Abstract
The Kuznets hypothesis, that is, inequality first rises and then falls as the economy advances, is often tested by regressing inequality on income and its squared term (along with other determinants). Findings of a significantly negative coefficient on income and a significantly positive estimate on the quadratic term are commonly taken as evidence supporting the inverted-U Kuznets curve. Although intuitive, Lind and Mehlum (Citation2010) argued that the conventional approach is flawed and proposed an appropriate test for a U-shaped association. We revisit the validity of the Kuznets hypothesis by applying the novel testing strategy of Lind and Mehlum (Citation2010) to annual US data over the period 1917 to 2007. Inconsistent with the Kuznets hypothesis, the test results overwhelmingly reject the combined null hypothesis of an inverted-U or monotone relationship in favour of a U-shaped linkage between income inequality and economic development. Moreover, the results are robust to changes in inequality measures and functional specifications.
Acknowledgements
The authors are grateful to Jo Thori Lind and Halvor Mehlum for making available the Stata code used in this article. Income inequality data are kindly provided by Emmanuel Saez.
Any remaining errors are our own responsibility.
Notes
1The description below follows closely that of Lind and Mehlum (Citation2010).
2In the quadratic case, the presence of a U indicates that β1 + 2β2ly
l
< 0 and β1 + 2β2ly
h
> 0, whereas in the inverse case, a U shape means that and
.
3Please see Piketty and Saez (Citation2003) and Atkinson et al. (Citation2011) for the methodology used to construct the top income shares.