What Do Corruption Indices Measure?

Apparently, not the level of corruption. This is the finding of a new paper (What_do_corruption_indices_measure?) by Dilyan Donchev (Harvard) and Gergely Ujhelyi (University of Houston).

The authors use data on actual corruption experiences and correlate it to reported corruption perceptions in surveys by Transparency International and the International Country Risk Guide. There is no correlation. Some of the factors commonly found to "reduce" corruption, such as economic development, democratic institutions or Protestant traditions, systematically bias corruption perception downward from corruption experience. In other words, people assume that there would be less corruption in rich countries or democratic countries, regardless of the actual experience.

In addition, perception indices are influenced by absolute (as opposed to relative) levels of corruption, which penalizes large countries. Why is that the case? Because in large countries the media reports more stories of corruption (in terms of absolute numbers) and so people who answer the surveys in, say, Brazil, think the country is more corrupt than, say, Argentina, even if in terms of the number of corruption instances per adult population Argentina may have more.

In sum, perceptions indices are like the king's new clothes: they are only in the eyes of the beholder.