Corruption: a non-parametric analysis

This paper considers the distributional dynamics of a well-known corruption index. Specifically, we are interested in evaluating whether corruption is best characterized as multimodal (i.e. pointing to clusters of countries with persistently different levels of corruption) and whether there have been significant changes (i.e. convergence or divergence) in the distribution of the perception of corruption across countries and over time. Using non-parametric kernel density methods, our findings lend support to concerns expressed in the theoretical literature – namely, that corruption can be highly persistent, and characterized by multiple equilibria. This highlights and corroborates the conclusion that anti-corruption campaigns must be sustained to be effective.