The Primacy of Education in Long-Run Development

The Primacy of Education in Long-Run Development Gregory P. Casey 1 (Cornerstone Research, Boston, Massachusetts, USA) And Patrick Kent Watson 2 Sir Arthur Lewis Institute of Social & Economic Studies, University of the West Indies, St. Augustine, Trinidad & Tobago Abstract A long literature suggests that institutional quality is the driving factor behind the divergence in the incomes of former European colonies, like those of the English-speaking Caribbean. The data, however, imply that education is the crucial determinant of growth in the very long run. Using two-stage least squares procedures, we show that, when controlling for the endogeneity of human capital, trade openness and institutions, only the first is found to have a significant effect on income per capita. These results are robust to controls for geography and legal origin and are consistent with unified growth theory, which stresses education as a crucial determinant of longrun growth. 1 Corresponding author. Tel: (617) 927-3019, e-mail: gregory.casey@fulbrightmail.org. 2 Tel: (868) 662-6965, Fax: (868) 645-6329, e-mail: patrick.watson@sta.uwi.edu. Casey 1

I. Introduction This paper attempts to add to a growing literature on the fundamental determinants of growth in the very long run. Recent work suggests that institutional quality is the key driver of long term economic growth and that institutions are more important than other commonly considered variables (e.g. Hall and Jones 1998; Easterly and Levine 2003). In a series of highly influential papers, Acemoglu, Johnson and Robinson (AJR) (2001, 2002, 2005) show that endowments at the time of colonization, including the disease environment and population density, caused European colonists to develop more extractive institutions that allowed them to expropriate wealth from the rest of the population through slavery, low wage labor and burdensome tax systems. According to their theory, without adequate property rights, the majority of society faces a classic hold-up problem because of the risk that the elites will simply expropriate any accumulated wealth. Thus, total investment in the society will be greatly reduced, leading to slower economic growth. However, early proponents of similar ideas, Engerman and Sokoloff (ES 1997, 2001, 2005, 2006), argue that the same factors that retard institutional development by creating a dominant economic elite will also lead to lower levels of schooling. Similar to the protection of private property, the advent of mass education will benefit the majority of society but limit the ability of the elite to maintain economic and political dominance. Thus, the results for institutional quality may be subject to omitted variable bias. Indeed, unified growth theory argues that human capital is an especially crucial driver of economic growth because it causes technological change and influences fertility decisions, providing theoretical support for the central role of education in long run growth (Galor 2005). Glaeser et al (2004) test education and Casey 2

institutional quality in 2SLS regressions similar to those performed by AJR and find the schooling is significant while institutional quality is not. Still other authors contend that trade openness best explains the relationship between natural endowments and economic growth (e.g. Dollar and Kraay 2003; Alcala and Ciccone 2004). Dollar and Kraay (2003) attempt to separate the effects of trade and institutions using 2SLS regressions but find that multicolinearity prevents them from identifying significant partial effects of either. Alcala and Ciccone (2004) use a similar technique and find that both variables are significant. In some evidence to the contrary, Rodrik Subramanian and Trebbi (2004) use similar regressions to show that institutional quality affects growth while trade does not. Bhattacharyya (2009) tests human capital against institutions and trade in regressions with two and three endogenous regressors but does not find any significant effects. Supporting the contentions of unified growth theory, this paper provides evidence that human capital is the most prominent factor in creating the divergence of incomes in former colonies. This paper uses two-stage least squares (2SLS) regressions to estimate the effects of schooling, institutions and trade on current income per capita. The instruments are drawn directly from previous work. Drawing on the work of Easterly (2007), we use the ratio of land suitable for growing wheat and land suitable for growing sugar. We also use the settler mortality from AJR (2001), which is quite popular in the institutional literature, as well as geographic determinants of propensity to trade developed by Frankel and Romer (1999). Using this new combination of instruments, we simultaneously compare the effects of education, institutional quality and trade openness on income per capita. We find strong evidence that schooling is the dominant factor in driving economic growth in the very long run. We also show that education is the only significant variable in specifications with two or three endogenous regressors. Thus, our Casey 3

paper adds to the growing literature on fundamental determinants of economic growth by separating out these three competing factors and indentifying education as the most important factor. II. Literature Review 2.1 The Institutional Literature Under the broad category of institutional economics, an important focus area is the underlying incentives to undertake economic activity. Studies in this area suggest that individuals will not invest unless they reap the returns of their effort (e.g. North 1991, 1994). This approach highlights the importance of private property protection. For example, AJR (2001, 2002, 2005) argue that, when governments have the power to expropriate wealth from citizens, the latter will not invest, preventing economic growth through a classic hold-up problem. They use settler mortality as an instrument for institutional quality and demonstrate that institutional quality leads to economic growth (AJR 2001). They argue that European colonizers set up more extractive institutions (i.e. those that did not respect property rights) in countries where they could not settle due to the disease environment. On the other hand, when Europeans could live safely in colonized areas, institutions that respected property rights were more likely to develop. For institutions to have effects on long-run growth, their effects must persist over time. AJR (2005, 2006a,b, 2008a,b) also develop a formal model of institutional persistence. They argue that political power emanates from two sources. Specifically, de jure political power emerges from formal political institutions and de facto political power is derived from economic superiority. In their theory, elites can use their total political power to determine the economic and political institutions in a country. They will do so in a way that maximizes their personal Casey 4

wealth and privileged economic status, which, in turn, preserves their ability to decide future institutions. Thus, an equilibrium income distribution and set of institutions emerge. Specifically, the elite will not support institutions of private property protection (the focus of AJR), which would limit their ability to expropriate wealth but lead to greater overall growth. ES (1997, 2001, 2005, 2006) propose a similar theory that focuses on differences in education. They argue that colonizers developed plantation-style economies in areas with agricultural endowments suitable for mass production, such as sugar. While this led to great initial wealth, it also concentrated political and economic power in the hands of a small number of wealthy families. These elites naturally resisted the implementation of mass voting and public schooling, which would allow the rest of the population to challenge their power. Conversely, areas like North America, which did not have crops whose production structure benefitted from economic of scale, had far less inequality. Thus, there was no powerful elite to block reforms that would benefit all citizens. In fact, these areas needed an influx of free labor to maintain economic progress, leading to the extensions of voting rights and schooling that would lure new settlers. Easterly (2001, 2007) has based empirical work on this premise. In the latter paper, he uses the ratio of land suitable for growing wheat and sugar as an instrument for inequality. He then demonstrates that inequality is a fundamental determinant of growth, good governance and schooling. 2.2 Trade as a Fundamental Source of Growth Another theoretically important determinant of growth is integration with the world economy (i.e. trade openness). While many economists view trade as important for economic growth, some view trade as a fundamental source of economic growth (e.g. Sachs and Warner Casey 5

1995; Dollar and Kraay 2003, 2004; Alcala and Ciccone 2004). According to Sachs and Warner (1995, pg. 3), trade promotes growth through a myriad channels: increased specialization, efficient resource allocation according to comparative advantage, diffusion of international knowledge through trade, and heightened domestic competition as a result of international competition. Several papers argue that trade openness best explains the link between natural endowments, such as disease environment and economic growth. For example, Dollar and Kraay (2003) attempt to add both institutions and trade openness into 2SLS regressions as simultaneous endogenous regressors. Given the colinearity between the two measures, neither is significant. More recently, Rodrik, Subramanian and Trebbi (2004) use geographical determinants of the propensity to trade, developed by Frankel and Romer (1999), to test the relative importance of trade and institutions. They find that institutions are the only significant determinants of growth. Alcala and Ciccone (2004), however, use similar methods and find that both institutions and trade are significant causes of growth. 2.3 The Role of Human Capital in Unified Growth Theory Unified growth theory represents an attempt to explain the course of economic development over the entirety of human history: see Galor (2005) for the authoritative review of this theory. Thus, the main goal of unified growth theory is to explain why countries transitioned from periods of prolonged Malthusian Stagnation, where any increase in productivity immediately led to a commensurate increase in fertility, into the Modern Period of sustained economic growth, where income rose dramatically and fertility grew less quickly, stalled or even declined (Galor 2005). Casey 6

Human capital plays a key role in unified growth theory. To explain the theory, it is best to start with the notion of Malthusian Stagnation. In this first regime, technological progress and population growth are very slow (Galor 2005). As total income increases, the mortality rate slows and the fertility rate increases because families can afford to have more children (Galor and Weil 1996, 1999; Galor 2005). In other words, families like having children, exhibiting a simple income effect. So, income per capita will remain stagnant even as total income and population slowly increase. With greater population, however, comes faster technological innovation, allowing countries to slowly emerge to a different structure of production (Galor and Weil 1996, 1999). More sophisticated forms of production increase the returns to human capital. This leads to increased investment in education through two channels. First, parents will have incentive to invest in education for their children (Galor and Weil, 1996, 1999). At the same time, wealthy capital owners have incentive to push for public education systems that will generate more skilled workers for their businesses (Galor and Moav 2006; Galor, Moav and Vollrath 2009). Education, in turn, leads to a demographic shift. First, when parents invest in more education for their children, they must substitute resources away from having more children (Galor and Weil 2000). Also, the shift away from purely manual labor raises women s relative earning potential, creating incentives for women to spend less time in child-care and more time working (Galor and Weil 1996). This process might not happen right away. In the earlier stages of technological progress, the income effect in fertility decisions may still dominate (Galor 2005). This is the transition period. Eventually, however, the substitution effects will take over, leading to the Modern Period. Casey 7

At the same time, the increased levels of human capital will drive even faster technological change, leading to sustained levels of economic growth (Galor and Tsiddon 1997; Galor and Moav 2004, 2006; Galor, Moav and Vollrath 2009). The positive benefits of technological change and education will continue as before, causing the economy to enter a new equilibrium characterized by constant growth, increases in human capital, low fertility and a negative relationship between growth and fertility (Galor 2005). This theory implies that human capital dominates the role of institutions in the process of development (Galor and Snowden 2008, 140). Given that our analysis focuses on economic growth in former colonies, it is important to indicate how differing forms of colonization could have led to such differing economic outcomes within this theory. Specifically, the theory must explain why plantation economies would end up with less investment in education. Two recent extensions of unified growth theory speak to these differences. First, in the standard theory, owners of capital invest in public education because they want to benefit from increased returns to human capital (Galor and Moav 2004). In reality, however, those elites benefitting from industrialization would be interested in greater levels of education, but the landed elite would want to keep labor abundant and wages low (Galor, Moav and Vollrath 2009). Thus, countries with high amounts of labor concentration and/or high relative power for landed elites versus capitalists would have less investment in human capital (Galor, Moav and Vollrath 2009). Plantation societies fit this bill perfectly. Similarly, Galor and Mountford (2008) argue that, when technologically advanced societies trade with less developed societies, specialization causes the former to shift towards advanced goods -- speeding investments in education and the transition to the period of sustained growth but creates incentives for poorer countries to stay with agriculture -- slowing the transition. Again, Casey 8

this describes the difference between plantation societies and neo-europes very well. Plantation societies were extractive in the sense that the colonizers forced them into the exportation of commodity goods, while the neo-europes, which had less valuable agricultural goods, were free to develop for the benefit of their own inhabitants. 2.3 Human Capital as a Fundamental Determinant of Growth Glaeser et al (2004) find that deep determinants used as instruments in the institutional literature also explain much of the variation in schooling levels. They also show that, when instrumenting for both schooling and institutional quality, schooling is found to be a significant cause of economic development while institutional quality is not. They do not, however, focus on this approach and, therefore, provide no robustness checks or check for other omitted variables. Most importantly, they do not include measures of trade. Further, one of their main instruments, legal origin, has been also shown by other literature to effect financial development, an important determinant of growth, through its effects on contracting institutions (Levine 1998, 1999, 2005). Since a long theoretical literature suggests that financial development is a determinant of schooling, the effect of the instruments on growth may still pass through institutions (Galor and Zeira 1993). Using panel data, Glaeser et al (2004) provide evidence that growth in schooling may actually cause institutional development. Bhattacharyya (2009) attempts the instrumental variable approach by adding human capital to specifications directly from the earlier literature, causing him to test education against both institutions and trade. In specifications using both two and three endogenous variables, he is unable to identify which variables cause growth because of high multicolinearity. Using panel data, he finds that both institutions and schooling affect level of per capita income independently Casey 9

of one another. The lack of focus on the fundamental determinants of long run growth by both of these previous papers has allowed the intuitionalist literature to continue without accounting for the effects of human capital (e.g. Acemoglu and Johnson 2005; AJR, 2005; Eicher and Luekert 2009; Chong and Gradstein 2007). III. Data and Methods 3.1 Empirical Strategy This paper attempts to add to a growing literature on the fundamental determinants of growth in the very long run (e.g. AJR 2001, 2002; Easterly and Levine 2003; Rodrik, Subramanian and Trebbi 2004; Easterly 2007). There are many proposed proximate causes of economic growth, including physical capital accumulation, technological change and fertility rates. As Easterly (2001, 2007) notes, however, this does not change the underlying question: if investment rates are the key to determining growth, then what determines cross-country differences in the investment rate? Thus, economists have begun to search for underlying base determinants of the more immediate causes of growth. As mentioned earlier, institutional quality, education and trade have all been proposed as fundamental determinants of growth. The structural relationship between education, institutional quality, trade, natural endowments and economic growth in former European colonies is quite complex. We believe that the following system of equations provides a reasonable representation of that relationship: = + + + + + (1) = + + + + + + + (2) = + + + + + (3) = + + + + + (4) Casey 10

In this set of equations, INST represents institutional quality (in particular, the protection of private property rights), EDU is educational attainment, LTRADE is trade openness, LGDPPC is income per capita (logarithm), LWHEATSUGAR is a measure of crop potential as discussed above, SM is settler mortality, LFR is a set of geographical determinants of propensity to trade (defined in more detail below) and Z is a vector of exogenous control variables. For our purpose, these controls include continent dummy variables, legal origin dummy variables and the share of land in the tropics. LWHEATSUGAR, SM and LFR are the exogenous instruments and will be discussed in greater detail in the next section. In addition to the instruments and controls, INST is determined by education and income (equation 1). Higher levels of education can lead to better institutional quality through a number of channels. First, literacy empowers people to learn more about the role of the government in their lives and challenge the (generally more educated) political elites. Education is also needed to support and build the civil society and judiciary aspects of society. Also, greater knowledge simply allows people to be exposed to a greater array of ideas that results in greater wellbeing. Glaeser et al (2004) present empirical support for this hypothesis. INST is also determined by LGDPPC. Higher income may lead to better institutional quality through similar channels and richer societies can afford to support civil societies that serve as a check on government. Barro (1999) provides empirical support for this hypothesis. Also, the vast literature on institutional quality and economic growth uses instrumental variables in order to avoid problems arising from this potential simultaneity. This hypothesis, however, is not without critics (Acemoglu et al 2008). Education, in turn, is determined by institutions, economic growth and trade in addition to the exogenous variables in the system (equation 2). The relationship between institutions and Casey 11

education is relatively simple: strong protection of property rights increases incentives to invest in education. Similarly, increased income can lead to higher levels of education simply because richer countries can afford to provide public financing or provision for education. The relationship between trade and income is also straightforward: countries with greater output have more goods and services to trade (equation 3). The relationship we are most interested in is the determinants of income per capita (equation 4). In our relatively sparse specification, LGDPPC is determined by education, institutions and trade. These are the three factors stressed in the IV literature. As in those papers, we also allow income per capita to be directly determined by other exogenous factors, like continent, tropical location and legal heritage. The partial relationships in this equation are straightforward and are discussed at length in earlier work: 1) Institutions cause growth by creating incentives for investment, 2) education causes economic growth through the accumulation of human capital, driving technological and productivity change and altering fertility decisions and, finally, 3) trade leads to increased growth through specialization and subsequent increased in total factor productivity. Equation 4 cannot be estimated by Ordinary Least Squares (OLS) because of the endogeneity of INST, EDU and LTADE. Instead, a two-stage least squares (2SLS) strategy is employed, which generates consistent estimators only if certain requirements are met. The first requirement is that the excluded instruments, LWHEATSUGAR, SM and LTRADE, must be uncorrelated with ε 4i. It is sometimes possible to test for this using Hansen s J test: the null hypothesis of this test in that the instruments are uncorrelated with the error term in the second stage regressions. Unfortunately, this test may be used only when there are more instruments than instrumented variables. Casey 12

The second requirement for the consistency of 2SLS is that the instruments excluded from the second stage regression are correlated with the instrumented variables. When that correlation is too low, problems of weak instruments arise. Weak instruments can lead to inflated confidence intervals and even bias in the coefficients. The Angrist-Pischke F-test may be used to test for the presence of weak identification in individual instrumented variables. The 2SLS procedure is standard in the literature that attempts to differentiate between fundamental determinants of economic growth. For example, AJR (2001, 2002) use settler mortality and population density to instrument for institutional quality as the sole determinant of long-run growth, and Easterly (2007) uses LWHEATSUGAR to instrument for inequality in income level regressions. Many other papers use two endogenous variables (e.g. Easterly and Levine 2003; Dollar and Kraay 2003; Rodrik, Subramanian and Trebbi 2004; Glaeser et al, 2004; Acemoglu and Johnson 2005). 3.2 Instruments Our first major instrument is the ratio of land suitable for growing wheat and land suitable for growing sugar (denoted LWHEATSUGAR). Technically, the variable is defined as log(1+area of land suitable for growing wheat/1+area of land suitable for growing sugar). ES (1997, 2000, 2006) argue that countries with more potential for exporting cash crops (which benefitted from economies of scale) during colonization were more likely to have developed highly stratified societies. In these situations, the economic elites would be more reluctant to invest in public schooling, which would challenge their economic superiority. Easterly (2007) shows that LWHEATSUGAR influences schooling and income per capita (and measures of good governance ) through income inequality. The increased potential for exporting would also lead Casey 13

to specialization and increases in current levels of trade, causing LWHEATSUGAR to influence trade openness as well. Given the similarities between the ES and AJR theories, LWHEATSUGAR may be expected to influence institutional quality. However, the ES theory is silent on the type of limited government measures of institutional quality stressed by AJR. As shall be seen, there is also little empirical evidence for this relationship. The other main instrument is SM: settler mortality rate in former colonies. AJR (2001) argue that in colonies with higher settler mortality rates, the European colonizers were more likely to set up extractive institutions that did not respect property rights. Since political arrangements are assumed to persist over time, this would affect the long term institutional quality within a country and, by extension, economic growth. AJR (2001) demonstrate this relationship empirically, a fact which has been confirmed by others (e.g. Easterly and Levine 2003; Rodrik, Subramanian and Trebbi 2004). It seems reasonable, however, to assume that when colonizers viewed territories as sources of natural resources for exploitation, they were also less likely to invest in human capital accumulation. Glaeser et al (2004) and Bhattacharyya (2009) show that settler mortality is also empirically associated with lower levels of schooling. Given the importance of unified growth theory in explaining the prominent role of education in long run economic growth, it is important that the theory can explain the relationship between the instruments and educational attainment. Indeed, both LWHEATSUGAR and SM could also lead to lower levels of schooling through the paths suggested by unified growth theory. Galor, Moav and Vollrath (2009) note that landed elites have interest in maintaining an abundance of low-skill laborers while capitalists have interest in promoting education to take advantage of higher returns to education brought about by innovation. Thus, societies with high land concentration and high power for landed elites vis-à-vis industrial elites Casey 14

will have lower investments in education. Thus, plantation societies, which evolved due to disease environment and sugar exporting potential, will have less investment in human capital. Similarly, Galor and Mountford (2008) note that specialization due to increased trade between technologically advanced and less developed countries will increase returns to education in the former and lower them in the latter, leading to differing levels of educational attainment. LFR is the log of the Frankel-Romer (1999) values, which predict propensity to trade based on geographical characteristics. The ability of these values to predict levels of trade is well established in the literature (Rodrik, Subramanian and Trebbi 2004; Dollar and Kraay,2004; Alcala and Ciccone 2004). Based on the theories of Galor and Mountford (2008) described in the previous paragraph, we may also see how these values may have led to lower levels of schooling, as the geographic propensity to export likely influenced decisions to invest in commodity exporting and lessened incentives to invest in education. 3.3 Endogenous Variables INST is the constraint on executive power in 1990 as developed by AJR (2001). This variable is measured on a scale of 1 to 7 with higher values indicating higher constraints. Both AJR and critics, such as Glaeser et al (2004), use this as the preferred measure of political institutions because it focuses on structure rather than outcome. It is important to stress that the type of institutions we are looking at are those that protect private property rights of citizens. EDU is an index of human capital and is measured as the average secondary schooling enrollment rates from 1998-2003. This data is also taken from Easterly (2007) and it originally comes from the World Bank s World Development Indicators (WDI). Following earlier work, including Dollar and Kraay (2003), the log of (imports+exports)/gdp, denoted LTRADE, is Casey 15

used as the measure of trade openness. Data covers the period 1998-2003 to match that of schooling rates and is taken from the WDI. LGDPPC, the log of real GDP per capita, is taken from Easterly (2007). The data is from 2002 and the original source is the WDI and Penn World Tables (PWT). Data definitions and summary statistics are reported in appendix tables 1 and 2. 3.4 Control Variables For controls, we use continent dummies. As the noted in the institutional literature, European countries did not simply wipe out the existing power arrangement upon the establishment of colonies. Instead, they utilized existing structures to their own advantage. Thus, cultural differences common to difference geographical areas may help explain power structures existing under colonization, which, in turn, help determine both institutional quality and levels of human capital. They also may capture effects of pre-existing propensity to trade. I use the continent dummies developed by Easterly (2007). They are specially constructed to account for the bias usually introduced by World Bank definitions, which separates regions at least implicitly based on income. The variables are, therefore, Western Hemisphere, Middle East/Africa and East Asia/South Asia/Pacific. TROPICAL, the percentage of arable land in a tropical location, is also used and indeed it is used regularly in the literature. Variables of this type control for other negative effects of geography that may directly affect GDP per capita. We also control for legal origin, which has been found to influence economic growth through a number of channels (e.g. La Porta et al 1998, 1999, 2008). IV. Results Casey 16

Table 1 below presents the results of the estimation of equation 4 using every possible combination of the control variables. Table 1: 2SLS Estimation of = + + + + + (1) (2) (3) (4) (5) (6) (7) (8) Log GDP Log GDP Log GDP Log GDP Log GDP Log GDP Log GDP Log GDP Schooling 0.023*** 0.023** 0.022*** 0.022*** 0.021*** 0.025*** 0.022*** 0.023*** (0.007) (0.009) (0.006) (0.008) (0.007) (0.006) (0.008) (0.006) Trade -0.084-0.086-0.104-0.103-0.138-0.123-0.105-0.114 (0.304) (0.315) (0.296) (0.295) (0.277) (0.331) (0.317) (0.283) Institutions -0.025-0.026-0.030-0.031 0.070 0.030 0.060 0.052 (0.155) (0.189) (0.158) (0.168) (0.125) (0.119) (0.143) (0.111) Middle East/ Africa -0.393-0.389-0.439-0.429 East and South Asia (0.495) (0.376) (0.446) (0.334) -0.177-0.172-0.212-0.197 (0.328) (0.205) (0.300) (0.194) Tropical 0.022 0.026 0.126 0.059 (0.227) (0.220) (0.207) (0.225) British Heritage 0.008 0.021-0.120-0.135 (0.302) (0.294) (0.178) (0.175) Constant 6.906*** 6.907*** 7.067*** 7.056*** 6.624*** 6.504*** 6.433*** 6.559*** (1.610) (1.636) (1.473) (1.417) (1.082) (1.303) (1.256) (1.110) Observations 52 52 53 53 53 52 52 53 R-squared 0.731 0.730 0.729 0.728 0.717 0.713 0.721 0.721 WEAK ID P-Value Schooling 0.0109 0.0089 0.0024 0.0028 0.0298 0.0242 0.0622 0.0115 Trade 0.0010 0.0009 0.0004 0.0004 0.0005 0.0009 0.0008 0.0003 Institutions 0.1569 0.3149 0.1556 0.2393 0.0486 0.0387 0.1033 0.0227 Robust standard errors in parentheses. Results presented are from the second stage regressions. Schooling, trade and institutions are the instrumented variables. The excluded instruments are LWHEATSUGAR, the log of settler mortality and the Frankel-Romer values, which quantify geographical determinants of trade. WEAK ID P-Value is the p-value from the Angrist-Pischke F-test for weak instruments. The null hypothesis is that an individual instrumented variable is only weakly identified. The results are striking. In all regressions, schooling is the only significant variable. Indeed, it is significant at the 1% in all regressions except column 2, where it is significant at the 5% level. The coefficient is consistently around 0.022. This implies that a one standard deviation change in the level of schooling raises the log of GDP by 1.58 standard deviations, an impressively large Casey 17

effect. In addition to being insignificant, the trade coefficient is always negative, while the coefficient for institutional quality is negative in about half of the specifications. Since we have an equal number of instruments and instrumented variables, it is not possible to calculate the Hansen s J statistics to test for overidentification. Looking at the results of the Angrist-Pischke (AP) first stage F-test, we see that both schooling and trade always pass the test at the 5% level, implying that there is no problem with weak identification. Institutions fail to pass the test in the first four columns. The main problem with weakly identified instruments, however, is that they are biased in the same way that OLS coefficients are biased and may have inflated confidence intervals. Since theory suggests that bias from OLS would be to overestimate the effects of institutions on GDP, our results are clearly not driven by weak instruments. Since both trade and institutional quality are not significant, equation 4 is re-estimated without the trade variable in the first instance, then without the Institution Quality variable. Tables 2 and 3 show the results. Columns 1-8 of Table 2, shown below, recreate the columns of Table 1 after dropping trade from the second stage regressions and LFR from the first stage regressions. Casey 18

Table 2: 2SLS Estimation of = + + + + (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Log GDP Log GDP Log GDP Log GDP Log GDP Schooling 0.026*** 0.027** 0.026*** 0.026*** 0.026*** 0.029*** 0.027** 0.026*** 0.025*** 0.025*** 0.031*** 0.028*** (0.008) (0.012) (0.010) (0.008) (0.008) (0.009) (0.012) (0.008) (0.007) (0.006) (0.008) (0.006) Institutions -0.122-0.131-0.133-0.142-0.024-0.059-0.038-0.025-0.078-0.082-0.106-0.073 (0.245) (0.303) (0.252) (0.283) (0.144) (0.170) (0.198) (0.144) (0.112) (0.116) (0.138) (0.112) Middle East /Africa -0.543-0.503-0.530-0.590-0.413-0.435* East and South Asia Log GDP Log GDP Log GDP Log GDP Log GDP (0.729) (0.575) (0.497) (0.712) (0.280) (0.259) -0.236-0.182-0.201-0.270-0.164-0.188 (0.434) (0.253) (0.217) (0.401) (0.208) (0.193) Tropical -0.012 0.026 0.126 0.079 0.006 0.145 (0.275) (0.260) (0.226) (0.258) (0.234) (0.241) British Legal Heritage 0.078 0.100-0.088-0.106 (0.404) (0.416) (0.211) (0.199) Constant 6.869*** 6.821*** 6.881*** 6.942*** 6.184*** 6.118*** 6.117*** 6.192*** 6.657*** 6.698*** 6.163*** 6.251*** (1.201) (1.044) (0.891) (1.101) (0.211) (0.259) (0.248) (0.205) (0.572) (0.489) (0.282) (0.212) Observations 53 53 54 54 54 53 53 54 53 54 53 54 R-squared 0.676 0.660 0.661 0.658 0.698 0.665 0.686 0.696 0.703 0.701 0.616 0.659 OIR 0.8376 0.7966 0.6955 0.6119 WEAK ID P-Value Schooling 0.0099 0.0210 0.0023 0.0015 0.0220 0.0535 0.0954 0.0158 0.0324 0.0059 0.1244 0.0353 Institutions 0.1909 0.3350 0.1838 0.2041 0.0318 0.0758 0.1492 0.0297 0.0376 0.0417 0.1160 0.0292 Legal Heritage as Instrument N N N N N N N N Y Y Y Y Robust standard errors in parentheses. Results presented are from the second stage regressions. Schooling and trade are the instrumented variables. The excluded instruments in columns 1-7 are LWHEATSUGAR and the log of settler mortality. Columns 8-12 add British legal heritage as an excluded instrument. WEAK ID P- Value is the p-value from the Angrist-Pischke F-test for weak instruments. The null hypothesis is that an individual instrumented variable is only weakly identified. OIR is the overidentification restriction. The null hypothesis is that the excluded instruments are not correlated with the Log GDP. Log GDP Log GDP Casey 19

The results are very similar: schooling remains the only significant variable. Following Glaeser et al (2004), columns 9-12 add a dummy for British legal heritage as another instrument. This variable has a significant effect on institutions in the first-stage regression but does not have a direct effect the log of GDP per capita in the second stage. This process increases the strength of the first stage regressions for institutions, allowing institutions to pass (or almost pass) the AP F- test in all of the four regressions. It also allows us to conduct a test for overidentification, which clearly indicates that overidentification is not a problem. The main results, however, are the same. The coefficient on schooling is positive and significant at the 1% level in all regressions. The coefficient on institutions remains insignificant and negative. The dummy variable for location in the Middle East/Africa is significant at the 10% level in column 8, which only adds continent dummies as controls. Table 3 below shows the results for equation 4 with INST removed. Casey 20

Table 3: 2SLS Estimation of = + + + + (1) (2) (3) (4) (5) (6) (7) (8) Log GDP Log GDP Log GDP Log GDP Log GDP Log GDP Log GDP Log GDP Schooling 0.022*** 0.022*** 0.021*** 0.021*** 0.025*** 0.026*** 0.026*** 0.025*** (0.006) (0.005) (0.004) (0.004) (0.003) (0.003) (0.003) (0.003) Trade -0.093-0.089-0.116-0.117-0.079-0.115-0.097-0.072 (0.298) (0.304) (0.287) (0.285) (0.283) (0.338) (0.339) (0.286) Middle East/ Africa -0.335-0.349-0.372-0.379* East and South Asia (0.329) (0.260) (0.273) (0.227) -0.146-0.162-0.178-0.188 (0.248) (0.191) (0.230) (0.184) Tropical 0.022 0.012 0.148 0.116 (0.218) (0.197) (0.202) (0.201) British Legal Heritage -0.021-0.013-0.104-0.104 (0.193) (0.178) (0.153) (0.149) Constant 6.845*** 6.846*** 7.009*** 7.016*** 6.480*** 6.495*** 6.441*** 6.455*** (1.464) (1.447) (1.342) (1.311) (1.150) (1.346) (1.348) (1.156) Observations 53 53 54 54 54 53 53 54 R-squared 0.736 0.737 0.735 0.736 0.707 0.701 0.704 0.708 OIR 0.8928 0.9070 0.8655 0.8685 0.5951 0.8253 0.7078 0.6747 WEAK ID P-Value Schooling 0.0367 0.0041 0.0061 0.0008 0.0000 0.0000 0.0000 0.0000 Trade 0.0039 0.0038 0.0015 0.0016 0.0004 0.0026 0.0024 0.0003 Robust standard errors in parentheses. Results presented are from the second stage regressions. Schooling and trade are the instrumented variables. The excluded instruments are LWHEATSUGAR, the log of settler mortality and the Frankel-Romer values, which quantify geographical determinants of trade. WEAK ID P-Value is the p-value from the Angrist-Pischke F-test for weak instruments. The null hypothesis is that an individual instrumented variable is only weakly identified. Once again, the results are quite similar. Education is always significant at the 1% level, while trade openness is never significant and always has a negative coefficient. The Africa/Middle East dummy variable is significant at the 10% in a one regression, but otherwise all controls are insignificant. Since we use all three instruments, we can also perform overidentification tests for all specifications. In all cases, we find no evidence that the instruments are directly correlated with GDP per capita, indicating that regressions are valid in all cases. Similarly, both endogenous regressors always pass the AP F-test at the 5% level. Casey 21

VI. Conclusion In this paper, evidence is provided for the primacy of human capital in determining longrun economic growth in former European colonies. This result directly contrasts the long literature arguing that proper institutions were the underlying determinants in cross-country income differences (AJR 2001, 2002; Easterly and Levine 2003; Rodrik, Subramanian and Trebbi 2004). In also goes against a smaller literature pointing to the importance of trade as a fundamental determinant of growth (Dollar and Kraay 2003; Alcala and Ciccone 2004). Based on these empirical results, it is safe to conclude that the level of human capital best explains the divergence of income in former European colonies. It is certainly possible that some omitted variable has influenced these results, but the inclusion of various controls based on the literature makes this less likely. It would also be a stretch to conclude that institutions and trade do not matter in the growth process. The horserace technique used here only compares the relative importance of each of the variables over the very long-run. It ignores, therefore, the potential interrelated nature of the variables. Thus, the results should be interpreted within the long literature using these techniques to determine the fundamental causes of long-run growth (e.g. Hall and Jones 1999; AJR 2001, 2002; Dollar and Kraay 2003; Easterly and Levine, 2003; Rodrik, Subramanian and Trebbi 2004; Gleaser et al, 2004; Acemoglu and Johnson 2005; Easterly, 2007; Bhattacharyya 2009). The results, however, do clearly imply that trade and institutions are not the primary cause of the enormous differences in income between former colonies. Importantly, the results also suggest that promoting human capital can be effective when it is not accompanied by trade openness or proper institutions. Casey 22

All of these findings are consistent with the assertions of unified growth theory, which tends to focus on the growth process in colonizing countries (Galor 2005). Specifically, this theory argues that growth is a fundamental determinant of changes in both productivity and fertility rates and, therefore, serves as a primary determinant of when countries switch from Malthusian Stagnation to a Modern Period of sustained growth. It also argues that access to schooling can increase independently of political reforms. Casey 23

Bibliography Acemoglu, Daron and James Robinson. 2006a. Economic Origins of Dictatorship and Democracy. Cambridge: Cambridge University Press. Acemoglu, Daron and James Robinson. 2006b. DeFacto Political Power and Institutional Persistence. American Economic Review 96(2): 325-30. Acemoglu, Daron and James Robinson. 2008a. Persistence of Powers, Elites and Institutions. American Economic Review 98: 267-93. Acemoglu, Daron and James Robinson. 2008b. The Persistence and Change of Institutions in the Americas. Southern Economic Journal 75(2): 282-99. Acemoglu, Daron and Simon Johnson. 2005. Unbundling Institutions. Journal of Political Economy 113(5): 949-95. Acemoglu, Daron, Simon Johnson and James Robinson. 2001. The Colonial Origins of Comparative Development. American Economic Review 91(5): 1369-1401. Acemoglu, Daron, Simon Johnson and James Robinson. 2002. Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution. Quarterly Journal of Economics, 117: 1231-1294. Acemoglu, Daron, Simon Johnson, and James Robinson. 2005. Institutions as the Fundamental Determinant of Long Term Growth. In Handbook of Economic Growth. Edited by Phillipe Aghion and Steven Durlauf. Elsevier. Alcala, Fransisco and Antonio Ciccone. 2004. Trade and Productivity, Quarterly Journal of Economics May 2004: 613-646. Barro, Robert. 1999. Determinants of Democracy, Journal of Political Economy 107 (6): 158-183. Bhattacharyya, Sambit. 2009. Unbundled Institutions, Human Capital and Growth. Journal of Comparative Economics 37: 106-20. Chong, Alberto and Mark Gradstein. 2007. Inequality and Institutions. The Review of Economics and Statistics 89(3): 454-465. Ciccone, A., and M. Jarocinski. 2007. Determinants of Economic Growth: Will Data Tell? In Economics Working Papers 1052. European Central Bank. Coase, Ronald H. 1960. The Problem of Social Cost. Journal of Law and Economics 3(1): 1-44. Casey 24

Djankov, Simeon, Caralee McLiesh and Andrei Shleifer. 2007. Private Credit in 129 Countries. Journal of Financial Economics 84(2): 299-329. Djankov, Simeon, Caralee McLiesh, Tatiana Nenova, and Andrei Shleifer. 2003c. Who Owns the Media? Journal of Law and Economics 46(2): 341-81. Djankov, Simeon, Rafael La Porta, Florencio Lopez-de- Silanes, and Andrei Shleifer. 2002. The Regulation of Entry. Quarterly Journal of Economics 117(1): 1-37. Djankov, Simeon, Rafael La Porta, Florencio Lopez-de- Silanes, and Andrei Shleifer. 2003b. Courts. Quarterly Journal of Economics 118(2): 453-517. Djankov, Simeon, Rafael La Porta, Florencio Lopez de- Silanes, and Andrei Shleifer. 2008b. The Law and Economics of Self-Dealing. Journal of Financial Economics 88: 430-465 Dollar, David and Aart Kraay. 2003. Institutions, Trade and Growth. Journal of Monetary Economics 50(1): 133-162. Dollar, David and Aart Kraay. 2004. Trade, Growth and Poverty. Economic Journal 114(493). Royal Economic Society: F22-F49. Easterly, William and Ross Levine. 2003. Tropics, Germs and Crops: The Role of Endowments in Economic Development. Journal of Monetary Economics 50(1): 3-39. Easterly, William. 2001. The Middle Class Consensus and Economic Development. Journal of Economic Growth 6(4): 317-336. Easterly, William. 2007. Inequality Does Cause underdevelopment: Insights From a New Instrument. Journal of Development Economics 84: 755-76. Eicher, Theor and Andreas Leukert. 2009. Institutions and Economic Performance: Endogeneity and Parameter Heterogeteity. Journal of Money, Credit and Banking 41(1): 197-219. Engerman, Stanley and Kenneth Sokoloff. 1997. Factor Endowments, Institutions and Differential Paths of Growth Among New World Economics: A View From Economic Historians of the United States. In How Latin America Fell Behind, edited by Stephen Haber. Stanford CA: Stanford University Press. Engerman, Stanley and Kenneth Sokoloff. 2000. Institutions, Factor Endowments and Paths of Development in the New World. Journal of Economic Perspectives 14(3): 217-32. Engerman, Stanley and Kenneth Sokoloff. 2005. The Evolution of Suffrage Institutions in the Casey 25

Americas. Journal of Economic History 6(4): 891-921. Engerman, Stanley and Kenneth Sokoloff. 2006. Colonialism, Inequality and the Long-Run Paths of Development. In Understanding Poverty, Edited by Abhijit Benerjee, Roland Benabou, and Dilip Mookerjee. New York: Oxford University Press. Frankel, J,. and D. Romer. 1999. Does Trade Cause Growth? American Economic Review 89: 379-99. Galor, Oded and David N. Weil. 2000. Population, Technology and Growth; From Malthusian Stagnation to the Demographic Transition and Beyond. Galor, Oded and Andrew Mountford. 2004. Trading Population for Productivity: Theory and Evidence. Review of Economic Studies 75(4): 1143-1179. Galor, Oded and Brian Snowdon. 2008. Towards a Unified Theory of Economic Growth. World Economics 9(2): 97-151. Galor, Oded and Daniel Tsiddon. 1997. The Distribution of Human Capital and Economic Growth. Journal of Economic Growth 2(1): 93-124. Galor, Oded and David N. Weil. 1996. The Gender Gap, Fertility and Growth. American Economic Review 86(3): 374-87. Galor, Oded and David N. Weil. 1999. From Malthusian Stagnation to Modern Growth. American Economic Review 89(2): 150-54. Galor, Oded and David N. Weil. 2000. Population, Technology and Growth: From Malthusian Stagnation to the Demographic Transition and Beyond. American Economic Review 90(4): 806-28. Galor, Oded and Joseph Zeira. 1993. Income Distribution and Macroeconomics. Review of Economic Studies 60(1): 35-52. Galor, Oded and Moav, Omer. 2004. From Physical to Human Capital Accumulation: Inequality and the Process of Development. Review of Economic Studies 71(4): 1001-26. Galor, Oded and Omer Moav. 2006. Das Human-Kapital: A Theory of the Demise of the Class Structure. Review of Economic Studies 73(1): 85-117. Galor, Oded, Omer Moav and Dietrich Vollrath. 2009. Inequality in Landownership, the Emergence of Human-Capital Promoting Institutions and the Great Divergence. Review of Economic Studies 76(1): 143-79. Casey 26

Galor, Oded. 2005. From Stagnation to Growth: Unified Growth Theory. In Handbook of Economic Growth, Edited by Philippe Aghion and Steven Durlauf. Elsevier. Edition 1, Vol. 1: 171-293. Galor, Oded. 2007. Multiple Growth Regimes Insight From Unified Growth Theory. Journal of Macroeconomics 29(3): 470-75. Glaeser, Edward L, Rafael La Porta, Florencio Lopez-de-Silane, and Andrei Shleifer. 2004. Do Institutions Cause Growth? Journal of Economic Growth (9)3: 271-303. Hall, Robert E., and Charles I. Jones. 1999. Why Do Some Countries Produce So Much More Output per Worker than Others? Quarterly Journal of Economics 114(1). Hausmann, Ricardo, Dani Rodrik and Andres Velasco. 2005. Growth Diagnostics. Cambridge: John F. Kennedy School of Government, Harvard University. Johnson, Simon, William Larson, Chris Papageorgiou, and Arvind Subramanian. 2009. Is Newer Better? Penn World Table Revisions and their Impact on Growth Estimates. NBER Working Papers 15455. La Porta, Rafael, Florencio Lopez-de-Silanes and Andrei Shleifer. 2008. The Economic Consequences of Legal Origins. Journal of Economic Literature 46(2): 285-332. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer,, and Robert W. Vishny. 1999. The Quality of Government. Journal of Law, Economics and Organization 15(1): 222-79 La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny. 1998. Law and Finance. Journal of Political Economy 106(6): 1113-55. North, Douglass C. 1991. Institutions. Journal of Economic Perspectives 5 (1). American Economic Association: 97-112. North, Douglass C. 1994. Economic Performance through Time. American Economic Review 84(3). American Economic Association: 359-68. Rodriguez, Francisco and Dani Rodrik. 2001. Trade Policy and Economic Growth: A Skeptic s Guide to the Cross-National Evidence. In NBER Macroeconomics Annual 2000, vol. 15. National Bureau of Economic Research: 261-338. Rodrik, Dani and Arvind Subramanian and Francesco Trebbi. 2004. Institutions Rule: The Primacy of Institutions Over Geography and Integration in Economic Development. Journal of Economic Growth 9(2): 131-65. Rodrik, Dani. 2005. Growth Strategies. In Handbook of Economic Growth 44(4). Edited by Phillipe Aghion and Steven Durlauf. Elsevier: 967-1014 Casey 27

Sachs, Jeffrey and A. Warner. 1995. Economic Reform and the Process of Global Integration. Brookings Papers on Economic Activity 1:1-118. Williamson, Oliver E. 2002. The Theory of the Firm as Governance Structure: From Choice to Contract. Journal of Economic Perspectives 16(3): 171-195. Casey 28