Kuznets's hypothesis and the data constraint

Abstract

Five decades ago, Simon Kuznets expressed an important hypothesis about the relationship between the degree of income inequality within a country and its level of economic development: the Kuznets's inverted-U hypothesis. The lack of longitudinal data has forced the use of cross-section or pooled datasets in order to draw conclusions about that relationship. In the present note we highlight the lack of international comparability of surveys where the measures of inequality are based, and we show two main findings: (1) data comparability goes on constituting a problem, particularly in what respects to the different welfare indicators used in national surveys, and (2) the procedure usually used to minimize the problem of noncomparability is likely to enforce the bias rather than to solve it.

Notes

¹Paukert (1973) and Lecallion et al., (1984) are other supporters of the Kuznets's hypothesis.

² See also Vicente and Borge (2000).

³ Where y denotes real GDP per capita.

⁴ According to UNU–WIDER (2005, p. 13) a re-examination of the sources for D&S ‘revealed several instances of mistakenly labelled ‘high quality estimates’, i.e. that did not, in fact, meet the criteria that had been set up’.

⁵ For a more complete discussion on quality and consistency in income distribution data both within and across countries, see Atkinson and Brandolini (2001).

⁶Li et al., (1998) following D&S (1996) report that, everything else being the same, income-based Ginis are on average greater than expenditure-based Ginis by some 6.6 Gini points. Consequently, in their regressions, they increase expenditure-based Ginis by 6.6 points.

⁷ If the country i presents only an observation in the 1995 to 2003 period, we use that number, if the country reports more than one value we use the average of such values.