Recently three IMF economists published a paper arguing that redistribution is in general pro-growth (Ostry et al. 2014 ). The paper caused a stir as it dismisses right-wing beliefs that redistribution hurts growth. However, even people sympathetic to the ideas of inclusive growth and equality of opportunity find this finding problematic. One reason is that the authors rely on a measure of redistribution that misrepresents the true cost of redistribution in an economy. Another has to do with the omission of factors that affect positively the income growth of the poor and vulnerable, such as employment. This omission would exaggerate the importance of equality through redistribution as a source of growth and underplay the importance of structural transformation and investments directed towards sectors that use unskilled labor more intensively, and therefore have the potential to generate inclusive growth and productive employment for the poor segments of the population.
I’ll start with the first point. Ostry et al. (2014) define redistribution as the difference between the gross (market) inequality and net inequality series available in the Solt (2009)  database. Net inequality reflects the effect on the Gini of redistributive policies such as taxes, subsidies, and social transfers. I took a look at Solt’s measures and in particular focused on the Middle East and North Africa region. My curiosity increased after I found out that the database features nearly all developing MENA countries. Typically, we struggle to find information on most of them.
What do the gross and net Gini numbers in the database reveal? They tell me that in MENA there is little redistribution when measured as the difference between gross and net Gini. However, it is well established that governments in MENA use a wide range of measures  from taxes and subsidies to various social payments to redistribute commodity (e.g. oil) rents. So, the difference between the gross and net Gini indexes simply tells us that redistribution in MENA has not worked well. Therefore, Ostry et al.’s measure of redistribution as the difference between gross and net inequality does not capture the true cost of redistribution in the MENA region (the correlation between their measure of redistribution and the actual cost of various redistributive policies presented in Table 1 of their paper must be really low in the case of MENA). This error in measuring the size of true redistribution differs by country and will bias toward zero the estimated effect of redistribution on growth. Indeed, the coefficient on the redistribution variable in their growth regression is insignificant and close to zero in all of their specifications. If this is the case, one of the messages in the paper that redistribution appears generally benign in terms of its impact on growth may not be applicable to MENA.
Another problem related to the findings of the Ostry et al. paper is the omitted variable bias. The authors use the GMM estimator which accounts for time-invariant unobservable factors. But the estimation does not control for time variant factors correlated both with inequality and growth, for example, policies that boost investment and growth in sectors that use unskilled labor most intensively.
Finally, one wonders about the accuracy of the Gini estimates. According to the summary data from the Solt (2009) database, the inequality measures (net or gross) are lowest in the countries with most severe civil unrest, except Yemen. Said differently, these measures would not have predicted the popular uprisings in Tunisia, Egypt, and Syria in 2010-11. Either the inequality indexes do not give us an accurate picture of inequality in these countries or other factors make inequality tolerance higher in countries such as Iran, Lebanon, and Morocco-- the countries with the highest Gini coefficients in the region. For the Middle East and North Africa, the research findings of Ostry et al. (2014) clearly pose more questions than they answer.