Show Me Everything But the Money

Every year, with some fanfare, Transparency International releases its Corruption Perceptions Index (CPI). This is perhaps the best known of a range of measures of a country’s level of sleaze. Such indicators and estimates have their uses — they highlight a significant global problem, one that can have a particularly large impact on poor people. What they aren’t terribly good at, paradoxically, is measuring how corrupt countries actually are — and understanding why that’s the case is important not only to solving developing countries’ corruption problems, but the rest of their problems as well.

Take the case of Peru. In 2000, tapes of National Intelligence Service head Vladimiro Montesinos bribing legislators, judges, TV station operators, and others — 1,600 people in all — led to Montesinos’s ouster and President Alberto Fujimori’s impeachment. They also precipitated a significant drop toward unclean status in the country’s ranking on the Corruption Perceptions Index. But the collapse in the CPI came after the tapes were released. There was no significant change in the index prior to the release of the tapes, when Peru was perceived as being cleaner than the Czech Republic, for example. Of course, before the tapes were released was when the actual corruption was going on. And the perceptions index kept on dropping even as survey evidence in Peru suggested the corruption clean-up was working.

It isn’t just Transparency International, of course. Measures of perceived corruption, or poor governance, or weak rule of law are as common as they are cheap to make and easy to get publicity for. But that doesn’t make them accurate. In fact, the Peru example demonstrates the big gap between perception and reality driven by our own expectations and biases. (And to be fair, Transparency International itself admits that you can’t easily use changes in the CPI over time to measure changes in levels of corruption.)

Six years ago, economist Benjamin Olken examined the relationship between people’s perceptions of corruption in road building in villages in Indonesia and actual amount of construction materials stolen from the projects. He found that villagers’ perceptions regarding corruption in a particular project in their own village had little to do with how much corruption was actually happening. Indeed, perceived levels of corruption were driven far more by general cynicism about the honesty of politicians, the education level of respondents, and the ethnic makeup of the village in question than they were by actual corruption. While perceptions of corruption were considerably higher in ethnically diverse villages, for instance, objective evidence of corruption was considerably lower.

If such problems plague perceptions of corruption in a project that survey respondents actually know something about, you can guess at the inaccuracy of surveys asking (as Transparency International does) for country-level perceptions from an international audience, many members of which may never have set foot in the country being judged. The available evidence suggests that the CPI tracks somewhat better with the level of petty corruption — grease payments for things like an electricity connection — than it does with country-level survey responses regarding bribes for government contracts. But even that relationship is weak. There are better measures of corruption available, to be sure — surveys asking entrepreneurs if firms in their line of business pay a bribe for government contracts or basic services, for instance, are surely a more accurate tool. But even these measures suffer from considerable noise. And it isn’t surprising that the links between perceptions, surveys, and reality can all be weak — corruption isn’t something people advertise, after all.

That weak link helps to explain the very fragile relationship between perceived corruption and outcomes like the quality of government services that you’d expect to be affected by rampant malfeasance. For example, where a country lands on Transparency International’s index has no relation to its levels of private investment in infrastructure, extent of infrastructure access, or use of telecommunications and electricity grids. In fact surveys of enterprises in developing countries suggest that countries that show more concern about the impact of corruption see more private investment in infrastructure, more mobile phones subscribers, and no fewer water or electricity connections than would be expected given income levels.

Does that mean the extent of corruption just isn’t that important to development outcomes? Far from it. But it does suggest a problem: If our measures of corruption are so weak, how can we know what works to stop it? The answer is to look at the very outcomes we think it affects. We’re concerned about corruption because it leads to services that are inefficient, low quality, and with limited reach — water systems that only reach the urban elite, for instance, and force the rest of the people in a city to walk ages to find a standpipe. And we have far better measures of the efficiency, quality, and extent of service provision than we do of corruption, for the simple reason that it is much harder to conceal outcomes like few connections and low quality of service than it is to hide corruption.

Take the impact of corruption in the water sector. MIT professor Jennifer Davis interviewed 1,400 people involved in water provision in South Asia, and found a system in which bribes were required for connections, contracts, and licenses; officials were paid to overlook substandard work; and people bought their way into government jobs where the opportunity to collect bribes was large. Surely this extensive corruption is one factor behind the fact that nearly a third of the country lacks access to safe drinking water and even those with easy access to a tap suffer incredibly high infection rates from water-borne disease — across the country there are 38 million cases annually.

We should measure the success of our anti-corruption interventions by monitoring numbers like this — not people’s imagined sense of how bad corruption is. If we introduce a reform — one directly aimed at reducing corruption — that improves the quality, efficiency, and reach of services, it is likely that we have successfully reduced corruption along with it. And, frankly, even if we haven’t reduced corruption, we’ve improved the outcomes we care about — hardly a disastrous mistake. How well a country actually does by its citizens is, after all, far more important than how good it is at keeping up appearances.