Abstract
Corruption is notoriously hard to measure directly, and cross-country corruption indices thus often rely on indirect information such as country experts’ or international businessmen’s perceptions. Transparency International’s Corruption Perception Index (CPI) is one such indicator that is often used by policy makers and researchers.
The CPI is a composite index that, in its 2019 version, draws on 13 different data sources for calculation, with a threshold of at least three available sources for a country to qualify for a ranking. Until now, however, it has not been clear whether the data sources it uses are the only suitable ones.
To assess this, we revisit the choice of these data sources and propose several improvements to the CPI methodology. Specifically, we identify up to five new data sources as potential candidates for inclusion.
We examine the effects of including these additional data sources in two simulations: including all five data sources or only the four most suitable ones. Our results are mixed: the inclusion of new data sources would lower the standard errors of the CPI, but we identify a lack of correlation between the CPI and some of the data sources.
We conclude by discussing trade-offs involved in including additional data sources in the CPI that may provide lessons for other composite policy indices.