Subgroup and Shapley Value Decompositions of Multidimensional Inequality: An Application to South East European Countries

Summary

Inequality is often discussed along a single dimension like income. This is reflected in the various decomposition approaches of inequality indices. However, inequality is a multidimensional phenomenon in reality. 

Using OLS-regression analysis on data drawn from a variety of sources, the study analyzes multidimensional inequality in three large South-East European countries, Serbia (2007) and Bulgaria and Romania (2008). In order to construct the multidimensional inequality index, it combines four dimensions: household income, household health, household education level and housing quality, and applies various decomposition methods to single  and multidimensional indices of inequality. The decomposition analysis of the Massoumi index underlines the outstanding importance of education differences in determining inequality in welfare levels in Bulgaria and Romania.

Results indicate that in Bulgaria and Romania, income and education inequality can be explained very well by the differences in the educational attainment level of the head of the household, the participation of household members in the labor market and the differences between rural and urban regions. The same characteristics stand out in the case of income inequality in Serbia but their explanatory power is much lower, while education inequality cannot be explained at all. Also, the decomposition analysis for the dimension household health points towards the importance of education and labor market participation. In all three countries, education turns out to be the most important determinant of the composite inequality measure, with employment status ranking second in Bulgaria and Romania and third in Serbia. As opposed to traditional approaches, the paper furthermore shows how a Shapely decomposition can be a valid approach to comparing inequality measures across countries. 

If you cite this resource, please notify communications@gdn.int with the subject line 'GDN citation'.