rainfall and economic output. • Ashraf and Michalopoulos (2011): variation in prehistoric temperature levels influenced the timing of the Neolithic revolution. • Durante (2010): high rainfall variation is positively correlated with contemporary trust among regions of Europe – Similar: effects of climate on social norms. – Different: values (individual responsibility) vs. beliefs (trust), European vs. global sample, no formal model II. The Model: Risk and Socialization • Basis: Coate and Ravallion (1993) – Two dynastic households with risky income – Self-enforcing risk sharing arrangement – Maximum transfer = future value of arrangement • Extension: – A parent may invest in the strength of her child’s social ties – A taste for collective responsibility serves as a commitment device that increases the maximum transfer in informal risk-sharing arrangement s – Optimal collectivism is increasing in the size and frequency of shocks. Household Utility • Lifetime Utility V (1 r ) t vt • Periodic Utility • • • • c = consumption q = taste for collective resp. x = status of relationship s = socialization effort • Socialization technology t 0 v(ct , qt , st , xt ) u(ct ) qt xt st u '(c) 0 and u "(c) 0 qt q(st 1 ) q(0) 0, q '( s) 0, q "( s) 0 Income and Risk • Periodic Income y A,B y1 , y2 y1 y and y2 y • Distribution of income pij Pr • Prob. of Asymmetric Shock: p12 p21 p y A , y B ( yi , y j ) Informal Insurance Arrangements • State contingent net transfers (A to B): 11 ,12 ,21 ,22 • Symmetric Insurance Arrangement: (0, , , 0) • Implimentability Constraint: gain from defection <= value of relationship ve (, s) ve (0, 0) u ( y ) u ( y ) q ( s ) r G( , s) r u ( y ) u ( y ) q ( s ) p u ( y ) u ( y ) u ( y ) u ( y ) s 0 Optimization max v e (Q), s.t. G(q ,s) £ 0 and s ³ 0 q ue ( s, ) G ( , s) 0 FOC : (1 r ) Gs ( , s) 0 G( , s) 0, 0 and G( , s) 0 Complementary Slackness: s 0, 0, s 0. Shadow value of collectivism = r Shadow value of socialization = Boundary Equilibrium • Complete risk pooling * • MU of transfer is zero v ( *) 0 • IC non-binding: G ( *) 0 • No socialization s* 0 • Shadow value of socialization = 0 * 0 Interior Equilibria • Incomplete risk-pooling: * • MU of transfer is positive: ue ( *) 0 • IC is binding: G ( *, s*) 0 • Optimal Socialization: 1 r ue (ˆ( s*))ˆ '( s*) H ( , s) 0 Graph: Interior Equilibrium Implementability constraint: G(s,)=0 * Optimal socialization: H(s,)=0 s* d ds H 0 s H s ( , s) ue rq "( s) 0 e H ( , s) (1 r )G Gs u Risk, Transfers and Socialization • Equilibrium Transfer: d * 1 H s G p Gs H p 0 dp d * 1 H s G Gs H 0, d • Equilibrium Socialization: The taste for collective responsibility is increasing in the frequency and size of income shocks. ds * 1 H p G G p H 0 dp ds * 1 H G G H 0. d III. Measuring Individual Responsibility and Rainfall Variation • Ronald Inglehart (2000): World Values Survey – Five waves, 1981-2008, covering over 250,000 individuals in 89 countries. – Question: Scale = 1 to 10: – “People should take more responsibility to provide for themselves” – “The government should take more responsibility to ensure that everyone is provided for.” – Averaged across individuals and then waves to generate a single measure of individual responsibility for each country. • Potential issue: – Government responsibility is only one potential form of collective responsibility. – Does this variable really just measure a taste for government? Individualism and individual responsibility • Gorodnichenko and Roland (2012): – Individualism and collectivism reflect deeply held understandings of the self, as independent or interdependent. – This distinction in the understanding of the self gives rise to two sets of closely related attitudes, beliefs and behaviors. – A taste for individual responsibility is, therefore, hard to distinguish (conceptually and empirically) from individualism. Hofstede’s Measure of Individualism • Hofstede (1985, 2001): – Survey of IBM employees in 70+ countries. – Hofstede uses factor analysis to identify dimensions of cultural variation. Individualism is the first factor. • Hofstede (2001, p. 225) contrasts an individualistic society in which “everyone is expected to look out for themselves,” with a collectivist society in which people are “integrated into strong, cohesive in-groups, which … protect them in exchange for unquestioning loyalty.” Schwartz (2006) measures of Individualism • Schwartz (1996, 2006) – Survey of K-12 teachers, college students in 50+ countries – Affective autonomy and intellectual autonomy measure the degree to which individuals find value in pursuing their own goals and experiences and their own ideas or beliefs, respectively. – Embeddedness: meaning in life comes from group identity and social interactions. • Schwartz (2006, 140): “In autonomy cultures, people are viewed as autonomous, bounded entities…. [C]ultures with an emphasis on embeddedness, people are viewed as entities embedded in the collectivity. Meaning in life comes largely through social relationships.” Rainfall Variation • Global Historical Climatology Network (GHCN) dataset: 12 million observations of monthly precipitation from over 20K weather stations, starting in 1697. Use data from 1900 forward. • Coefficient of variation rather than standard deviation: a given change in rainfall matters more where rain is scarce. • Deserts: Omit observations from low rainfall stations/months, e.g. < 1 cm per month. • Monthly rather than annual variation: Timing matters for agriculture. Rain in November is not a perfect substitute for rain in May. Rainfall Variation • Intertemporal vs. Interspatial Rainfall variation? – The model is more consistent with interspatial income shocks, but – Rosenzweig and Binswanger (1993) find that intertemporal rainfall shocks generate contemporaneous HH income shocks. – GHNC database does not have good data on spatial location of rainfall series. • Contemporary or Historical Rainfall Variation? – Historical: Better fit for the theory – Contemporary: More counties, less prone to selection bias. – Generate both historical (pre 1900) and contemporary (post 1900) measures of rainfall variation • Generate four measures of rainfall variation (corr. 0.68-0.88) Linguistic Instrument • Pronoun Drop: The primary language permits speakers to drop pronouns. Kashima and Kashima (1998) – Example: • Spanish permits pronoun drop: “Yo amo” or “Amo.” • English does not: “I love.” – Licht, et al. (2007, p. 672) argue that the “grammar of a language may transmit and reproduce culture and social categories.” – Intuition: the subject is understood in context collectivist. • Abdurazokzoda and Davis (2014): revise and augment KK data with linguistic data from the World Atlas of Language Structures. IV. Rainfall Variation and Individualism • Theory: expect a negative relationship between rainfall variation and individualism • Empirics: – Consider four measures of rainfall variation: intertemporal and interspatial, contemporary and historical – Control for land area and number of rainfall series used to compute variable – Consider a falsification test, based on off-season rainfall variation – Exclude high migration countries: Latin America, Caribbean, neoEuropes. – Consider alternative measures of individualism from Hofstede and Schwartz Table 2: Rainfall Variation and Individualism VARIABLES lncovraincm1 (1) indresp (2) indresp (3) indresp (4) indresp -1.182** (-2.582) sdlnspacecm1 0.00112* (1.931) 0.000929 (1.374) pre1900stations Observations R-squared • (8) schw_auto -1.191** (-2.062) -1.191** (-2.311) -60.44*** (-3.476) -1.503** (-2.099) 0.0408 (0.126) 0.00110* (1.888) 0.000552 (1.470) 0.0514*** (2.803) 0.000198 (0.448) -0.537 (-0.713) off6_lncovraincm1 Constant (7) hof_idv -1.421*** (-3.253) pre1900sdlnspacecm1 area (6) indresp -0.238 (-0.465) pre1900lncovrain1cm stations (5) indresp -5.44e-07 (-1.520) 4.685*** (20.39) -5.48e-07 (-1.429) 5.269*** (15.83) 0.00339*** (3.689) -8.69e-07** (-2.266) 4.724*** (21.04) 89 0.130 87 0.053 65 0.240 0.00297*** (2.864) -1.05e-06** (-2.474) 5.668*** (11.71) -4.90e-07 (-1.222) 4.688*** (20.08) -5.17e-07 (-1.272) 4.622*** (17.34) -7.21e-06 (-0.510) 19.40* (1.937) -3.38e-07 (-0.644) 7.306*** (15.24) 58 0.158 88 0.113 73 0.107 64 0.390 49 0.238 Column 6 sample excludes Latin America, Caribbean and neo-Europes. Two concerns with Table 2 • It may be that rainfall variation is correlated with aspects of geography or climate that affect individualism, directly or indirectly. – Control for geography and climate • Perhaps the dependent variable reflects a taste for government, rather than individual responsibility per se. Control for – variables known to influence the size of the welfare state – Proxies for economic and political ideology Table 3: Controlling for Climate and Geography VARIABLES lncovraincm1 Continents F-stat p-value lndisteq (1) indresp (2) indresp (3) indresp (4) indresp (5) indresp (6) indresp (7) indresp -1.134** (-2.201) Yes 0.102 -1.243** (-2.526) No -1.324** (-2.223) No -1.073** (-2.088) No -1.222** (-2.315) No -1.325*** (-2.832) No -1.333*** (-2.725) No -0.0158 (-0.207) -0.454 (-1.266) -0.447 (-1.487) landlock mich_distsea mich_elev -1.023*** (-3.883) 1.348*** (3.948) mich_sd_elev mich_precav 0.00308** (2.041) mich_tempav 0.000207 (0.0154) mich_soil -1.027* (-1.926) 1.884 (0.989) mich_sdsoil mich_clim -0.616* (-1.677) -0.874 (-0.932) mich_sdclim Observations R-squared 89 0.174 80 0.214 80 0.234 80 0.161 80 0.137 80 0.197 80 0.169 Notes: All regressions control for the number of weather stations and land area. Coefficients not reported. Results using historical rainfall variation are similar. Table 4: Controlling for Politics and Ideology VARIABLES lncovraincm1 (1) indresp (2) indresp (3) indresp (4) indresp (5) indresp (6) indresp (7) indresp (8) indresp -1.182** (-2.582) -1.737*** (-3.780) -1.094*** (-5.143) -1.709*** (-3.091) -1.164*** (-4.767) -0.000837 (-1.118) -0.000229 (-0.0654) -1.248*** (-3.009) -0.871*** (-4.573) -2.184*** (-5.175) -1.030*** (-5.086) -1.186** (-2.385) -0.823*** (-3.415) -1.538*** (-3.668) -0.925*** (-4.010) -1.634*** (-4.007) -0.888*** (-4.370) 0.292* (1.833) 0.379*** (3.505) -0.262** (-2.570) 0.330* (1.843) 76 0.355 71 0.498 leg_socialist mich_nmbrlang trade90s wvs_trust 2.418*** (4.011) wvs_rightwing 0.358*** (2.803) wvs_stateown -0.174 (-1.544) wvs_compbad Observations R-squared 89 0.130 89 0.350 79 0.375 80 0.413 76 0.432 77 0.305 Notes: All regressions control for the number of weather stations and land area. Coefficients not reported. Results using historical rainfall variation are similar. V. Individualism and Economic Development • Endogeneity of Individualism: • Development increases individualism: – Markets may crowd out informal exchange – Urbanization and mobility may weaken family ties. – OLS estimate will be biased upward. – Development reduces individualism: – Rise of the welfare state. – OLS estimates will be biased downward. • Empirical Strategy: Two-Stage Least Squares with rainfall variation and other instruments (to allow OIR tests). Table 5: Individual Responsibility and Economic Development VARIABLES indresp (1) OLS (2) IV 0.520*** (4.780) 0.956*** (4.039) lncovraincm1 (3) First stage (4) IV (5) First stage 1.466*** (3.514) (6) IV (7) First stage 1.725*** (33.76) (8) IV 1.005*** (4.985) -1.737*** (-3.780) (11) First stage 1.104*** (3.769) -1.701*** (-3.957) -1.001** (-2.068) sdlnspacecm1 -1.016* (-1.782) pronoundrop Observations R-squared First stage F-stat OIR test (10) IV -1.125* (-1.944) pre1900lncovrain1cm leg_socialist (9) First stage -0.211 (-0.900) 0.155 (0.537) -1.094*** (-5.143) 0.400 (0.907) -0.894*** (-3.973) 0.769** (2.238) -0.921*** (-3.607) -0.0159 (-0.0558) 89 0.298 89 0.166 14.28 -- 89 0.350 65 65 0.376 87 87 0.204 79 0.087 11.98 p = 0.76 15.66 -- 3.17 -- -0.530* (-1.977) -0.763*** (-2.783) 79 0.354 0.105 (0.319) 60 -0.634** (-2.366) -0.544* (-1.927) 60 0.418 11.45 p = 0.088 Notes: All regressions control for the number of weather stations and land area. Coefficients not reported. Economic Effect of Individual Responsibility • The IV coefficient estimates are two to three times as large as the OLS estimates. This is consistent with 1) measurement error or 2) economic development increases taste for collective responsibility (rise of the welfare state). • The effect of individual responsibility on economic development is large: – Using the estimate in column (2A), a one-standard deviation increase in individualism is associated with an income ratio of exp(0.956*0.966) = 2.54. – The maximum observed difference in individual responsibility in our sample is associated with an income ratio of exp(0.956*3.80) = 37.82, which is roughly equal to the income differential between the richest and poorest five countries. VI. Robustness Tests • I consider two arguments that rainfall variation may influence economic development through channels other than its influence in the level of individualism. • Rainfall variation may be correlated with variables that influence economic development through agricultural outcomes. • Rainfall variation may be correlated with financial or institutional development or transfer. Robustness: Rainfall and Agriculture • Concern: Rainfall variation may be related to current output through its impact on agricultural outcomes, rather than its relationship to preindustrial social norms. • Controls: – Current Ag. Productivity: relative labor prod. in ag., alone and interacted with landlocked, Matsuyama (1992). – Historical Ag. Development: pop. density in 1500 and technology in 1000 CE, Ashraf and Galor (2011) – Ag. Productivity: log of average rainfall, Miguel et al. (2004) – Suitability for agriculture: average and s.d. of suitability of the climate and soil for agriculture, Michalopoulos (2012) – Contemporary importance of rain-fed Ag.: Exclude SSA. Table 6: Controlling for Agricultural Development VARIABLES indresp (1A) IV 1.333*** (4.339) lncovraincm1 lnagprod landagprod (1B) 1st Stage 0.259* (1.902) 0.251 (0.472) (2A) IV (2B) 1st Stage 0.951*** (4.109) -1.801*** (-4.652) -0.179 (-1.603) -0.697** (-2.620) ln_pd1500 Six measures of agricultural suitability NO NO NO NO Observations R-squared First stage Fstat 62 62 0.522 79 0.201 13.96 79 0.389 (5A) IV -1.732*** (-3.587) -0.523** (-2.412) NO 0.00344 (0.0196) NO 89 89 0.350 12.86 (5B) 1st Stage 1.158*** (5.792) -0.00901 (-0.108) 0.213 (0.224) lnavrain 21.64 (3B) 1st Stage 1.385*** (3.719) -1.792*** (-3.736) -0.0242 (-0.244) 0.176 (0.163) ln_CEtech1K (3A) IV 0.315 (1.190) 0.471 (0.922) 0.148 (0.948) -0.520 (-0.349) -0.439* (-1.665) Yes 54 0.271 21.42 (6A) IV (6B) 1st Stage 0.968*** (4.993) -2.036*** (-4.730) -0.339*** (-3.618) -0.919*** (-2.798) -0.116 (-0.977) 0.100 (0.0909) 0.00244 (0.0150) Yes 54 0.722 -1.879*** (-3.785) 79 0.060 14.33 79 0.393 Note: All regressions include controls for the number of weather stations, area, socialist history and average rainfall. Robustness: Climate, Finance and Institutions • Concern: Rainfall variation may influence financial development or be correlated with dimensions of climate that influence institutional quality • Controls: • Financial development: private credit as a share of GDP, Levine (2005) • Early Institutions: constraints on the executive in 1950 and 1900-20. Note: Including these controls may lead to an under-estimate of the effect of individualism on development, e.g. Tabellini (2005). • Tropics: Distance from the Equator, Hall and Jones (1999). Note: Poor overlap with AJR (2001) sample on settler mortality. • Legal Origin: British, French and socialist legal origin, Djankov (2002) • Crop Choice: natural log of the share of land suitable for wheat relative to that suitable for sugar. Engermann and Sokoloff (1998, 2002), Easterly (2007). Table 7: Controlling for Financial and Institutional Development VARIABLES (1A) IV (1B) 1st Stage indresp 1.180*** (2.710) lncovraincm1 -1.491** (-2.528) 0.00788** 0.00346 (2.010) (1.149) privatecredit (2A) IV (2B) 1st Stage 1.049** (2.523) exconst1950 -1.720*** (-2.814) 0.00702 (0.0664) (6B) 1st Stage 0.831** (1.979) -1.556*** (-2.788) 0.0592 (0.799) leg_british -0.0788 -0.339 (-0.180) (-0.728) 0.00893 -0.209 (0.0223) (-0.465) 0.350* 0.254*** (1.913) (2.703) 1.007* -0.147 (1.874) (-0.206) -0.00203** -0.000398 (-1.993) (-0.439) leg_french lndisteq lwheatsugar mich_nmbrlang Observations R-squared 1st Stage F-Stat (6A) IV 83 0.195 6.39 83 0.371 53 7.92 53 0.391 71 0.502 7.76 71 0.432 Note: All regressions include controls for the number of weather stations, area, socialist history and average rainfall. Conclusions • I argue that rainfall variation, a measure of exogenous agricultural risk, influenced the development of preindustrial social norms regarding individualism and collectivism and that these norms persist to some degree to the present. • I develop a formal economic model showing that risk increases the return to collectivist social norms and present evidence that rainfall variation is negatively related to contemporary measures of individualism. Conclusions (cont.) • Using rainfall variation as an instrument, I find a positive relationship between the exogenous variation in individualism and economic development. • These relationships are robust to controls for current and historical agricultural sector development, measures of early institutional quality, measures of climate linked to institutional transfer and development, and financial development.