Political centralization in pre-colonial Africa

Abstract

In this paper we investigate the empirical correlates of political centralization using data from the Standard Cross-Cultural Sample. We specifically investigate the explanatory power of the standard models of Eurasian state formation which emphasize the importance of high population density, inter-state warfare and trade as factors leading to political centralization. We find that while in the whole world sample these factors are indeed positively correlated with political centralization, this is not so in the African sub-sample. Indeed, none of the variables are statistically related to political centralization. We also provide evidence that political centralization, where it took place, was indeed associated with better public goods and development outcomes. We conclude that the evidence is quite consistent with the intellectual tradition initiated in social anthropology by Evans-Pritchard and Fortes in the 1940s which denied the utility of Eurasian models in explaining patterns of political centralization in Africa.

Introduction

Most of the poor people in the world live in Sub-Saharan Africa (henceforth Africa). The income per-capita of the poorest countries such as Ethiopia or Sierra Leone differ from those of prosperous OECD countries by a factor of about 40 and these income differences come along with huge differences in welfare, health, economic opportunities and life chances. Two hundred and fifty years ago, before the ‘great divergence’ we know that these differences were much smaller. Parts of the world which have now very different levels of income per-capita were indistinguishable according to this metric in 1750. But how poor relatively was Africa compared to other parts of the world? Some, like Hopkins, 1973, Thornton, 1992, Jerven, 2010, Ehret, 2012, see few historical differences in institutional dynamics and prosperity between Africa and the rest of the world. Others are more selective, arguing that while Africa may have been behind Eurasia it was ahead of the Americas (e.g. Inikori, 2012). Africa was certainly behind the rest of the world, even the Americas, technologically (Goody, 1971, Austen and Headrick, 1983, Law, 1980) and Acemoglu and Robinson, 2010, Acemoglu and Robinson, 2012 argue that more generally it was economically backward in many dimensions at the start of the early modern period, and in particular did not have the economic and political institutions necessary to generate modern economic growth.

If one accepts that Africa lagged the rest of the world in term of developing basic economic institutions and public goods which might have stimulated technological change or adoption in the early modern period what might have been behind that? This question has been answered in many ways. Easterly and Levine (1997) argued that the great ethnic fragmentation of Africa was one reason for its relative poverty, Sachs and Warner (1997) suggested Africa had adverse geography, while Inikori, 1992, Nunn, 2008, following a large literature in Africa studies, focused instead on the deleterious impact of the slave trade. Other scholars have instead emphasized more recent factors potentially causing poor economic performance in Africa, such as the arbitrary nature of post-colonial national boundaries (Englebert, 2000).

One thing that everyone seems to agree on is that state institutions have been dysfunctional in Africa. Much of the political science literature saw economic decline after independence as being closely related to states that were unable or unwilling to provide public services or encourage economic activity (Callaghy, 1984, Turner and Young, 1985, on the totemic case of Zaire, Young (1994) for a synthesis). Moreover, the lack of effective centralized states is clearly a potential factor not just in explaining poor economic performance in Africa since 1960, but also over a much longer duree. Whatever the impact of the colonial period might have been on state formation in Africa, at a factual level the evidence seems to suggest that Africa developed centralized states later than the rest of the world. Though Africa certainly did have states and quite a few emerged and consolidated in the 18th and 19th century, this process seems to have definitely lagged behind Eurasia and at least parts of the Americas (Central America and Andean South America). One can get some quantitative picture of this via the data coded by Louis Putterman and his collaborators (Bockstette et al., 2002). Fig. 1 plots their state antiquity index from 1000 to 1500. This captures the extent to which a country in the world has been under the control of a centralized state. It shows that though Africa did have greater state antiquity historically than the Americas or Oceania, it lagged behind the rest of the world, particularly Eurasia.

A large literature in social science places the functioning of the state at the heart of economic development. Ultimately this view comes from the work of Max Weber and recent comparative work has claimed that the difference between developmental successes and developmental failures is indeed that the former have effective states while the latter do not (see for example Evans (1995), or the recent literature by economists: Acemoglu, 2005, Acemoglu et al., 2013, Acemoglu et al., 2011, Besley and Persson, 2011). In the context of Africa the most ambitious version of this argument is due to Herbst, 2000, Bates, 2001. Indeed, as we discuss in Section 2, the case study literature in Africa does indeed suggest that the absence of centralized state authority is a potent source of poor economic institutions and the absence of public good provision, potentially helping to explain lagging economic development of Africa.

But what could explain the differential development of centralized states in Africa? This topic has been researched at least since the famous volume edited by the social anthropologists Evans-Pritchard and Fortes (1940) and has taken two broad lines. Many scholars, for example Diamond (1997), Herbst (2000), Bates (2001) and Reid, 2012, Reid, 2012, take what they see as the successful models which have supposedly explained political centralization in Europe and apply them to Africa. Here the key would be the absence of the factors which led to the formation of states in Europe, usually warfare, high population density and trade. For example, factors unique to Africa such as a very adverse disease environment or lack of domesticable plant and animal species kept population density low which retarded the development of states. Other scholars, exemplified even by Evans-Pritchard and Fortes (1940) and more recently by the essays in McIntosh (1999a) particularly McIntosh, 1999b, Vansina, 1999, deny the applicability of Eurasian model to explain the dynamics of political institutions in Africa. Their main point is that it is not simply that there is less political centralization in Africa, but that the development of political institutions took a different path historically creating qualitatively very different structures from those seen in Eurasia.

Though the causal mechanisms linking population density, warfare and trade to political centralization may be widely regarded as plausible they suffer from many problems. Most obviously to our knowledge there have been only three empirical studies of the determinants of political centralization in Africa. Though it was not the focus of his research, Nunn (2008) found an important negative correlation between the intensity of the slave trade and political centralization. More recently, Fenske (2012) argues for a causal relationship between ecological diversity and state formation in Africa arguing that this picks up the potential for specialization and trade. Alsan (2012) shows that the incidence of tse-tse fly is negatively correlated with political centralization which she argues works through various channels, including lower population density and the inability to use draught animals. None of these papers uses the data we use in this paper or considers other potential hypotheses. Reflecting this lack of research there is also considerable controversy about what the basic correlations are. Evans-Pritchard and Fortes (1940) famously denied that there was any correlation between state formation and population density in Africa. Noting (p. 7).

“size of population should not be confused with density of population. There may be some relations between the degree of political development and the size of population, but it would be incorrect to suppose that governmental institutions are found in those societies with great density.”

This view was contested by Stevenson (1968), reaffirmed by Vengroff (1976), contested by Bates (1983), but recently reaffirmed once more by McIntosh (1999b), though none of these papers actually looked systematically at the most complete data available. Other issues, such as the connection between warfare and state formation in Africa seem never to have been systematically investigated. If the correlations are controversial, this must a fortiori be true about causal relationships. For example, even if it were true that there is a positive correlation between population density and state formation, this does not imply that higher density of populations make it more likely for states to form. It could well be the other way round. Indeed, much case study evidence from Africa suggests that state formation is followed by population expansion rather than the other way round.¹

In this paper we use the Standard Cross Cultural Sample to conduct some preliminary tests of the conventional hypotheses about the drivers of state formation in the whole world and Africa. This dataset was first constructed by Murdock and White (1969) using materials from the more frequently used Murdock Ethnographic Atlas, but covering far fewer “cultures”. The rough idea was that the societies in the Atlas were not independent cultures since many of them had common roots and were subject to cultural dissemination and contagion. The SCCS has been greatly added to over time by many different anthropologists and now includes 186 cultures. This is far fewer than the number of societies in the Atlas but unfortunately that dataset does not include most of the main explanatory variables in which we are interested. Though this dataset is less than ideal, and very incomplete since any variable usually has many missing observations, it does contain quite a lot of important variation in Africa both spatially and in terms of political centralization (see Fig. 2). For instance it includes the!Kung bushmen and the Hadza, two hunter–gatherer societies; a few classic unpolitically centralized societies in Africa, these include the Tallensi or Northern Ghana, the Kikuyu of Kenya and the Tiv of eastern Nigeria; it also includes some salient politically centralized societies such as the Asante of Ghana, the Ganda of Uganda, and the Hausa of Nigeria. Finally there are intermediate societies such as the Bemba of Zambia and the Mende on Sierra Leone. Nevertheless, it only includes 40 cultures in sub-Saharan Africa.²

Our findings are quite contrary to the existing literature which has stressed the applicability of Eurasian models of political centralization in Africa. Our initial examination of the data uses Ordinary Least Squares (OLSs) regressions and thus consists only of conditional correlations. However, we find that while for the whole world sample there is a very robust positive correlation between population density and political centralization and at least some evidence of a positive correlation between trade and warfare and political centralization, this is not so for the Africa sub-sample. Indeed, within Africa all of these potential explanatory variables are uncorrelated with political centralization. Though the sample size is small and these results should obviously be regarded as preliminary it is quite striking that the theories fail to predict the correlations in the data so completely.

We investigate this further by developing an identification strategy to cope with the potential endogeneity of population density. Since the SCCS contains a variable for the agricultural potential of the area inhabited by a particular society we use this as a potentially exogenous source of variation in population density. In the first stage there is a strong positive correlation between agricultural potential and population density in the world, though this is weaker in the African sub-samples. In the second stage our Two Stage Least Squares (2SLSs) results confirm our initial findings: there is a robust positive correlation between population density and political centralization in the whole world but not in Africa.

Finally, since the SCCS also includes various measures of public goods, particularly the use of money, writing and various modern forms of infrastructure such as roads, we can further use it to investigate whether the correlations are consistent with political centralization being important for development. We find very strong positive correlations between political centralization and all these public goods outcomes suggesting that the relative lack of political centralization in pre-colonial Africa may indeed be an important part of the story about African underdevelopment.

Though these results might be thought to be somewhat negative we feel they are an important test of the recent literature on the origins of the state in Africa. They suggest, as social anthropologists have long argued, that historical states in Africa were not only less centralized but developed according to different logics than the states of Eurasia. In consequence the presence or absence of the factors that created states in Eurasia are not relevant for determining the variation in political centralization in Africa. A different theoretical approach is required. One alternative which flows from the anthropological literature is that Africa differs from Eurasia in terms of social structure and this may be an important influence on political centralization. For example, the types of age structures and organizations common to many stateless societies in Africa (e.g. Prins, 1953, Bernardi, 1985) may create cross-cutting cleavages which make it very difficult for one group to accumulate political power and build a state (see McIntosh (1999b) for related ideas). We are investigating these ideas in ongoing research (Acemoglu et al., 2012). Our argument is as follows. It is natural to think of a society lacking political centralization as divided into different relatively autonomous groups, often based on kinship. Put simply, political centralization involves one of these groups dominating the others and removing this autonomy. Though many things may influence the incentives and constraints which face a particular group attempting to dominate the others, we hypothesize that in Africa a significant important factor is the existence of cross-cutting cleavages which link groups together. There are many types of such institutions but the canonical example might be age sets the most famous definition of which is due to Radcliffe-Brown (1929, p. 21) as

“a recognised and sometimes organized group consisting of persons (often male persons only) who are of the same age .. In Africa, at any rate in East and South Africa,an age-set is normally formed of all those males who are initiated at one time … Once a person enters a given age-set, whether at birth or initiation, he remains a member of the same age-set for the remainder of his life .. In East Africa, where the age-organization is highly elaborated, each age-set normally passes from one grade to another as a whole.”

But they also include many other types of institutions such as secret societies, cult groups, titling societies and various other types of associations. The hypothesis Acemoglu et al. (2012) is that it is unique elements of the social structures of African societies which make it very difficult to centralize power because they make it difficult for one group to dominate others, a precondition for political centralization. Though this theme surfaces at many places in African studies to our knowledge it has not been precisely formulated before and tested. It dates back at least to a series of critiques of Evans-Pritchard and Fortes’s dichotomy between state and non-state (lineage) societies. Brown (1951) challenged this on the basis that “associations” defined as “an organized and corporate group, membership in which does not follow automatically from birth or adoption into a kin or territorial unit” were critical links between lineages in many non-state societies, such as the Igbo or Tallensi in West Africa (a point made independently for East Africa by Bernardi (1952)). But she points out (p. 270) that “no associations were found in Ashanti, and that the associations of Dahomey and Nupe did not use sanctions against non-members”. Strikingly, of all the places she studied, these were the ones that had centralized states. Similar issues arise in McIntosh’s critique of the application of non-African models of political complexity to understanding the development of political institutions in Africa. She notes “The distribution of power among several corporate entities (e.g. lineages, secret societies, cults, age grades) can be regarded as a strategy that has successfully resisted in a variety of ways the consolidation of power by individuals” (1999b, p. 4). However, her emphasis is not on how these block the creation of political centralization but on how they create qualitatively different complex political organizations. In more specialized literature, particularly on East African stateless societies, there is a huge amount of emphasis on such cross-cutting institutions, particularly age-sets and the political role they play (e.g. Prins, 1953, Dyson-Hudson, 1963, Dyson-Hudson, 1966, Bernardi, 1985). At the same time in societies with states, such as the Buganda, the Basoga, or Rwanda there is often mention of the absence of these cross-cutting associations (e.g. Fallers, 1964, Fallers, 1965 on the former two and Vansina (2004) on the latter), but the comparative picture never seems to have been put together.

The paper proceeds as follows. In the next section we discuss the data we use in detail. In Section 3 we them explore the issue of whether polities with greater degrees of political centralization have better development, particular public goods, outcomes. This exploration is important for motivating our study of political centralization and it is interesting to undertake it with the SCCS. We also discuss some case study evidence that the lack of political centralization in Africa has been responsible for poor economic institutions and poor economic performance. In Section 4 we then turn to a very preliminary analysis of the basic hypotheses using OLS regressions. Section 4.2 then discusses our identification strategy for population density and provides our 2SLS estimates. Section 5 concludes.