For all their flaws, the major cross-country corruption indexes—Transparency International’s Corruption Perceptions Index (CPI), the World Bank Institute’s Worldwide Governance Indicators (WGI), and the like—have been quite useful, both for research (at least when used appropriately) and for advocacy. But one important limitation of these datasets is that by focusing on corruption (or perceived corruption) at the country level, they may obscure the fact that there can be substantial within-country variation in the level of (perceived) corruption. This variation may occur across government institutions—the same country may have quite different degrees of corruption in the health sector, the police force, the judiciary, customs, etc. More pertinent here, there may also be significant heterogeneity across regions, particularly in large countries with substantial political decentralization. Indeed, numerous studies have exploited within-country regional variation in corruption levels to test various hypotheses about corruption’s causes and consequences; such studies include research on Italy, Russia, China, the Philippines, and the United States, among others. But these studies typically make use of particular data sets that are not reproduced year-to-year.
As we’re starting to see rapidly diminishing returns from the major cross-country corruption datasets, it is high time for those organizations with the resources and capacity to compile information on corruption perceptions on an ongoing basis to turn their focus to within-country regional variation in corruption. I propose the creation of a sub-national corruption perceptions index (snCPI), starting with the so-called BRICS countries (Brazil, Russia, India, China, and South Africa), which would gather and compile data (primarily perception-based data, perhaps supplemented with more objective data when available) on perceived corruption levels within the major sub-national units (states/provinces, autonomous regions, and municipalities) within each of those countries.