2.3It is also argued that Africa’s middleclass is strongest in countries that haverobust and growing private sectors
. Africa’smiddle class is not only crucial for economicgrowth but is also essential for the growth of democracy.
Definition of the Middle Classin the Context of this Study
2.4This brief uses an absolute definition of per capita daily consumption of $2-$20 in2005 PPP US dollars to characterize themiddle class in Africa. The study providesthree sub-categories of the middle class. Thefirst sub-category is that of the “floatingclass” with per capita consumption levels of between $2-$4 per day. Individuals at thislevel of consumption, which is only slightlyabove the developing-world poverty line of $2 per person per day (the second povertyline) used in some studies
, remain largelyvulnerable to slipping back into poverty in theevent of some exogenous shocks. Thiscategory is crucial because it is a hingebetween the poor and lower middle classcategory. This class is vulnerable andunstable, but it reflects the direction of change in population structure through time. The second sub-category is that of the“lower-middle” class with per capitaconsumption levels of $4-$10 per day. Thisgroup lives above the subsistence level andis able to save and consume non essentialgoods. The third sub-category is the “upper-middle class” with per capita consumptionlevels of $10-$20 per day.
2.5A variety of data sources were used tocreate the population distributions anddetermine the size of the African middle class(section 3). For African countries, the primarysource for the distribution data was obtainedfrom the World Bank’s PovcalNet database,which provides detailed distributions of eitherincome or household consumptionexpenditures by different percentiles basedon actual household survey data. In addition,PovcalNet provides information on meanhousehold per capita income orconsumption levels in 2005 PPP dollars.2.6The database primarily provides sampledistributions based on consumption except ininstances in which only income measuresexist. At lower income levels, the differencebetween consumption and income is small.But this difference tends to grow with wealthand thus should be considered a potentialmeasurement error in the analysis. Still weexpect that these differences are relativelyminor as there is a high correlation betweenincome and consumption especially at lowerlevels and thus should have little effect onoverall computations. We also focus onconsumption as it better captures individualwelfare and is less prone to fluctuationscaused by negative and positive shocks.2.7The data used in the identification of the characteristics of the middle class(section 4) was obtained mainly from the AfDB Data Platform database and otherinternational sources.
2.8The tabulated distributions and theirmeans are used to generate a Lorenz curvefor each country to show the share of incomefor each proportion of the population. Theline of equality depicts a situation of perfectequality in a society where a given proportionof the society would have an equally givenproportion of the income. The extremeopposite would be where an individual wouldhave all the income while everyone else hasnone. The Lorenz curve is normally used torepresent inequality in the distribution of income among the population. The furtherthe curve is away from the line of equality, themore unequal the income distribution is in asociety. The gini coefficient, also measureslevels of inequality, and is derived from theLorenz curve. It represents the proportion of the area between the line of equality and theLorenz curve and the total area above andbelow the Lorenz curve. The literature on theestimation of Lorenz curves provides anumber of different functional forms. Two of the best performers among them are thegeneral quadratic (GQ) Lorenz curve
andwhat may be called the Beta Lorenz curve
. There is some evidence for Indonesia that theBeta model yields somewhat more accuratepredictions of the Lorenz ordinates at thelower end of the distribution, though thesame study found that the GQ model is moreaccurate over the whole distribution
. TheGQ model, however, does have onecomparative advantage over the Beta model,namely, that it is computationally simpler
.2.9This study therefore uses the Betaapproach for estimating the various points of the Lorenz curve showing the proportion of the population and their share of nationalincome (see Box 2 in Appendices). Using asystem of six equations, each representing atrapezoid-shaped area under the Lorenzcurve for a series of coordinates, we obtainthree unique values representing theproportions of the population associated withper capita daily income levels $4, $10 and$20 (see chart 20 in Appendices). Theserepresent the “floating class”, “lower-middle”and the “high-middle” class categoriesdefined by this study.
3Africa’s EmergingMiddle Class
3.1 The results of this study reveal that Africa’s middle class has increased in sizeand purchasing power as strong economicgrowth in the past two decades has helped
African Development Bank
e Middle of the Pyramid:Dynamics of the Middle Classin Africa
Market Brief • April 20, 2011 • www.afdb.org
See Ravillion, Chen, and Sangraula, 2008.
See Box 2 in the Appendices for more details on the methodological approach.
Villasenor and Arnold 1984, 1989.
Ravallion and Huppi, 1990.