Egypt ranks as one of the world’s most equal countries judging by official estimates of income and consumption inequality. Estimates of inequality, like estimates of poverty, are derived from national household surveys that collect detailed income and/or consumption data for a sample of households, assumed to be representative of the country’s population.
We consulted the World Bank’s PovcalNet, a repository of household income and consumption surveys from around the world, to obtain estimates of the Gini coefficient for 135 countries around 2008-09 (the most recent period for which we have survey data for Egypt). The result is presented in Figure 1; The Gini coefficient is arguably the most commonly used measure of inequality; it ranges between 0 and 1 with higher values indicating more inequality. The most unequal countries are on the left; South Africa’s Gini for example is 0.63 (not highlighted), and we move to the right as inequality declines. As can be seen, Egypt’s Gini is just over 0.3, which is low by international standards. It is even low by the standards of the Middle East.
This will come as a surprise to many. There is a sense that economic success in Egypt is reserved for the privileged, and that opportunities for the majority are largely missing. For example, Assaad (2013) estimates that a child from a disadvantaged family has a one in ten chance of enrolling in university while a child from a privileged family is virtually guaranteed to make it to university. Efforts to redistribute income from the rich to the poor are also limited. Taxes have become more regressive, particularly since 2005, as has been documented in a recent op-ed in Mada Masr by Diab (2016). There have been some efforts to restore a degree of progressivity but these have encountered resistance. During this period, signs of inequality have become increasingly more visible: “In Egypt, where luxury hotels and upscale neighborhoods abut sprawling informal settlements, inequality is out in the open, bringing with it the constant potential for social unrest.” (Diab, 2016)
In early 2011 social unrest indeed broke out. The increased level of inequality may not have been the decisive issue in the revolution although some argued it might have contributed, see e.g. Ncube and Anyanwu (2012) and Hlasny and Verme (2013). Protesters were arguably motivated by a mix of economic and political grievances (see for example Kinninmont, 2012; Arampatzi et al., 2015). It is conceivable however that many of these grievances are related to inequality one way or another. Think of the deterioration of public education and healthcare, chronic corruption, crony capitalism, police brutality, poor working conditions, low pay, and lack of accountability. Those who can afford it rely increasingly more on private service provision in education and healthcare which contributes to a widening of the cleavage between the privileged and the rest of society, something van der Weide and Milanovic (2014) refer to as “social separatism” (see also their VoxEU column). These persistent government failures arguably served as the dominant grievance. This did not go unnoticed by the leading political parties who made economic inequality a defining issue in their election manifestos after the revolution had successfully brought down Mubarak, making place for Egypt’s first democratic election (Kinninmont, 2012).
Could it be then that inequality is higher in Egypt than official estimates make us believe? Alvaredo and Piketty (2014) conjecture that inequality is indeed being underestimated. This should come as no surprise. Household surveys often fail to capture top incomes, either due to non-response of the rich or under-reporting of their incomes or both, which then leads to an underestimation of income inequality; e.g. Atkinson et al., 2011; Alvaredo and Piketty, 2014). A popular solution is to estimate the top tail of the income distribution from income tax record data, estimate the rest (often 99 percent) of the income distribution using the household survey, and then combine the two to obtain an estimate of the complete income distribution (Atkinson et al., 2011; Alvaredo, 2011; Alvaredo and Londoño Vélez, 2013; Anand and Segal, 2015).
Income tax records arguably denote the ideal source of data as far as top incomes are concerned. Unfortunately, tax record data are hard to come by, and Egypt is no exception to that rule. Hlasny and Verme (2013) made an attempt to re-estimate inequality by explicitly modeling the non-response among top income households which introduces a re-weighting of the available observations. This did not yield a meaningful correction, possibly because their approach does not consult a second (external) source of data. If the main problem is that top income households are simply missing from the household survey, then no adjustment that relies solely on the survey will resolve the downward bias in estimates of inequality.
In the absence of tax record data, we explore the feasibility of using data on house prices to estimate the top tail of the income distribution. The house price database is compiled from real estate listings that are available in the public domain. This data is used to estimate the top tail of the house price distribution. We also estimate the relationship between the house price and household income using the household survey. The two are then combined to obtain an estimate of the top tail of the income distribution (which in turn is combined with an estimate of the rest of the income distribution). The study finds evidence that inequality is indeed being underestimated in Egypt by a considerable margin. The Gini coefficient for urban Egypt is found to increase from 0.36 to 0.47 after correcting for the missing top tail. Note that this potentially moves Egypt to the mid-range in Figure 1.
Relying on predictors of top incomes rather than actual incomes derived from tax records is not without caveats. For example, one needs to make assumptions about the functional form of the relationship between the house price and household income, and about the functional form of the upper tail of the house price distribution. In addition it is assumed that one house constitutes one household and that all houses are domestically owned. Therefore, in cases where tax record data are available these should undoubtedly be considered first. Having said that we believe that this approach will provide more reliable estimates of inequality than estimates obtained using survey data alone. (This is confirmed in a robustness analysis provided in the annex to van der Weide, Lakner and Ianchovichina (2016). The perfect should not be the enemy of the good.
It is conceivable that inequality estimates for other countries may similarly be downward biased. When household surveys are combined with income tax data, the Gini coefficient for (i) the United States in 2006 increases from 0.59 to 0.62 (Alvaredo, 2011), (ii) Colombia in 2010 from 0.55 to 0.59 (Alvaredo and Londoño Vélez, 2013), and (iii) Korea in 2010 from 0.31 to 0.37 (Kim and Kim, 2013). Ideally, one would apply these corrections to all countries. Where Egypt would rank in this hypothetical version of Figure 1, nobody knows.
This post has been adapted from an original post on VoxEU.org