How Deep Are the Roots of Economic Development? PDF

NBER WORKING PAPER SERIES HOW DEEP ARE THE ROOTS OF ECONOMIC DEVELOPMENT? Enrico Spolaore Romain Wacziarg Working Paper 18130...

NBER WORKING PAPER SERIES HOW DEEP ARE THE ROOTS OF ECONOMIC DEVELOPMENT? Enrico Spolaore Romain Wacziarg Working Paper 18130 http://www.nber.org/papers/w18130 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 June 2012 The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. At least one co-author has disclosed a financial relationship of potential relevance for this research. Further information is available online at http://www.nber.org/papers/w18130.ack NBER working papers are circulated for discussion and comment purposes. They have not been peer- reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. © 2012 by Enrico Spolaore and Romain Wacziarg. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. How Deep Are the Roots of Economic Development? Enrico Spolaore and Romain Wacziarg NBER Working Paper No. 18130 June 2012 JEL No. N10,O1,O33,O4,O5,O57 ABSTRACT The empirical literature on economic growth and development has moved from the study of proximate determinants to the analysis of ever deeper, more fundamental factors, rooted in long-term history. A growing body of new empirical work focuses on the measurement and estimation of the effects of historical variables on contemporary income by explicitly taking into account the ancestral composition of current populations. The evidence suggests that economic development is affected by traits that have been transmitted across generations over the very long run. This article surveys this new literature and provides a framework to discuss different channels through which intergenerationally transmitted characteristics may impact economic development, biologically (via genetic or epigenetic transmission) and culturally (via behavioral or symbolic transmission). An important issue is whether historically transmitted traits have affected development through their direct impact on productivity, or have operated indirectly as barriers to the diffusion of productivity-enhancing innovations across populations. Enrico Spolaore Department of Economics Tufts University Braker Hall 315 8 Upper Campus Road Medford, MA 02155 and NBER [email protected] Romain Wacziarg Anderson School of Management at UCLA C-510 Entrepreneurs Hall 110 Westwood Plaza Los Angeles, CA 90095-1481 and NBER [email protected]  "The further backward you look, the further forward you can see" (attributed to Winston Churchill).1 1 Introduction Why is income per capita higher in some societies and much lower in others? Answers to this perennial question have evolved over time. Decades ago, the emphasis was on the accumulation of factors of production and exogenous technological progress. Later, the focus switched to policies and incentives endogenously a¤ecting factor accumulation and innovation. More recently, the attention has moved to the institutional framework underlying these policies and incentives. Pushing back the debate one more degree, a key question remains as to why the proximate determinants of the wealth of nations vary across countries. A burgeoning literature seeks to better understand the deep causes of development, rooted in geography and history. As the empirical literature has moved from studying the proximate determinants of growth and development to analyzing ever deeper, more fundamental factors, important questions have arisen: How much time persistence is there in development outcomes? How far back in time should we go in order to understand contemporary economic development? Through what speci…c mechanisms do long-term geographic and historical factors a¤ect outcomes today? If economic development has deep historical roots, what is the scope for policy to a¤ect the wealth of nations? This ar- ticle discusses the current state of knowledge on these issues, focusing on recent empirical work shedding light on the complex interactions among geography, history, and comparative develop- ment. Throughout, we illustrate the major milestones of the recent literature in a uni…ed empirical framework for understanding variation in economic development. Our starting point is the long-standing debate on geography and development. There is no doubt that geographic factors, such as latitude and climate, are highly correlated with development, but the interpretation of this correlation remains hotly debated. While some of the e¤ects of geography may operate directly on current productivity, there is mounting evidence that much of the correlation operates through indirect mechanisms, i.e. through the historical e¤ects of initial geographic conditions on the spatial distribution of human characteristics, such as institutions, 1 This is the usual form of the quote attributed to Winston Churchill - for instance, by Queen Elizabeth II in her 1999 Christmas Message. According to Richard Langworth (2008, p. 577), Churchill’s words were "the longer you can look back, the farther you can look forward." 1 human capital, social capital and cultural traits, a¤ecting income and productivity over the long run.2 We review the literature on the legacy of geographic conditions in Section 2. A major theme emerging from the recent literature is that key human characteristics a¤ecting development are transmitted from one generation to the next within populations over the long run, explaining why deep historical factors still a¤ect outcomes today. A growing body of new empirical work has focused on the measurement and estimation of long-term e¤ects of historical variables on contemporary income by explicitly taking into account the ancestral composition of current populations (Enrico Spolaore and Romain Wacziarg, 2009; Louis Putterman and David Weil, 2010; Diego Comin, William Easterly and Erick Gong, 2010). We survey contributions to this new literature in Section 3. In Section 4 we provide a general taxonomy to discuss di¤erent channels through which in- herited human characteristics may impact economic development. Our discussion builds on an extensive evolutionary literature on the complex interactions among genetic, epigenetic, and cul- tural transmission mechanisms, and on the coevolution of biological and cultural traits (Luigi Luca Cavalli-Sforza and Marcus W. Feldman, 1981; Robert Boyd and Peter J. Richerson, 1985; Richerson and Boyd, 2005; Eva Jablonka and Marion J. Lamb, 2005), as well as on a growing literature on cultural transmission and economic outcomes (e.g., Alberto Bisin and Thierry Verdier, 2000, 2001; Alberto F. Alesina, Paola Giuliano and Nathan Nunn, 2011; Guido Tabellini, 2008). An important issue is whether historically transmitted characteristics a¤ect economic development through their direct impact on productivity, or operate indirectly as barriers to the di¤usion of technological and institutional innovations across populations. 2 Geography and Development 2.1 Long-Term E¤ects of Geography The hypothesis that geographic factors a¤ect productivity and economic development has a long pedigree, going back to Niccolò Machiavelli (1531), Charles-Louis de Montesquieu (1748) and Alfred Marshall (1890). A vast empirical literature has documented high correlations between current levels of income per capita and a series of geographic and biological variables, such as climate and 2 For recent discussions of these issues from di¤erent perspectives see Oded Galor (2005, 2011) and Acemoglu, Jonhson and Robinson (2005). 2 temperature (Gunnar Myrdal, 1968; Andrew M. Kamarck, 1976; William Masters and Margaret McMillan, 2001; Je¤rey D. Sachs 2001), the disease environment (David E. Bloom and Sachs, 1998; Sachs, Andrew D. Mellinger, and John L. Gallup, 2001; Sachs and Pia Malaney, 2002), natural resources (Sachs and Andrew M. Warner, 2001), and transportation conditions (Jordan Rappaport and Sachs, 2003). In order to illustrate the main empirical …ndings of the contributions discussed herein, we punc- tuate this paper with our own empirical results based on a uni…ed dataset, regression methodology and sample. This analysis is not meant to be an exhaustive recapitulation of existing results, but simply to illustrate some important milestones in the recent literature. We use, alternately, log per capita income in 2005 (from the Penn World Tables version 6.3) as a measure of contemporary economic performance, and population density in 1500 (from Colin McEvedy and Richard Jones, 1978) as a measure of economic performance in 1500, and regress these on a variety of proposed determinants of development, starting here with geographic factors.3 Table 1, column 1 shows that a small set of geographic variables (absolute latitude, the percent- age of a country’s land area located in tropical climates, a landlocked country dummy, an island country dummy) can jointly account for 44% of contemporary variation in log per capita income, with quantitatively the largest e¤ect coming from absolute latitude (excluding latitude causes the R2 to fall to 0:29). This result captures the ‡avor of the above-cited literature documenting a strong correlation between geography and income per capita. While the correlation between geography and development is well-established, the debate has centered around causal mechanisms. A number of prominent economists, including Myrdal (1968), Kamarck (1976), and Sachs and co-authors, argued that geographic factors have a direct, contem- poraneous e¤ect on productivity and development. In particular, Sachs (2001) claims that economic underdevelopment in tropical countries can be partly explained by the current negative e¤ects of their location, which include two main ecological handicaps: low agricultural productivity and a high burden of diseases. Tropical soils are depleted by heavy rainfall, and crops are attacked by pests and parasites that thrive in hot climates without winter frosts (Master and McMillan, 2001). 3 As is well-known, in the pre-Industrial, Mathusian era population density is the appropriate measure of a society’s economic performance since any technological improvement leads to increases in population rather than to increases in per capita income. For a theoretical and empirical analysis of the relationship between population size, population density and long-term growth in Malthusian times see Michael Kremer (1993). For in-depth discussions of this topic see Galor (2005) and the recent contribution by Ashraf and Galor (2011). 3 Warm climates also favor the transmission of tropical diseases borne by insects and bacteria, with major e¤ects on health and human capital. In sum, according to this line of research, geography has direct current e¤ects on productivity and income per capita. Other scholars, in contrast, claimed that geography a¤ects development indirectly through historical channels, such as the e¤ects of prehistoric geographic and biological conditions on the onset and spread of agriculture and domestication (Jared Diamond, 1997; Ola Olsson and Douglas Hibbs, 2005), and the e¤ects of crops and germs on the settlement of European colonizers after 1500 (Stanley Engerman and Kenneth Sokolo¤, 1997 and 2002; Daron Acemoglu, Simon Johnson and James Robinson, 2001, 2002; William Easterly and Ross Levine, 2003). Jared Diamond (1997) famously argued that the roots of comparative development lie in a series of environmental advantages enjoyed by the inhabitants of Eurasia at the transition from a hunter-gather economy to agricultural and pastoral production, starting roughly in 10,000 BC (the Neolithic Revolution). These advantages included the larger size of Eurasia, its initial biological conditions (the diversity of animals and plants available for domestication in prehistoric times), and its East-West orientation, which facilitated the spread of agricultural innovations. Building on these geographic advantages, Eurasia experienced a population explosion and an earlier acceleration of technological innovation, with long-term consequences for comparative development. According to Diamond, the proximate determinants of European economic and political success ("guns, germs, and steel") were therefore the outcomes of deeper geographic advantages that operated in prehistoric times. The descendants of some Eurasian populations (Europeans), building on their Neolithic advantage, were able to use their technological lead (guns and steel) and their immunity to old- world diseases (germs) to dominate other regions in modern times - including regions that did not enjoy the original geographic advantages of Eurasia. In order to test Diamond’s hypotheses, Olsson and Hibbs (2005) provide an empirical analysis of the relation between initial biogeographic endowments and contemporary levels of development.4 They use several geographic and biological variables: the size of continents, their major directional axis (extent of East-West orientation), climatic factors, and initial biological conditions (the number of animals and plants suitable to domestication and cultivation at each location 12,000 years ago). We revisit their empirical results in columns 2 through 5 of Table 1. In order to reduce the e¤ect of post-1500 population movements, the Olsson-Hibbs sample excludes the neo-European countries 4 See also Hibbs and Olsson (2004). 4 (Australia, Canada, New Zealand and the United States), as well as countries whose current income is based primarily on extractive wealth. Column 2 replicates the estimates of column 1 using this restricted sample - the joint explanatory power of geographic variables rises to 55%, since the new sample excludes regions that are rich today as a result of the guns, germs and steel of colonizing Europeans rather than purely geographic factors. Columns 3-5 add the two main Olsson-Hibbs geographic variables, …rst separately and then jointly: a summary measure of biological conditions and a summary measure of geographic con- ditions.5 Both geographic and biological conditions variables are highly signi…cant when entered separately. When entered jointly, the geographic conditions variable remains highly signi…cant and the overall explanatory power of the regressors remains large (52%). These empirical results provide strong evidence in favor of Diamond’s hypotheses, while suggesting that the geographic component of the story is empirically more relevant than the biological component. Column 6 goes further in the attempt to control for the e¤ect of post-1500 population movements, by restricting the sample to the Old World (de…ned as all countries minus the Americas and Oceania). The e¤ect of geography now rises to 64% - again highly consistent with Diamond’s idea that biogeographic conditions matter mostly in the Old World.6 2.2 The Legacy of the Neolithic Transition The long-term e¤ects of geographic and biogeographic endowments also play a central role in the analysis of Quamrul Ashraf and Oded Galor (2011). While their main goal is to test a central 5 These are the …rst principal components of the above-listed factors. Since latitude is a component of the geo- graphic conditions index, we exclude our measure of latitude as a separate regressor in the regressions that include geographic conditions. 6 Olsson and Hibbs also …nd that geographic variables continue to be positively and signi…cantly correlated with income per capita when they control for measures of the political and institutional environment. They show that such political and institutional measures are positively correlated with geographic and biogeographic conditions, consistent with the idea that institutions could mediate the link between geography and development. As they notice (p. 934), controlling for political-institutional variables raises well-known issues of endogeneity and reverse causality (for instance, richer countries can have the resources and ability to build better institutions). They write: "Researchers have struggled with the joint endogeneity issue, proposing various instrumental variables to obtain consistent estimates of the proximate e¤ects of politics and institutions on economic performance, along with the related question of how much in‡uence, if any, natural endowments exert on economic development independent of institutional development. None of these attempts is entirely persuasive in our view." We return to these important issues below. 5 tenet of Malthusian theory (that per capita income gains from technological improvements in the pre-industrial era were largely dissipated through population growth), their approach leads them to provide further evidence relating to Diamond’s hypotheses and the legacy of geography. Ashraf and Galor demonstrate that the spread of agriculture (the Neolithic transition) was driven by geographic conditions (climate, continental size and orientation) and biogeographic conditions (the availability of domesticable plant and big mammal species). They empirically document how geographic factors in‡uenced the timing of the agricultural transition. They also show that biogeographic variables, consistent with Olsson and Hibbs (2005), are strongly correlated with population density in 1500, but argue that the only way these variables matter for economic performance in pre-industrial times is through their e¤ect on the timing of the adoption of agriculture. This paves the way to using biogeographic factors as instruments for the timing of the Neolithic transition in a speci…cation explaining population density in 1500. Table 2 illustrates these …ndings in our uni…ed empirical setup. In column 1, we regressed the number of years since the Neolithic transition (obtained from Chanda and Putterman, 2007) on a set of geographic variables - i.e. this is the …rst stage regression.7 These geographic conditions account for 70% of the variation in the date of adoption of agriculture, and most enter with a highly signi…cant coe¢ cient. Column 2 shows the reduced form - again, geographic factors account for 44% of the variation in population density in 1500, consistent with the results of Table 1 for the contemporary period.8 Ashraf and Galor (2011) argue that, while geographic factors may have continued to a¤ect eco- nomic development after the introduction of agriculture, the availability of prehistoric domesticable wild plant and animal species did not in‡uence population density in the past two millenia other than through the timing of the Neolithic transition. Therefore, they use these variables, obtained from the Olsson and Hibbs (2005) dataset, as instruments to estimate the e¤ect of the timing of 7 For comparability we use the same set of variables as above, except instead of the Olsson-Hibbs summary indices of geographic and biological conditions, we directly include the number of annual or perennial wild grasses and the number of domesticable big mammals, so as to maintain consistency with Ashraf and Galor (2011). 8 Interestingly, the e¤ect of latitude is negative. Ashraf and Galor (2011) indeed observe that: "in contrast to the positive relationship between absolute latitude and contemporary income per capita, population density in pre- industrial times was on average higher at latitudinal bands closer to the equator". Thus, the e¤ects of geographic factors have varied over di¤erent periods of technological development, in line with the idea that the e¤ects of geography on development are indirect. 6 the Neolithic transition on population density. The results of column 3 (OLS) and column 4 (IV) of Table 2 illustrate their …ndings: years since the agricultural transition has a strong, statistically signi…cant positive e¤ect on population density in 1500. Interestingly, the IV e¤ect is quantitatively larger than the OLS estimate.9 The magnitude of the e¤ect is large, as a one standard deviation change in years of agriculture is associated with 63% of a standard deviation change in log popula- tion density in 1500 (OLS). The corresponding standardized beta coe¢ cient using IV is 88%. All the other regressors feature much smaller standardized e¤ects. In addition to providing strong support in favor of the Malthusian view that technological improvements impact population density but not per capita income in pre-industrial societies, the results in Ashraf and Galor (2011), as summarized in Table 2, add an important quali…er to the Olsson and Hibbs (2005) results. They show, not only that an earlier onset of the Neolithic transition contributed to the level of technological sophistication in the pre-industrial world, but also that the e¤ect of Diamond’s biogeographic factors may well operate through the legacy of an early exposure to agriculture. 2.3 Reversal of Fortune and the Role of Institutions Diamond’s book as well as the empirical work by Olsson and Hibbs and Ashraf and Galor suggest an important role for geography and biogeography in the onset and di¤usion of economic development over the past millennia. However, these analyses leave open the question of whether the e¤ects of geography operate only through their historical legacy, or also a¤ect contemporaneous income and productivity directly. Nunn (2009) makes a closely related point when discussing Nunn and Puga (2007), an attempt to estimate the magnitude of direct and indirect (historical) e¤ects of a speci…c geographic characteristic: terrain ruggedness, measured by the average absolute slope of a region’s surface area. Nunn and Puga (2007) argue that ruggedness has a negative direct e¤ect on agriculture, construction and trade, but a positive historical e¤ect within Africa because it allowed protection from slave traders. They …nd that the historical (indirect) positive e¤ect is twice as 9 Ashraf and Galor (2011, p. 2016) argue that, in regressions of this type: "reverse causality is not a source of concern, (...) [but] the OLS estimates of the e¤ect of the time elapsed since the transition to agriculture may su¤er from omitted variable bias (....)". The sign of the expected OLS bias therefore depends on the pattern of correlations between the omitted factors, the dependent variables and the included regressors. Finding an IV e¤ect that is larger than the OLS e¤ect is also broadly consistent with IV partly addressing measurement error in years since the agricultural transition, although care must be exercised with this inference in the multivariate context. 7 large as the negative (direct) contemporary e¤ect. A broader issue with Diamond’s geographic explanation is that it denies a role for speci…c di¤erences between populations, especially within Eurasia itself. For example, Joyce Appleby (2010, p. 11) writes: "How deep are the roots of capitalism? [... ] Jared Diamond wrote a best-selling study that emphasized the geographic and biological advantages the West enjoyed. Two central problems vex this interpretation: The advantages of the West were enjoyed by all of Europe, but only England experienced the breakthrough that others had to imitate to become capitalistic. Diamond’s emphasis on physical factors also implies that they can account for the speci…c historical events that brought on Western modernity without reference to the individuals, ideas, and institutions that played so central a part in this historic development." We return to these important questions below. Acemoglu, Johnson and Robinson (2002) address the issue of whether geography may have had a direct e¤ect on development by documenting a "reversal of fortune" among former European colonies. This reversal of fortune suggests that the e¤ect of geography was indirect. The simplest geography story states that some geographic features are conducive to development, but this story is inconsistent with the reversal of fortune since the same geographic features that made a society rich in 1500 should presumably make it rich today.10 More sophisticated geography-centered arguments rely on the idea that geographic features conducive to development vary depending on the time period. A reversal of fortune would be consistent with non-persistent direct e¤ects of geography on productivity: features of geography that had positive e¤ects on productivity in the past could have become a handicap in more recent times. However, such shifts would then have to be explained by speci…c changes in non-geographic factors (e.g., a technological revolution). To proxy for levels of economic productivity and prosperity in a Malthusian world, Acemoglu, Johnson and Robinson (2002) use data on urbanization patterns and population density. Contem- porary income per capita is regressed on these measures of economic performance in 1500 to assess whether a reversal of fortune has occurred. The bottom panel of Table 3 mirrors their main results: in various samples that all exclude European countries, the relationship between population density 10 Acemoglu, Johnson and Robinson (2002) state that: "The simplest version of the geography hypothesis empha- sizes the time-invariant e¤ects of geographic variables, such as climate and disease, on work e¤ort and productivity, and therefore predicts that nations and areas that were relatively rich in 1500 should also be relatively prosperous today." (p. 1233) 8 in 1500 and log per capita income in 2005 is negative. In the regression that corresponds to their baseline (column 3), looking only at former European colonies, the e¤ect is large in magnitude and highly signi…cant statistically: the standardized beta on 1500 density is 48% and the t-statistic is 7. Similar results hold for the whole World minus Europe (column 1), and also when restricting attention only to countries not currently populated by more than 50% of their indigenous popula- tion (columns 5 and 7).11 These important …ndings suggest that the observed correlation between geographic variables and income per capita are unlikely to stem from direct e¤ects of geography on productivity. In contrast, they point to indirect e¤ects of geography operating through long-term changes in non-geographic variables. Acemoglu, Johnson and Robinson (2002) argue that the reversal re‡ects changes in the insti- tutions resulting from European colonialism: Europeans were more likely to introduce institutions encouraging investment in regions with low population density and low urbanization, while they introduced extractive, investment-depressing institutions in richer regions. This interpretation is consistent with Acemoglu, Johnson and Robinson (2001), where the focus is on an indirect biogeo- graphic channel: European settlers introduced good (productivity-enhancing) institutions in regions where they faced favorable biogeographic conditions (low mortality rates), and bad institutions in regions where they faced unfavorable biogeographic conditions (high mortality rates).12 This line of research is part of a body of historical and empirical work emphasizing institutional di¤erences across societies, including seminal contributions by Douglass North and Robert Thomas (1973), North (1981, 1990) and Eric L. Jones (1988), and more recently Stanley L. Engerman and Kenneth L. Sokolo¤ (1997, 2002), Sokolo¤ and Engerman (2000), and Acemoglu, Johnson and Robinson (2001, 2002, 2005). In particular, Engerman and Sokolo¤ (1997) provided a path-breaking investigation of the interplay between geographic and historical factors in explaining di¤erential growth performance in the Americas (United States and Canada versus Latin America). They pointed out that Latin American societies also began with vast supplies of land and natural resources per capita, and “were among the most prosperous and coveted of the colonies in the seventeenth and eighteenth century. Indeed, so promising were these other regions, that Europeans of the time 11 To de…ne whether a country’s population today is composed of more than 50% of descendents of its 1500 popula- tion we rely on the World Migration Matrix of Putterman and Weil (2010), which we discuss in much greater detail in Section 3. 12 For a critical reassessment of the empirical strategy in Acemoglu, Johnson and Robinson (2001), see David Albouy (2011). 9 generally regarded the thirteen British colonies of the North American mainland and Canada as of relatively marginal economic interest –an opinion evidently shared by Native Americans who had concentrated disproportionally in the areas the Spanish eventually developed. Yet, despite their similar, if not less favorable, factor endowment, the U.S. and Canada ultimately proved to be far more successful than the other colonies in realizing sustained economic growth over time. This stark contrast in performance suggests that factor endowment alone cannot explain the diversity of outcomes.”(Engerman and Sokolo¤, 1997, p. 260). Their central hypothesis was that di¤erences in factor endowments across New World colonies played a key role in explain di¤erent growth patterns after 1800, but that those e¤ects were indirect. Di¤erent factor endowments created “substantial di¤erences in the degree of inequality in wealth, human capital, and political power,” which, in turn, were embodied in persistent societal traits and institutions. Societies that were endowed with climate and soil conditions well-suited for growing sugar, co¤ee, rice, tobacco and other crops with high market value and economies of scale ended up with unequal slave economies in the hands of a small elite, implementing policies and institutions that perpetuated such inequality, lowering incentives for investment and innovation. In contrast, a more equal distribution of wealth and power emerged in societies with small-scale crops (grain and livestock), with bene…cial consequences for long-term economic performance. An alternative to the institutional explanation for the reversal of fortune is rooted in the com- position of world populations. For while Europeans may have left good institutions in former colonies that are rich today, they also brought themselves there. This point is stressed by Edward L. Glaeser, Rafael La Porta, Florencio Lopez-de-Silanes and Andrei Shleifer (2004, p. 274), who write: “[Acemoglu, Johnson and Robinson’s] results do not establish a role for institutions. Specif- ically, the Europeans who settled in the New World may have brought with them not so much their institutions, but themselves, that is, their human capital. This theoretical ambiguity is consistent with the empirical evidence as well.” The top panel of Table 3 shows that when Europe is included in the sample, any evidence for reversals of fortune disappears: the coe¢ cient on 1500 population density is essentially zero for the broadest sample that includes the whole World (column 1). For countries that were not former European colonies, there is strong evidence of persistence, with a positive signi…cant coe¢ cient on 1500 density. The evidence of persistence is even stronger when looking at countries that are populated mostly by their indigenous populations (the evidence is yet stronger when de…ning 10 "indigenous" countries more strictly, for instance requiring that more than 90% of the population be descended from those who inhabited the country in 1500).13 In other words, the reversal of fortune is a feature of samples that exclude Europe and is driven largely by countries inhabited by populations that moved there after the discovery of the New World, and now constitute large portions of these countries’ populations - either European colonizers (e.g. in North America and Oceania) or African slaves (e.g. in the Caribbean). These regularities suggest that the broader features of a population, rather than institutions only, might account for the pattern of persistence and change in the relative economic performance of countries through history. Of course, the quality of institutions might be one of the features of a population (perhaps not the only feature) that makes it more or less susceptible to economic success, but the basic lesson from Table 3 is that one cannot abstract from the ancestral structure of populations when trying to understand comparative development. This central idea is the subject of Sections 3 and 4, so we will say little more for now. Recent work casts additional doubt on the view that national institutions are paramount. In a paper on African development, Stelios Michalopoulos and Elias Papaioannou (2010) …nd that national institutions have little e¤ect when one looks at the economic performance of homogeneous ethnic groups divided by national borders. They examine the e¤ects on comparative development of national contemporary institutions structures and ethnicity-speci…c pre-colonial societal traits, using a methodological approach that combines anthropological data on the spatial distribution of ethnicities before colonization, historical information on ethnic cultural and institutional traits, and contemporary light density image data from satellites as a proxy of regional development. Overall, their …ndings suggest that long-term features of populations, rather than institutions in isolation, play a central role in explaining comparative economic success.14 13 The R2 we obtain in the regressions of Table 3 are commensurate in magnitude to those obtained from compa- rable speci…cations in Acemoglu, Johnson and Robinson (2002). As expected, as the sample exands beyond former European colonies, the explanatory power of past development for current development falls, and correspondingly the R2 falls. In general, R2 s are quite low because we are regressing two di¤erent measures of development on each other (per capita income and population density in 1500), and both variables (particularly historical population density) are measured with signi…cant amounts of error. 14 The e¤ects of ethnic/cultural di¤erences on economic outcomes within a common national setting are also docu- mented by Beatrix Brügger, Rafael Lalive and Josef Zweimüller (2009), who compare di¤erent unemployment patterns across the language barrier in Switzerland, and …nd that job seekers living in Latin-speaking border communities take about 18% longer to leave unemployment than their neighbors in German- speaking communities. 11  In sum, the evidence on reversal of fortune documented by Acemoglu, Johnson and Robinson (2002) is consistent with an indirect rather than direct e¤ect of geography on development, but is open to alternative interpretations about the mechanisms of transmission. A key issue is whether the di¤erential settlement of Europeans across colonies after 1500 a¤ect current income in former colonies exclusively through institutions, as argued by Acemoglu, Johnson and Robinson (2001, 2002, 2005), or through other relevant factors and traits brought by Europeans, such as human capital (Glaeser, La Porta, Lopez-de-Silanes and Shleifer, 2004) or culture (Landes, 1998). Disentangling the e¤ects of speci…c societal characteristics, such as di¤erent aspects of insti- tutions, values, norms, beliefs, other human traits, etc., is intrinsically di¢ cult, because these variables are conceptually elusive to measure, deeply interlinked, and endogenous with respect to economic development. In spite of these intrinsic di¢ culties, a growing body of historical and empirical research, focusing on natural experiments, has attempted to provide insights on the com- plex relationships between geography and human history and their implications for comparative development (for example, see the contributions in Diamond and Robinson, 2010). As we discuss in the next two sections, recent contributions stress the importance of persistent characteristics transmitted intergenerationally over the long run. This literature is consistent with anthropological work, such as by Carmela Rosalba Guglielmino, Carla Viganotti, Barry Hewlett, and Luigi Luca Cavalli-Sforza (1995), showing in the case of Africa that cultural traits are trans- mitted intergenerationally and bear only a weak correlation with environmental characteristics: "Most traits examined, in particular those a¤ecting family structure and kinship, showed great conservation over generations. They are most probably transmitted by family members." 3 Development and the Long-Term History of Populations 3.1 Adjusting for Ancestry Historical population movements play a central role in the debate regarding the mechanism linking geography and economic development, as well as the interpretation of reversals of fortune. Recent research has focused on the measurement and estimation of the long-term e¤ects of historical factors on contemporary income by explicitly taking into account the ancestral composition of current populations. We review some of these contributions in this section. An important contribution within this line of research is Louis Putterman and David Weil 12 (2010). They examine explicitly whether it is the historical legacy of geographic locations or the historical legacy of the populations currently inhabiting these locations that matters more for contemporary outcomes. To do so, they assemble a matrix showing the share of the contemporary population of each country descended from people in di¤erent source countries in the year 1500. The de…nition of ancestry is bound to have some degree of arbitrariness, since it refers to ancestral populations at a speci…c point in time. However, choosing 1500 is sensible since this date occurs prior to the massive population movements that followed the discovery of the New World, and data on population movements prior to that date are largely unavailable. Building on previous work by Valerie Bockstette, Areendam Chanda and Louis Putterman (2002) and Chanda and Putterman (2007), they consider two indicators of early development: early state history and the number of years since the adoption of agriculture. They then construct two sets of historical variables, one set representing the history of the location, the other set weighted using the migration matrix, representing the same variables as they pertain not to the location but the contemporaneous population inhabiting this location. Inevitably, measuring these concepts is fraught with methodological issues. For instance, when it comes to state antiquity, experience with centralization that occurred in the distant past is discounted exponentially, while no discounting is applied to the measure of the years of agriculture. While these measurement choices will surely lead to future re…nements, it is the comparison between the estimates obtained when looking at the history of locations rather than populations that leads to interesting inferences. According to this approach, the United States has had a relatively short exposure to state centralization in terms of location, but once ancestry-adjusted it features a longer familiarity with state centralization, since the current inhabitants of the United States are mostly descended from Eurasian populations that have had a long history of centralized state institutions.15 Clearly, in 15 Germany and Italy, two countries from which many ancestors of current Americans originate, have ‡uctuated over their histories between fractured and uni…ed states. For instance, Italy was a uni…ed country under the Ro- man Empire, but a collection of city-states and local polities, partly under foreign control, prior to its uni…ca- tion in 1861. The index of state antiquity for such cases discounts periods that occurred in the distant past (see http://www.econ.brown.edu/fac/louis_putterman/Antiquity%20data%20page.htm for details on the computuation of the index). Due to lengthy periods of uni…cation or control of substantial parts of their territories by domestic regional states (such as the Republic of Venice and Prussia), however, Italy and Germany do not display state antiq- uity indices that are that di¤erent from other European countries. The United States overall has a state antiquity index roughly commensurate with that of European countries, despite the addition of populations, for instance de- scended from Native Americans or African slaves, that may have had limited exposure to centralized states. While the 13 this work the New World plays a big role in identifying the di¤erence in the coe¢ cients between historical factors and their ancestry-adjusted counterparts, because outside the New World, every- one’s ancestry is largely from their own location. Putterman and Weil explore how their two historical variables - each either ancestry adjusted or not - a¤ect the level of income per capita and within-country income inequality in the world today. Their key …nding is that it is not as much the past history of locations that matters as it is the history of the ancestor populations. Tables 4 and 5 illustrate their approach in our uni…ed empirical framework. Table 4 starts with simple correlations. The correlations between state history and years of agriculture, on the one hand, and per capita income in 2005, on the other hand, are of the expected positive signs, but are much larger when ancestry-adjusting - almost doubling in magnitude. These results are con…rmed in the regressions of Table 5. In these regressions, we start from the speci…cation that controls for the baseline set of four geographic variables, and add the Putterman and Weil variables one by one, either ancestry-adjusted or not. The variables representing the history of the locations enter with an insigni…cant coe¢ cient (columns 1 and 3), while the ancestry-adjusted variables enter with positive, statistically signi…cant coe¢ cients (columns 2 and 4). A one standard deviation change in ancestry-adjusted years of agriculture can account for 17% of a standard deviation of log per capita income, while the corresponding …gure is almost 22% for state history. To summarize, a long history of centralized states as well as an early adoption of agriculture are positively associated with per capita income today, after ancestry adjustment.16 Putterman and Weil also …nd that the variance of early development history across ancestor populations predicts within-country income inequality better than simple measures of ethnic and linguistic heterogeneity. For example, in Latin America, countries that are made up of a lot of Europeans along with a lot of native Americans tend to display higher income inequality than countries which are made up mostly of European descendants. Finally, to further elucidate why correcting for ancestry matters, measurement of state antiquity can be questioned on several grounds, there is little doubt that ancestry adjustment implies that the United States had a longer experience with centralized states than the history of Native Americans would suggest. 16 Interestingly, Paik (2010) documents that within Europe, an earlier onset of agriculture is negatively correlated with subsequent economic performance after the Industrial Revolution, contrary to the worldwide results of Putterman and Weil. Paik argues that the mechanism is cultural: a late adoption of agriculture is associated with individualist values that were conducive to economic success in the Industrial era. 14 they also show that a variable capturing the extent of European ancestry accounts for 41% of the variation in per capita income, a topic to which we turn in the next subsection. Putterman and Weil’s results strongly suggest that the ultimate drivers of development can- not be fully disembodied from characteristics of human populations. When migrating to the New World, populations brought with them traits that carried the seeds of their economic performance. This stands in contrast to views emphasizing the direct e¤ects of geography or the direct e¤ects of institutions, for both of these characteristics could, in principle, operate irrespective of the popula- tion to which they apply. A population’s long familiarity with certain types of institutions, human capital, norms of behavior or more broadly culture seems important to account for comparative development. 3.2 The Role of Europeans William Easterly and Ross Levine (2009) con…rm and expand upon Putterman and Weil’s …nding, showing that a large population of European ancestry confers a strong advantage in development, using new data on European settlement during colonization and its historical determinants. They …nd that the share of the European population in colonial times has a large and signi…cant impact on income per capita today, even when eliminating Neo-European countries and restricting the sample to countries where the European share is less than 15% - that is, in non-settler colonies, with crops and germs associated with bad institutions. The e¤ect remains high and signi…cant when controlling for the quality of institutions, while it weakens when controlling for measures of education. Table 6 captures the essence of these results. Still controlling for our four baseline geographic variables, we introduce the share of Europeans (computed from the Putterman and Weil ancestry matrix) in a regression explaining log per capita income in 2005. The e¤ect is large and statistically signi…cant (column 1), and remains signi…cant when con…ning attention to a sample of countries with fewer than 30% of Europeans. Introducing the Putterman and Weil ancestry-adjusted his- torical variables (columns 3 and 4), we …nd that years of agriculture and state history remain signi…cant after controlling for the share of Europeans, suggesting that historical factors have an e¤ect on contemporary development over and beyond the e¤ect of European ancestry. In other words, while the traits characterizing European populations are correlated with development, the historical legacy of state centralization and early agricultural adoption matters independently. 15  Easterly and Levine (2009) interpret these …ndings as consistent with the human-capital ar- gument by Glaeser, La Porta, Lopez-de-Silanes and Shleifer (2004) that Europeans brought their human capital, and the Galor and Weil (2000) and Galor, Moav and Vollrath (2009) emphasis on the role of human capital in long-run development. However, Easterly and Levine (2009, p. 27) also write: "Of course, there are many other things that Europeans carried with them besides general education, scienti…c and technological knowledge, access to international markets, and hu- man capital creating institutions. They also brought ideologies, values, social norms, and so on. It is di¢ cult for us to evaluate which of these were crucial either alone or in combination." This exempli…es the di¢ cult issue of disentangling, with the imperfect data that must be used to study comparative development, the e¤ects of di¤erent human characteristics. The bottom line, however, is that human traits are important to account for comparative development patterns, quite apart from the e¤ects of geographic and institutional factors. 3.3 The Persistence of Technological Advantages The deep historical roots of development are at the center of Comin, Easterly and Gong (2010). They consider the adoption rates of various basic technologies in 1000 BC, 1 AD, and 1500 AD. in a cross-section of countries de…ned by their current boundaries. They …nd that technology adoption in 1500, but also as far back as 1000 BC, is a signi…cant predictor of income per capita and technology adoption today. The e¤ects of past technology continue to hold when including continental dummies and other geographic controls. At the level of technologies, then, when examining a worldwide sample of countries (including European countries), there is no evidence of a reversal of fortune. Interestingly, Comin, Easterly and Gong (2010) also …nd that the e¤ects of past technologi- cal adoption on current technological sophistication are much stronger when considering the past history of technology adoption of the ancestors of current populations, rather than technology adoption in current locations, using the migration matrix provided in Putterman and Weil (2010). Hence, Comin, Easterly and Gong’s results provide a message analogous to Putterman and Weil’s: earlier historical development matters, and the mechanism is not through locations, but through ancestors - that is, intergenerational transmission. The basic lesson from Putterman and Weil (2010), Easterly and Levine (2009), and Comin, Easterly and Gong (2010) is that historical factors - experience with settled agriculture and with former political institutions, and past exposure to frontier technologies - predict current income 16 per capita and income distribution within countries, and that these factors become more important when considering the history of populations rather than locations. These contributions point to a key role for persistent traits transmitted across generations within populations in explaining development outcomes over the very long run. 3.4 Genetic Distance and Development Genealogical links among populations over time and space are at the center of Spolaore and Wacziarg (2009), where we emphasized intergenerationally transmitted human traits as important determinants of development. The main goal of this paper was to explore the pattern of di¤usion of economic development since the onset of the Industrial Revolution in Northwestern Europe in the late 18th century and early 19th century. The idea is to identify barriers to the adoption of these new modes of production, with a speci…c focus on human barriers (while controlling for geo- graphic barriers). The bottom line is, again, that human traits matter, but the paper emphasizes barrier e¤ects stemming from di¤erences in characteristics, rather than the direct e¤ect of human characteristics on economic performance. We compiled a data set, based on work by L. Luca Cavalli-Sforza, Paolo Menozzi and Alberto Piazza (1994), providing measures of genetic distance between pairs of countries, using informa- tion about each population’s ancestral composition.17 Genetic distance is a summary measure of di¤erences in allele frequencies between populations across a range of neutral genes (chromosomal loci). The measure we used, FST genetic distance, captures the length of time since two populations became separated from each other. When two populations split apart, random genetic mutations result in genetic di¤erentiation over time. The longer the separation time, the greater the genetic distance computed from a set of neutral genes. Therefore, genetic distance captures the time since two populations have shared common ancestors (the time since they were parts of the same popu- lation), and can be viewed as a summary measure of relatedness between populations. An intuitive 17 To accommodate the fact that some countries are composed of di¤erent genetic groups (e.g. the USA), we computed a measure of "weighted genetic distance," representing the expected genetic distance between two randomly chosen individuals, one from each country, using the genetic distances associated with their respective ancestor populations. That is, we do not consider the inhabitants of countries composed of di¤erent genetic groups as a new homogeneous "population" in the biological sense, but treat each of those countries as formed by distinct populations, to accurately capture the di¤erences in ancestor-transmitted traits within and across countries. This is the measure used in the empirical work discussed below. 17 analogue is the concept of relatedness between individuals: two siblings are more closely related than two cousins because they share more recent common ancestors - their parents rather than their grandparents. Figure 1 (from Cavalli-Sforza, Menozzi and Piazza, 1994, p. 78) is a phylogenetic tree illustrating how di¤erent human populations have split apart over time. Such phylogenetic trees, constructed from genetic distance data, are the population analogs of family trees for individuals. In this tree, the greatest genetic distance observed is between Mbuti Pygmies and Papua New Guineans, where the FST distance is 0:4573, and the smallest is between the Danish and the English, where the genetic distance is 0:0021.18 To properly interpret the e¤ect of genetic distance on di¤erences in economic outcomes, two important clari…cations are in order. First, since genetic distance is based on neutral change, it is not meant to capture di¤erences in speci…c genetic traits that can directly matter for survival and …tness. Hence, we emphasize that empirical work using genetic distance provides no evidence for an e¤ect of speci…c genes on income or productivity. Evidence of an "e¤ect of genetic dis- tance" is not evidence of a "genetic e¤ect." Rather, it can serve as evidence for the importance of intergenerationally-transmitted traits, including traits that are transmitted culturally from one generation to the next. Second, the mechanism need not be a direct e¤ect of those traits (whether culturally or geneti- cally transmitted) on income and productivity. Rather, divergence in human traits, habits, norms, etc. have created barriers to communication and imitation across societies. While it is possible that intergenerationally transmitted traits have direct e¤ects on productivity and economic performance (for example, if some parents transmit a stronger work ethic to their children), another possibility is that human traits also act to hinder development through a barrier e¤ect: more closely related societies are more likely to learn from each other and adopt each other’s innovations. It is easier for someone to learn from a sibling than from a cousin, and easier to learn from a cousin than from a stranger. Populations that share more recent common ancestors have had less time to diverge in a wide range of traits and characteristics - many of them cultural rather than biological - that 18 Among the more disaggregated data for Europe, also used in Spolaore and Wacziarg (2009), the smallest genetic distance (equal to 0.0009) is between the Dutch and the Danish, and the largest (equal to 0.0667) is between the Lapps and the Sardinians. The mean genetic distance across European populations is 0.013. Genetic distances are roughly ten times smaller on average across populations of Europe than in the world data set. 18 are transmitted from a generation to the next with variation. Similarity in such traits facilitates communication and learning, and hence the di¤usion and adaptation of complex technological and institutional innovations. Under this barriers interpretation, di¤erences in traits across populations hinder the ‡ow of technologies, goods and people, and in turn these barriers hurt development. For instance, histor- ically rooted di¤erences may generate mistrust, miscommunication, and even racial or ethnic bias and discrimination, hindering interactions between populations that could result in a quicker di¤u- sion of productivity-enhancing innovations from the technological frontier to the rest of the world. The barriers framework in Spolaore and Wacziarg (2009) predicts that, ultimately, genetic distance should have no residual e¤ect on income di¤erences (unless another major innovation occurs), as more and more societies, farther from the frontier, come to imitate the frontier technology. This is consistent with the di¤usion of economic development as emerging from the formation of a human web, gradually joined by di¤erent cultures and societies in function of their relative distance from the technological frontier (McNeill and McNeill, 2003). We test the idea that genealogical relatedness facilitates the di¤usion of development in our uni…ed empirical framework. Table 7, columns 1 and 2 introduce genetic distance to the USA in our basic income level regression, controlling for the baseline geographic variables.19 Genetic distance as of 1500, re‡ecting the distance between indigenous populations, is negatively and signi…cantly related to log income per capita in 2005. The e¤ect rises in magnitude when considering genetic distance to the USA using the current genetic composition of countries. In other words, ancestry- adjusted genetic distance once more is a better predictor of current income than a variable based on indigenous characteristics, consistent with the results in Table 5. Column 3 of Table 7 introduces genetic distance alongside the share of Europeans, showing that genetic distance to the USA bears a signi…cant partial correlation with current income that is not entirely attributable to the presence of Europeans. 19 Since several countries in our sample, especially the technological frontier (the United States) are composed of several distinct genetic groups, we used a weighted measure of genetic distance, capturing the expected genetic distance between two individuals, randomly selected from each of the two countries in a pair. Formally, the weighted FST genetic distance between countries 1 and 2 is de…ned as: W X I X J F ST12 = (s1i s2j dij ) i=1 j=1 where ski is the share of group i in country k, dij is the FST genetic distance between groups i and j. 19  While these simple regressions are informative, a better test of the hypothesis that genetic distance captures human barriers to the di¤usion of development relies on a bilateral approach, whereby absolute log income di¤erences are regressed on bilateral genetic distance, analogous to a gravity approach in international trade. This was the main approach in Spolaore and Wacziarg (2009), and is re‡ected in Tables 8 and 9. The bilateral approach o¤ers a test of the barriers story: if genetic distance acts as a barrier, it should not be the simple distance between countries that matters, but their genetic distance relative to the world technological frontier. In other words if genetic distance acts as a barrier, it should not be the genetic distance between, say, Ecuador and Brazil that should better explain their income di¤erence, but their relative genetic distance to the United States, de…ned as the absolute di¤erence between the Ecuador-USA genetic distance and the Brazil-USA genetic distance. The speci…cations we use are as follows: First, we estimate the e¤ect of simple weighted genetic distance, denoted F STijW , between country i and country j, on the absolute di¤erence in log per capita income between the two countries, controlling for a vector Xij of additional bilateral variables of a geographic nature: W 0 jlog Yi log Yj j = 0 + 1 F STij + 2 Xij + "ij (1) Second, we estimate the same speci…cation, but using as a regressor relative genetic distance rather than simple genetic distance. Genetic distance relative to the frontier (the United States) between countries i and j is de…ned as: F STijR = jF STi;U W S W j: F STj;U S R 0 jlog Yi log Yj j = 0 + 1 F STij + 3 Xij + ij (2) Third, we conduct a horserace between F STijW and F STijR : R W 0 jlog Yi log Yj j = 0 + 1 F STij + 2 F STij + 3 Xij + ij (3) The prediction of the barrier model is that the e¤ect of F STijR should be larger in magnitude than the e¤ect of F STijW ( 1 > 1 ), and that F STijR should "win out" in a horserace ( 1 > 2 ). Consistent with this prediction, in Table 8, columns 1 and 2 show that relative genetic distance enters with a larger magnitude than simple genetic distance, and column 3 demonstrates that, when both measures are entered together, relative genetic distance trumps simple genetic distance. The magnitude of the e¤ect is substantial, with a one standard deviation increase in genetic distance 20 increasing economic distance by between 25% and 30% of a standard deviation in the absolute di¤erence in log per capita income, depending on the speci…cation. Column 4 shows once again that the e¤ect of genetic distance is robust to including the absolute di¤erence in the share of Europeans. In other words, genetic distance accounts for comparative development over and above the role played by the historical advantage of European populations. Finally, column 5 attempts to control for the possible endogeneity of post-1500 migrations (as well as possible measurement error in contemporary genetic distance) by using genetic distance in 1500 as an instrument for contemporary genetic distance. The magnitude of the beta coe¢ cient increases to 45%. Table 9 examines the relationship between genetic and economic distances through history, providing further evidence for the barriers interpretation. Here, we consider the relative genetic distance to the English population in a sample going back to 1820, using Maddison’s data on per capita income. We continue to control for a large number of measures of geographic distance, climatic di¤erences and transportation costs. Since the availability of data changes through time, we report both standardized betas for the full samples and for the sample common to all dates. Focusing on the latter for comparability across time, we see that the magnitude of the e¤ect of ge- netic distance is maximal in 1870, in the wake of the Industrial Revolution. The e¤ect then declines steadily from the peak of 16% in 1870 to 7:8% in 2005. This pattern provides a further suggestive test of the barriers model: in the wake of a big innovation occurring in Northwestern Europe, relative genetic distance to the frontier strongly predicts income di¤erences, but as more and more countries join the ranks of industrialized countries, the e¤ect declines. As already mentioned, this is consistent with a barrier model, which predicts that, ultimately, unless another major innovation occurs, relative genetic distance should have no residual e¤ect on income di¤erences, as more and more societies, increasingly distant from the frontier society, imitate the frontier technology. Thus, these …ndings are consistent with our interpretation of genetic distance as capturing barriers to the long-term di¤usion of development. What traits are captured by genetic distance? By its very de…nition, genetic distance is a mea- sure of genealogical relatedness between human populations. It is important to stress again that while e¤ects of genetic distance point to the importance of intergenerational links, they are not ev- idence of direct e¤ects of speci…c genes or genetically transmitted traits on income or productivity. Rather, genetic distance captures genealogical relations between populations, and hence di¤erences in traits that are transmitted vertically from one generation to the next through a variety of mech- 21 anisms, biologically but also culturally, as well as through the interactions of the two inheritance systems (gene-culture co-evolution). We detail these di¤erent mechanisms in the next section. 4 The Intergenerational Transmission of Development 4.1 Mechanisms of Intergenerational Transmission The empirical literature on geography and the reversal of fortune and the more recent contributions on the role of ancestor populations suggest that, while there is signi…cant persistence in develop- ment, this persistence is a characteristic of human populations and not of geographic locations. The work discussed so far points to a key role for traits transmitted from one generation to the next within populations over the long run. In this section we provide a general taxonomy of the di¤erent channels and mechanisms through which these traits can be transmitted across generations and a¤ect economic development. We then use this framework to discuss recent contributions to the economics literature. The starting point for our discussion is a classi…cation of di¤erent mechanisms of intergenera- tional transmission. The more recent literature on heredity and evolution stresses that inheritance mechanisms are diverse and cannot be reduced to the old nature vs. nurture dichotomy. On the contrary, people and societies inherit traits from their ancestors through a complex interaction of biological and cultural mechanisms, with an essential role played by environmental factors. Follow- ing Eva Jablonka and Marion J. Lamb (2005), we consider four inheritance dimensions: genetic, epigenetic, behavioral, and symbolic. For convenience and to keep the taxonomy relatively simple, we refer to the …rst two dimensions (genetic and epigenetic) as "biological" and the remaining two dimensions (behavioral and symbolic) as "cultural." The genetic dimension has its molecular basis in DNA and its replication. Modern genetics stresses that the genome is a complex and dynamic system, and that genes alone do not determine individual characteristics. For example, Jablonka and Lamb (2005, p. 7) write: "The stretch of DNA that is a ‘gene’has meaning only within the system as a whole. And because the e¤ect of a gene depends on its context, very often a change in a single gene does not have a consistent e¤ect on the trait that it in‡uences." In general, it is useful to view genetic transmission as part of a broader system, interacting with other factors, including our second dimension: epigenetics. The epigenetic dimension, even though important in biology and medicine, is not as well- 22 known among non-specialists. Epigenetic inheritance systems refer to the ways cells with identical genetic information can acquire di¤erent phenotypes and transmit them to their daughter cells through the inheritance of epigenetic markers (for example, methylation patterns). The epigenetic dimension is vital for the biological development of individual organisms. In addition, some scholars argue that it may play an important role in evolution over time. An insightful and entertaining discussion of this view is provided in Jablonka and Lamb (2005, chapter 4). As they explain (p. 113), "A person’s liver cells, skin cells, and kidney cells, look di¤erent, behave di¤erently, and function di¤erently, yet they all contain the same genetic information. With very few exceptions, the di¤erences between specialized cells are epigenetic, not genetic. They are the consequences of events that occurred during the developmental history of each type of cell and determined which genes are turned on, and how their products act and interact (...) Although their DNA sequences remain unchanged during development, cells nevertheless acquire information that they can pass to their progeny. This information is transmitted through what are known as epigenetic inheritance systems." Even though the extent and relevance of heritable epigenetic e¤ects is debated in the biological literature, there is increasing evidence that changes in the epigenome can be inherited across generations ("paramutation"). Reviews of the evidence are provided, for instance, in Vicki Chandler and Mary Alleman (2008) and, for humans, in Daniel K. Morgan and Emma Whitelaw (2008).20 This mechanism could provide an explanation for rapid changes in populations that could not be brought about by genetic selection. In general, it is conceptually appropriate to consider both genetic and epigenetic transmission, and their interactions, when discussing biological inheritance systems.21 Clearly, humans do not inherit traits only biologically from their ancestors. Typically, people ac- quire all sorts of traits through cultural transmission (for example their mother tongue and all kinds of views and beliefs about the world). We can distinguish between behavioral transmission, which is 20 Morgan and Whitelaw (2008) mention the well-known Dutch Famine Birth Cohort Study (Lambert H. Lumey, 1992), which reported that children born during famine in World War II were smaller than average and that the e¤ects could last two generations. However, they also point out that a subsequent report by Aryeh D. Stein and Lambert H. Lumey (2002) failed to reproduce some of the …ndings. 21 In a di¤erent context, the relevance of epigenetics for the study of economic outcomes has been emphasized in the microeconomic literature on human capital formation. For instance, Flavio Cunha and James Heckman (2007, p. 32), write: "the nature versus nurture distinction is obsolete. The modern literature on epigenetic expression teaches us that the sharp distinction between acquired skills and ability featured in the early human capital literature is not tenable." 23 also common among some animals (for instance, monkeys), and symbolic transmission, which some scholars view as uniquely human - the philosopher Ernst Cassirer (1944) famously de…ned man as "the symbolic animal." Both mechanisms involve social learning. Behavioral transmission refers to learning by direct observation and imitation: I learned how to cook spaghetti by watching my dad in the kitchen. In contrast, symbolic transmission allows learning via systems of symbols: I learned how to cook goulash from a cook book. Human norms, habits, values, etc. tend to be passed across generations both behaviorally (by example) and symbolically (using language, art, writing and so on). In general, culture can be de…ned as "information capable of a¤ecting individuals’ behavior that they acquire from other members of their species through teaching, imitation, and other forms of social transmission." (Peter J. Richerson and Robert Boyd, 2005, p. 5).22 In their insightful discussion of the "evolution of cultural evolution," Joseph Henrich and Richard McElreath (2003, p. 123) write: "While a variety of local genetic adaptations exist within our species, it seems certain that the same basic genetic endowment produces arctic foraging, tropical horticulture, and desert pastoralism [...]. The behavioral adaptations that explain the immense success of our species are cultural in the sense that they are transmitted among individuals by social learning and have accumulated over generations. Understanding how and when such culturally evolved adaptations arise requires understanding of both the evolution of the psychological mechanisms that underlie human social learning and the evolutionary (population) dynamics of cultural systems."23 While it is conceptually useful to distinguish between biological and cultural transmission, we must keep in mind that in reality those dimensions are interconnected in complex ways. An increasingly in‡uential literature within population genetics has emphasized that human outcomes often stem from the interaction of biological and cultural factors. Both genes and culture are informational entities that are transmitted at di¤erent rates across generations with variations, and can be studied within a uni…ed framework that focuses on the interaction between biological and cultural inheritance systems. This approach is known as dual inheritance theory or gene- culture coevolution (Cavalli-Sforza and Feldman, 1976, 1981; Boyd and Richerson, 1985; Richerson and Boyd, 2005). In such a framework, individual outcomes (phenotypes) are a mix of genetically 22 Of course, this is one among many possible de…nitions of culture. It is well known that the word "culture" has multiple meanings. In a widely cited study, Alfred L. Kroeber and Clyde Kluckhohn (1952) provided 164 de…nitions of culture. 23 An interesting example of analysis of cultural evolution with long-term implications for economic development is Shari¤, Norenzayan and Henrich (2009). 24 and culturally transmitted traits, a¤ecting the transmission rates of di¤erent genetic and cultural information. As Richerson and Boyd (2005, p. 194) point out, genes and culture can be seen as "obligate mutualists, like two species that synergistically combine their specialized capacities to do things that neither can do alone. [... ] Genes, by themselves can’t readily adapt to rapidly changing environments. Cultural variants, by themselves, can’t do anything without brains and bodies. Genes and culture are tightly coupled but subject to evolutionary forces that tug behavior in di¤erent directions." Proponents of dual inheritance theory believe that gene-culture evolution has played an impor- tant role in the evolution of human social psychology, including the evolution of social norms and institutions (e.g., Richerson and Boyd, chapter 6). This view informs a broad literature on the co- evolution of preferences, institutions and behavior, such as the analyses of the evolution of altruistic behavior by Boyd, Herbert Gintis, Samuel Bowles, and Richerson (2003) and Gintis, Bowles, Boyd and Ernst Fehr (2003). General discussions of the emergence of prosperity-generating behavior from an evolutionary perspective are provided by Paul Seabright (2010) and Matt Ridley (2010). A famous example of gene-culture coevolution is the evolution of adult tolerance for milk in some, but not all, human populations (Fredrick J. Simoons, 1969, 1970).24 Most people, like most other mammals, can digest milk as infant but not as adults, because they lack the enzyme to digest lactose. However, there are several populations where most adults can indeed digest milk. The largest concentration of lactose absorbers can be found in Northwestern Europe, where less than 10 15% of the population is lactose intolerant. Low levels of lactose intolerance are also found among Indians and some African populations (Tutsi and Fulani). In contrast, few Far Easterners, Bantu Africans, Paci…c Islanders and Native American can digest milk as adults. It is now well-understood that the adult ability to digest milk evolved in response to a cultural innovation: dairying. The (dominant) gene controlling lactose absorption spread rapidly among populations that kept cows, sheep or goats, making those practices even more valuable from an evolutionary perspective. It also spread, but to a lesser extent, among Mediterranean people who consume milk in the form of cheese and yogurt, from which the lactose has been removed, but it did not spread among populations without a dairying tradition. To predict whether a current population would have a high or low tolerance for milk, one must look at the history of dairying among the population’s ancestors, no matter where they lived, rather than to the history of 24 See also William H. Durham (1991, chapter 5) and Richerson and Boyd (2005, chapter 6). 25 dairying in that population’s current territory. For instance, within the United States, it has been observed that the percentage of lactose intolerant adults is almost 100% among Native Americans, 90% among Asian Americans, 75% among African Americans, and only 12% among European Americans (Norman Kretchmer, 1972; Nabil Sabri Enattah et al., 2002). This is consistent with the intergenerational transmission of the lactose absorption trait over an extended historical span, through genetic and cultural interaction.25 In general, the economic e¤ects of human characteristics are likely to result from interactions of biological and cultural factors, with the e¤ects of genetic or epigenetic characteristics on eco- nomic outcomes changing over space and time depending on cultural characteristics, and vice versa. Consider, for example, di¤erences across individuals within a given population (say, the US) with respect to a clearly genetic feature, such as having two X chromosomes, the purely genetic charac- teristic associated with the female gender. This characteristic is likely to have had very di¤erent e¤ects on a person’s income, life expectancy and other outcomes in the year 1900 and in the year 2000, because of changes in culturally transmitted characteristics over the century. This is a case where the impact of genes on outcomes varies with a change in cultural characteristics.26 Con- versely, we can think of the di¤erential impact of a given cultural characteristic –for example, the habit of drinking alcohol –on individuals with di¤erent genetic traits, such as variation in alcohol dehydrogenase, the alcohol-metabolizing enzyme. In sum, we can aggregate inheritance dimensions and their complex interactions in three broad sets: (a) biological transmission (genetic and epigenetic, and their interaction), (b) cultural trans- mission (behavioral and symbolic, and their interaction), and (c) dual, capturing the interactions between biological and cultural transmission. 4.2 Direct and Barrier E¤ects When considering how traits transmitted along these di¤erent channels may have a¤ected develop- ment, we must introduce an additional distinction, orthogonal to the categories discussed up to this point. No matter how traits are transmitted, their e¤ects on economic performance might operate 25 More examples of gene-culture interaction, and a taxonomy, are provided by William H. Durham (1991). 26 This is a variation on an example by Alison Gopnik in her comment to debate between Steven Pinker and Eliza- beth Spelke at http://www.edge.org/discourse/science-gender.html#ag. Steven Pinker’s response is also available at http://www.edge.org/discourse/sciencegender.html. 26 either directly or as barriers to the di¤usion of development. One possibility is that intergenerationally transmitted traits have direct e¤ects on productiv- ity and economic performance. A slow-changing cultural trait developed in early history could be conducive directly to high incomes in modern times if it is transmitted from parents to kids within populations, either behaviorally or through complex symbolic systems (e.g., by religious teaching). For example, a direct e¤ect stemming from cultural transmission would be Max Weber’s (1905) argument that the Protestant ethic was a causal factor in industrialization (a recent criti- cal reassessment of this hypothesis has been provided by Sascha Becker and Ludwig Woessmann, 2009). As we discussed in the previous section, another possibility is that human traits act to hinder development through a barrier e¤ect. In this case, it is not the trait itself that directly a¤ects economic performance. Rather, it is di¤erences in inherited characteristics across populations that create barriers to the ‡ow of technological and institutional innovations, ideas, etc., and, conse- quently, hurt development. Historically rooted di¤erences may generate barriers - e.g., via cultural, racial and ethnic bias, discrimination, mistrust, and miscommunication - hindering interactions between populations that could result in a quicker di¤usion of productivity-enhancing innovations across populations, as in Spolaore and Wacziarg (2009). A focus on barriers can explain why dif- ferences in inherited traits may matter, even though many new ideas and innovations are learned "horizontally," from individuals and populations that are not directly related, rather than "verti- cally," from one’s close relatives and ancestors. The fact is that, when barrier e¤ects exist, vertically transmitted traits also a¤ect horizontal learning and di¤usion. People are more likely to learn new ideas and adopt new technologies from other people who, while not directly related to them, share more recent common ancestors and, consequently, on average, a larger set of inherited traits and characteristics. The microeconomic literature on the di¤usion of innovations (Everett M. Rogers, 2003) is con- sistent with a major role for subjective barriers between groups and populations. As Rogers points out, summarizing the lessons from decades of research, most people depend upon a subjective eval- uation of an innovation that is conveyed to them from other individuals like themselves, who have previously adopted the innovation. This dependence on the experience of near peers suggests that, at the heart of the di¤usion process, we can often …nd potential adopters’imitation of their network partners. 27 4.3 A General Taxonomy The distinction between barrier e¤ects and direct e¤ects on the one hand, and the di¤erent forms of intergenerational transmission on the other hand, can be conveniently captured in the following matrix: Direct E¤ect Barrier E¤ect Biological Transmission Quadrant I Quadrant IV (genetic and/or epigenetic) Cultural Transmission Quadrant II Quadrant V (behavioral and/or symbolic) Dual Transmission Quadrant III Quadrant VI (biological-cultural interaction) In what follows we provide examples of each of these quadrants from contributions to the social sciences, with a particular focus on research that stresses the role of intergenerationally transmitted traits on economic development. As we will see, mechanisms from all six of the quadrants in the …gure above have been studied in the economic literature. However, so far there have been no systematic attempt to quantify the respective power of each of the major explanations, nor has there been enough work on the precise traits that can account for either direct or barrier e¤ects on development. These are the major avenues for promising future research in this emerging …eld. In light of the advances in the scienti…c literature that we have mentioned above, which have emphasized the interconnection and coevolution of biological and cultural traits, it may be very hard, or even meaningless, to separate biological and cultural e¤ects in the long run. Therefore, while the rows in our taxonomy matrix provide useful ideal benchmarks from a theoretical perspec- tive, a more productive empirical approach might be to focus on whether such intergenerational mechanisms - whether biological or cultural (or dual) - operate directly or as barriers to the dif- fusion of technological and institutional innovations - i.e. to identify the respective roles of each column rather than each row in the matrix that summarizes our proposed taxonomy. We return to these points below, as we discuss speci…c contributions to the economic literature. 4.3.1 Quadrants I, II and III A theory of development centered on a direct e¤ect of biologically transmitted characteristics (Quad- rant I) is provided in Oded Galor and Omer Moav (2002). These authors suggest that there may exist an inter-geenrationally transmitted trait a¤ecting humans’fertility strategies, with some indi- 28 viduals inheriting traits that induce them to follow a quantity-biased strategy (having a high number of children), while others would lean towards a quality-biased strategies (high parental investment in a smaller number of children).27 Galor and Moav (2002) then argue that the evolutionary dy- namics of these traits had important implications for the onset of the Industrial Revolution and the following demographic transition. Their starting point is the pre-industrial world, where everybody was caught in a Malthusian trap: technological improvements just led to larger populations but not an increase in income per capita and standards of living.28 In such a world, a positive shock to productivity is associated with an expansion of the population, and hence selective pressure in favor of productivity-enhancing traits, such as a focus on parental investment. As the genetic predilec- tion to having fewer children spread as a result of these selective pressures, a transition out of the Malthusian regime endogenously occurred. In sum, Galor and Moav (2002) provide a theoretical argument for a direct e¤ect of intergenerationally transmitted traits on the onset of the Industrial Revolution. While the focus of their paper is on genetic transmission, Galor (2005, p. 250) also points out that "the theory is applicable for either social or genetic intergenerational transmission of traits. A cultural transmission is likely to be more rapid and may govern some of the observed di¤erences in fertility rates across regions." With this broader interpretation, the theory in Galor and Moav (2002) spans Quadrants I, II and III. Another contribution that allows for a direct e¤ect of inherited characteristics on development is Gregory Clark’s (2007) book on the advent of the Industrial Revolution in England. As in Galor and Moav (2002), Clark’s starting point is the pre-industrial world caught in a Malthusian trap. In such a world, unlike in modern industrialized societies, economic success translated into reproductive success: the richer individuals had more surviving children, while the poorer individuals had so few surviving children that their families were often dying out. Therefore, over time the children of the richest individuals tended to replace the children of the poorest. Through this form of selection, traits that would "ensure later economic dynamism," such as patience, hard work, and innovativeness, spread "biologically throughout the population" (Clark, 2007, p.8). In Clark’s view, this long-term process worked through all pre-industrial agrarian societies caught in a Malthusian trap, but was especially powerful in England because of "accidents of institutional stability and 27 The authors are careful to notice that such di¤erences do not imply that one strategy is "better" than the other, in the same way as we cannot say that giant sequoias (which follow an extreme quality-biased strategy) are "better" than humans. 28 Empirical evidence on these e¤ects is provided by Ashraf and Galor (2011) - see our discussion in Section 2 above. 29 demography: in particular the extraordinary stability of England back to at least 1200, the low growth of English population between 1300 and 1760, and the extraordinary fecundity of the rich and economically successful" (Clark, 2007, p. 11). According to Clark, this process played an essential role in allowing England to break out of the Malthusian trap after 1800. Clark’s argument, which is based on a detailed analysis of the historical record, is related to Galor and Moav’s theoretical contribution, insofar as selection of traits in the Malthusian era sets the stage for future intensive economic growth. However, the speci…c inherited traits stressed by Clark are di¤erent from those at the center of Galor and Moav’s story - Clark’s focus is mainly on attitudes towards work.29 Clark tends to emphasize biological mechanisms, but he does not take a de…nitive stance as to whether the human traits that caused the English to experience the Industrial Revolution were primarily biological or cultural. Hence, in a more general sense, Clark’s suggested mechanisms also belong to Quadrants I, II and III. A di¤erent set of channels through which genetic forces may a¤ect economic development is explored by Ashraf and Galor (2010). In that study the authors focus on genetic diversity within populations. Genetic diversity is a di¤erent concept from genetic distance, as it was used in Spolaore and Wacziarg (2009). Genetic distance refers to genetic di¤erences between populations, while genetic diversity is de…ned in terms of heterogeneity within populations. Because of the serial- founder e¤ect, genetic diversity tended to decline as human populations moved from the ancestral lands where Homo Sapiens originally emerged (East Africa) to the rest of the World (Eurasia, Oceania and the Americas).30 Hence, genetic diversity is highest among African populations and lowest among Amerindian populations. Ashraf and Galor (2010) show that genetic diversity bears a non-monotonic relationship with development outcomes between years 1 and 1500 AD. They argue 29 It is also worth noting that Clark, like Galor and Moav, is not making any claim that inherited traits made some populations generally "superior" to others: "This is not in any sense to say that people in settled agrarian economies on the eve of the Industrial Revolution had become “smarter” than their counterparts in hunter-gatherer societies.... For the average person the division of labor... made work simpler and more repetitive. The argument is instead that it rewarded with economic and hence reproductive success a certain repertoire of skills and dispositions that were very di¤erent from those of the pre-agrarian world, such as the ability to perform simple repetitive tasks hour after hour, day after day. There is nothing natural or harmonic, for example, in having a disposition to work even when all basic needs of survival have been achieved" (Clark, 2007, pp. 187-188). 30 As explained in Ashraf and Galor (2010), subsets of populations left to establish new settlements, they carried with them only part of the overall genetic diversity of their parental populations, therefore reducing the heterogeneity of populations that settled farther from the original cradle of humankind. 30 that this relation is causal, because of a trade-o¤ between the bene…cial and the detrimental e¤ects of diversity of traits on productivity. In their view, a high level of genetic heterogeneity within populations comes with both costs and bene…ts for economic development. Heterogeneity can be costly because it may reduce trust and coordination among individuals that are less similar and less closely related. However, genetic diversity also comes with advantages: "in an economy where the labor force is characterized by genetic heterogeneity in a wide array of traits, to the extent that some of these traits lead to specialization in task-oriented activities, higher diversity will increase productivity for society as a whole, given complementarities across di¤erent tasks" (Ashraf and Galor, 2010, p. 2 ft. 1). Given such con‡icting e¤ects of heterogeneity, Ashraf and Galor argue that intermediate levels of genetic diversity are most conducive to the accumulation of wealth. These can be interpreted as direct e¤ects of inherited traits working through the biological channel (Quadrant I), and possibly its interaction with the cultural channel (Quadrant III). While the focus of Ashraf and Galor (2010) is on genetic diversity and biological mechanisms, a similar trade-o¤ between costs and bene…ts could emerge with respect to diversity in culturally transmitted traits. Ashraf and Galor (2011b) explicitly focus on the costs and bene…ts of cultural diversity (Quadrant II) when providing a framework to understand the e¤ects of prehistoric measures of geographic isolation on economic development.31 While there are relatively few contributions in the social sciences that focus on biological trans- mission, there is a much larger literature on culturally-transmitted traits and development (Quad- rant II). Among recent contributions, Guido Tabellini (2008, 2010) argues that speci…c cultural traits, such as generalized trust and individualism, can account causally for variation in insti- tutional development across regions of Europe, and hence indirectly for variation in the level of economic development there. An extensive comparative and historical analysis of the role of cul- 31 Regarding the mechanisms in Quadrant I, there is also a literature in political science, discussed by James H. Fowler and Darren Schreiber (2008), arguing that direct biological e¤ects are relevant in the study of political attitudes and behavior. Fowler and Schreiber (2008, p. 914) write: "The new science of human nature demands that we recognize that genes are the institutions of the human body. [...] we cannot fully appreciate their function in humans without understanding their role in the very complex political and social interactions that characterize our species." Among the biological evidence cited in their survey is the involvement of neuroreceptors in speci…c political behavior, such as the link between the DRD2 gene, which codes for a dopamine receptor, and voter turnout (Christopher T. Dawes and J. H. Fowler, 2009). These claims are controversial among political scientists, and have been criticized, for example, by Evan Charney and William English (2012). For a recent contribution on the molecular-genetic-based heritability of economic and political preferences, see Daniel J. Benjamin et. al. (2012). 31 ture, institutions and development is provided in Avner Greif (1994, 2006), who o¤ers a uni…ed conceptual framework for analyzing the persistence of institutions, their endogenous change, and the impact of past institutions on subsequent development. As Greif (2006, chapter 1) points out, culturally transmitted traits, such as beliefs and norms, play a key role in determining which formal rules are followed and what is the actual economic impact of an institutional organization. Luigi Guiso, Paola Sapienza and Luigi Zingales (2008) explain the persistence of di¤erences in culturally- transmitted beliefs regarding collective action and cooperation across Italian cities by successfully testing Robert D. Putnam’s (1993) hypothesis that those di¤erences re‡ect the impact of historical variables on local civic values.32 Cultural transmission plays a key role in the analysis of divergent paths by China and Europe in Greif and Tabellini (2010), who argue that China’s and Europe’s distinct cultural and institutional trajectories during the last millennium re‡ect the impact of di¤erent initial moral systems and kinship organizations. According to Greif and Tabellini (2010): “The Chinese clan is a kinship- based hierarchical organization in which strong moral ties and reputation among clan’s members are particularly important in sustaining cooperation. In Medieval Europe, by contrast, the main example of a cooperative organization is the city. Here cooperation is across kinship lines and external enforcement plays a bigger role. But morality and reputation, although weaker, also matter and extend beyond one’s kin.”This analysis sheds light on why China and other advanced societies in East Asia have taken so long to catch up with the Industrial Revolution, in spite of positive historical preconditions and a signi…cant technological lead in early periods, as documented in the aforementioned studies by Comin, Easterly and Gong (2010) and Putterman and Weil, (2010).33 Direct cultural mechanisms are also at the heart of the already cited study by Ashraf and Galor (2011b), who argue that the interplay between the forces of cultural assimilation and cultural di¤u- sion contributed to the long-term patterns of relative development in Europe and Asia. According to these authors, cultural rigidity was an advantage in earlier phases of development but a hindrance 32 Evidence on the long-term persistence of pernicious cultural traits is provided by Nico Voigtländer and Hans- Joachim Voth (2011), who use data on anti-Semitism in Germany and …nd continuity at the local level over six centuries: anti-Semitic pogroms during the Black Death in 1348-50 are a strong and robust predictor of violence against Jews in the 1920s and of the vote share of the Nazi Party. 33 This is an issue that Jared Diamond also had to face in his book (1997, chapter 16). Classic references on this topic are Rosenberg and Birdzell (1987) and Mokyr (1990). 32 at later stages.34 Along similar lines, Yuriy Gorodnichenko and Gérard Roland (2011) study the interplay between cultu