Building the integrated inequality database and the seven sins of inequality measurement in Sub-Saharan Africa

The favourable growth performance of SSA over the last twenty years (Figure 1) - emphatically referred to by some as ‘the SSA Renaissance’ or ‘Africa Rising’ - has been accompanied by a perceptible, but still modest decline in poverty, from 59 to 48 percent over 1993-2010, i.e. much less than that recorded in South Asia (Ferreira 2014). Such aggregate trend however conceals substantial cross country variations. How to explain then such differences in poverty reduction rates? The standard approach (Bourguignon 2003) shows that the percentage change in poverty rates can be decomposed in the percentage change in GDP per capita growth rates and the percentage changes in the Gini coefficient, plus a (generally small) residual.2 In this regard, it must be noted that in SSA the average GDP growth per capita oscillated in a narrow range, i.e. between 1.7 percent in non-resource rich countries and 2.6 percent in resource-rich ones. The reason why poverty declined at different rates is therefore to be found in the divergence of inequality trends experienced by the countries of the region. This paper as well as Cornia (2014) and the literature quoted therein argue in fact that over the last 20 years the Gini index of inequality rose in several countries but simultaneously fell in a similar number of them. A proper documentation of inequality trends in the region becomes therefore essential to explain the above mentioned differences in poverty reduction. This task however is hindered by the limited and at time conflicting inequality data in the region and by the lack of a comprehensive database of good-quality and consistent inequality statistics. This situation is even more penalizing when considering that over the last two decades policy formulation has become increasingly ‘evidenced-based’, i.e. based not only on ideological and doctrinal priors but also on the empirical evidence provided by a growing number of household budget surveys (HBS), demographic and health surveys, wealth surveys, multiple indicator cluster surveys, multipurpose living standard measurement studies, and other surveys. The field of studies that has benefitted the most from such increase in the number of surveys is that concerning poverty alleviation and the control of inequality. In most developed and developing regions academic and policy institutions have by now built databases tracing the evolution of the Gini coefficient over at least the last 20 years, as in the case of LIS for the OECD countries, SEDLAC and CEPALSTAT for Latin America, TRANSMONEE for the European economies in transition and so on. Finally, during the same period global inequality databases were also created, including WID, WIID, SWIID, Allgini and others which are discussed below. In view of the problems caused by few and scattered inequality data and the lack of an assessment of their quality and pitfalls, this paper aims at doing two things:4 • First, in Section 2 it describes the Integrated Inequality Database (IID-SSA) obtained by comparing the Gini coefficients included in the existing databases, and selecting the least biased Gini’s. IID-SSA thus summarizes hopefully in the least distorted and systematic way the existing information on inequality, permitting in this way to analyze the changes recorded in the region in this field during the last two decades, and to draw policy recommendations. The IID-SSA dataset is illustrated in detail in Annex 1. It provides a summary of all Gini coefficients from all international databases and national sources not included already in the former, it selects the best time series of country Gini coefficients for the years 1993-2011 on the basis of a standard protocol, and plots their time trend for the 29 countries with at least four good-quality and well-spaced Gini points. Annex 1 also provides summary information on Gini availability for countries with only 1-3 Gini data. The time series for the 29 above countries can be used for a variety of analytical and policy purposes, be they the calculation of changes in poverty rates over time or panel regressions of Gini trends. Yet, given the data limitations and biases discussed in Section 3, this information has to be used with a pinch of salt, i.e. checking the results they may generate against those predicted by economic theory, economic history and other statistical sources (such as the national accounts) and by introducing whenever feasible the statistical adjustments indicated below. • Second, in Section 3 it discusses the limitations and biases of the data included in IID-SSA and tries when possible to measure the extent of such biases with the purpose of alerting the researchers of African inequality of the ‘seven sins of inequality measurement’5 most commonly met in the region. Section 3 also presents the approaches currently followed to remedy – when possible - such problems. Such seven problems concern: differences in the design of successive surveys within a country; differences in survey design across countries; under-sampling of top incomes; possible inconsistency between Gini data derived from surveys and the ‘labour share’ computed on the bais of the national accounts; the neglect of incomes generated by assets held abroad; the distributive impact of different dynamics of food prices and CPI; and the neglect of the public social services in kind provided by the state when calculating the Gini coefficient. In a way, Section 3 represents a ‘checklist of possible biases’ that researchers, statisticians and policy makers aiming at computing the ‘real Gini coefficient’ of a country should take into account. Indeed, the usual way the inequality data are computed often constitutes an oversimplification that mostly leads to an underestimate of inequality and lack of policy action. Yet, the corrections suggested in this paper require the availability of survey micro-data (not available to us) and are labor- and assumptions-intensive But carrying out such corrections allows to compute Gini data that are more precise than those included in IID-SSA data, and get in this way a better grasp of the real distributive situation of a country. Academic economists and staff of UNDP and World Bank are well advised to introduce such corrections when working on poverty and inequality at the country level