3. Growth in Africa
As previously explained, my main focus is on the differential growth performances within Africa. Among growth determinants, I start by discussing standard economic factors. Rodrik (1999) is among the first to adopt the growth regressions approach for a sample of African countries. With the goal of understanding the specific impact of trade policy, he replicates the specification chosen by Sachs and Warner (1997) for a sample of sub-Sahara African countries over the 1965-90 period. The dependent variable is per capita growth and the explanatory variables preliminarly considered are initial per capita income, dummies for tropical climate and landlocked countries, life expectancy, public savings, institutional quality, measures of openness, and population growth. A more parsimonious specification that accounts for limited data availability shows a significant effect of initial per capita income, life expectancy, public savings, population growth, and export taxation. Thus, he establishes that even within sub-Saharan Africa there is evidence of convergence, while growth differentials are explained by a combination of human resources and macroeconomic policy. These findings are broadly in line with standard predictions from growth theory, suggesting that the sources of underdevelopment in sub-Saharan Africa are not specific to this region. However, the same exercise repeated over three sub-periods (1964-74, 1975-84, and 1985-94) reveals a much worse fit of each regression and a loss of significance of trade policy and demographics. This suggests that over shorter horizons growth rates tend to be unstable and that their determinants may vary over time, more widely in Africa if compared with the rest of the world.
Another attempt to measure growth within Africa is presented by Bertocchi and Canova (2002), who adapt the benchmark growth regressions initially proposed by Barro (1991). Over the 1960-88 period, they select a specification including the combination of economic and sociopolitical variables which displays the best explanatory power for average growth of per capita income for a cross section of African countries. Such a combination includes the initial condition (and its square), the investment-output ratio, the percentage of working age population in secondary school, the index of political rights, the index of ethnic fractionalization, and a dummy for oil producing countries. As in Rodrik (1999), in the full sample these variables tend to be associated with significant coefficients with the expected sign. However, over sub-periods, once again the picture varies considerably: in the 1960-1973 sample only investment matters, while only the index of ethnic fractionalization is significant in the 1974-1980 sample, and very little significance is left in the final 1981-1988 sample. This non-robustness denies the existence of a single cause for Africa’s poor growth performance over the period under consideration and suggests that different factors may matter for subsequent stages of development. In the initial stage investment in physical capital appears to be the most important driver, while later on human capital accumulation and political rights emerge as crucial for growth. Bertocchi and Guerzoni (2011) produce updated evidence on the growth performance of the area by running comparable growth regressions over a yearly 1999-2004 panel. They find a tendency to convergence and that economic development is facilitated by schooling and government expenditures, while it is retarded by inflation and ethnic fractionalization. Over this time frame, the relationship between civil liberties and growth is non-linear, suggesting higher growth for intermediate regimes, even though with marginal statistical significance.
In all the investigations reported above, geographical variables such as latitude and a dummy for landlocked countries, which usually matter in a world context, turn out not to contribute to the understanding of the local growth experience, possibly because they exhibit limited regional variation.
To conclude, while the analysis of standard growth factors does confirm, for Africa, conditional convergence, at the same time the influence of regressors other than the initial condition tends to vary significantly across samples, leading to unsatisfactory results that suggest a potential role for omitted variables.
4. History and colonization
The conclusion from the previous section is that standard growth determinants, even those reflecting current institutional characteristics, cannot provide a complete and robust description of the African case. One of the most promising avenues undertaken by subsequent research has been to gauge the potential impact of the history of the continent, with special emphasis on its colonization experience. The basic conjecture behind this avenue is that colonization may be the reason both for low average growth rates in Africa and, at the same time, for the observed heterogeneities across African countries. Africa represents a particularly appropriate setting to analyze the impact of colonial rule on growth because, historically, nowhere else was colonization so far-reaching and time-homogeneous in nature as in the African experience that began at the end of the nineteenth century, despite significant differences across individual countries and colonization regimes. The prevailing wisdom from the previous, huge literature outside of economics was that colonization was bad for colonial economies. According to the drain of wealth thesis, most of the colonial surplus was extracted by the metropolitan countries. Exploitation also distorted the colonial economies in many ways, by reducing physical and human capital accumulation and by generating dysfunctional institutions. An alternative point of view emphasizes instead the positive modernization impulses that came from the metropolises and the advantages deriving from the integration of the colonies into the world economic system. Within the economic literature, Lucas (1990) and Grossman and Iyigun (1995) develop static models of colonial domination, while early empirical contributions are represented by La Porta et al. (1998), who focus on the legal origins associated with colonial heritage; Alam (1994), who compares the growth rates of sovereign countries and colonies, but with the exclusion of Africa; and Grier (1999), who studies the relationship between the length of colonial rule and growth.
For the 1960-88 period, Bertocchi and Canova (2002) explore the empirical relevance of colonial variables for a sample of African countries within a standard growth regressions framework. To overcome the obstacle of data availability for the colonial period proper, they employ post-war data and historical information to identify the consequences of colonial domination for current performances. To this end, they classify African countries according to a number of indicators: their political status (i.e., colony vs. dependency vs. independent country) during the colonial period; their metropolitan ruler during domination; and the degree of economic penetration they were exposed to, as captured by the ratio of GNP to GDP at the end of the colonial period. Controlling for standard determinants, they find that in Africa the identity of the metropolitan ruler and economic penetration do add explanatory power in cross-sectional growth regressions. Namely, British colonies have superior growth performances if compared with the former colonies of France, Italy, and Portugal, while higher economic penetration is detrimental. Moreover, the colonial indicators are correlated with measures of human capital accumulations and political distortions. Hence, several decades after the end of colonization, its legacy still exerts a significant impact on growth in Africa, both directly and indirectly. These results support the conjecture that colonial rule may indeed represent the omitted factor behind the relationship between local average growth rates and economic and sociopolitical factors. While the above conclusions are based on cross-country data for the 1960-88 period, Bertocchi and Guerzoni (2011) review the evidence over a yearly 1999-2004 panel, to find that colonial indicators no longer contain any explanatory power, which suggests that the lasting influence of the colonial era may finally have faded in more recent times.