The spillover of Africa’s slave trades on the economic performances of the receiving countries is the theme developed by Engerman and Sokoloff (1997, 2002), on the basis of historical evidence they provide for the New World. Their influencial hypothesis can be summarized as follows. They first argue that factor endowments (climate, soil, crops, etc.) determined the suitability and hence the adoption of plantation slavery. In turn the use of slave labor caused extreme economic inequality which shaped the evolution of local institutions (including voting rights, the taxation system, and educational policy) in a way that hampered long term economic development. With the goal of testing Engerman and Sokoloff’s hypothesis, Nunn (2008b) examines the relationship among slavery, inequality and economic development for a sample of 29 American countries and finds that the fraction of slaves over population in 1750 is indeed negatively correlated with per capita income in 2000. A similar negative relationship emerges for slave use, in each decade between 1790 and 1860, across states and counties within the U.S., although no evidence is found that this relationship is driven by large scale plantation slavery, i.e., by factor endowments. For the U.S. case, Nunn (2008b) also tests whether inequality is the channel through which slavery manifests itself on underdevelopment. The findings are once again mixed since, while it is true that slavery in 1860 is positively associated with contemporaneous land inequality, land inequality in 1860 is not correlated with income in 2000. Overall, these results confirm that slavery was detrimental for economic development, even though they question both the link between slavery and factor endowments and the role of inequality as the channel of transmission between slavery and current development. A closer look at the U.S. is taken by Bertocchi and Dimico (2010), who find that the negative influence of the slave share on current income is actually not robust to the inclusion of geographic controls which capture structural differences among regions, and even turns positive when state fixed effects are included. On the other hand, they find a negative impact of a dummy for slave states which, rather than the intensity of slave use, should rather reflect institutional differences, possibly linked to the Black Codes and the Jim Crow Laws. Moreover, they find that slavery has a positive and robust effect on current income inequality, i.e., those U.S. counties that displayed a higher share of slaves over population are not necessarily poorer, but more unequal, in the present day. They also show that the impact of slave use on current income inequality runs through racial inequality and that the channel of transmission from slavery to inequality is human capital accumulation. In other words, current inequality is primarily influenced by slavery through the unequal educational attainment of blacks and whites. Finally, for a sample of Mississippi counties, Bertocchi and Dimico (2011) analyze the link between slavery and political institutions and show that the former, rather than de jure provisions such as poll taxes and literacy tests, is the main driver of blacks’ restricted suffrage at the end of the nineteenth century. Consistently, race and the legacy of slavery, rather than suffrage, emerge as the main determinants of a broad set of indicators capturing multiple aspects of current development.
A provisional conclusion I can draw from the above findings is that, despite the fact that the Engerman and Sokoloff’s hypothesis is only partially supported, across different samples and specifications, still there is evidence that Africa’s slave trades had a long lasting influence not only on Africa but also on the countries that used Africa as a source of slave labor.
7. Conclusion
The scope of this essay was to reconstruct the main steps along the path leading to the discovery of the determinants of growth in Africa, taking into account subsequent waves of a large literature which initiated from standard Solowian factors and ended up enclosing a wide array of additional considerations. While this literature had a broader scope, to account for the case of Africa was one of its main challenges. With the words of Easterly (2002), I can conclude that, in and out of Africa, the quest for growth has indeed been quite an elusive one. The list of proposed determinants of growth, or lack of it, has included, in order of appearance, physical capital, demographics, human capital, macroeconomic policies, geography, ethnic division, disease, and a large variety of historical and institutional factors. Fragility, a complex mix of dysfunctions, has been the latest newcomer to this list. Accounting for Africa’s slave trades has allowed to broaden the perspective to growth spillovers even outside Africa itself. While some progress has been achieved, many questions are still open. While the list of potential growth correlates may not be exhausted yet, at least part of the responsibility for the absence of definite answers may actually fall not on lack of imagination about additional hypotheses, but on the underlying growth regressions approach. The latter certainly has its limits even after the advancements obtained through more sophisticated techniques including instrumental variables and panel estimation. Still, as acknowledged in a critical essay by Wacziarg (2002), despite their trouble in identifying causal links and the lack of robustness of their results, growth regressions have represented a first step toward a deeper understanding of what may underlie simple empirical correlations based on reduced forms. Durlauf et al. (2005) also recognize that most of the growth literature has simply attempted to investigate whether or not particular hypotheses can find any support in the data and to highlight systematic patterns. This approach, however, fails to capture the underlying channels of transmission. In order to establish causation, the estimated parameters must correspond to precisely identified links within a coherent framework derived from economic theory. Consider, for instance, the relationship between slavery and development. Even if it exhibited a robust statistical correlation, which is actually not the case, in order to understand its economic significance we would need a theory justifying the underlying mechanisms at work. In other words, why should past slavery still matter today? Because it reflects geography and initial factor endowments, because it shaped human capital accumulation in an unequal fashion, or else because it promoted divisive political institutions? To conclude, structural models based on rigorous theoretical predictions are the next — harder — step for future research on the empirics of growth.