A (Very) Brief History of Corruption

It now appears that more than 300,000 people were killed by collapsed buildings in the devastating Haiti earthquake of January 2010. That earthquake, which measured 7.0 on the standard scale, was powerful, to be sure — but it was only the world’s 14th most powerful since 2000.

The 13 more powerful quakes are estimated to have killed a combined 165,000 through building collapse. Some far more powerful earthquakes than Haiti’s led to far less devastation. For example, later in 2010, a magnitude 8.8 earthquake — the second most powerful since 1980 — killed fewer than 1,000 in Chile.

Recent research argues that the difference in earthquake devastation caused by building collapse is not simply a result of the location of the quake relative to population concentrations, nor is it directly the result of poverty’s effect on building quality.

Nicholas Ambraseys of Imperial College London and Roger Bilham of the University of Colorado at Boulder argue that the extent of corruption tells us more than poverty about the consequences of earthquakes.

Their recent research found that an earthquake of a given magnitude is especially devastating where corruption is especially virulent. This is because in highly corrupt societies, bribes make it possible for buildings to be thrown up without regard to codes and regulations. When shaken by earthquakes, these buildings are completely obliterated, killing their inhabitants.

There could hardly be a more dramatic demonstration of the costs of corruption. But it is likely that the corrosive effects of corruption bring down not only buildings, but social institutions as well. It is reasonable to conjecture that corruption destroys both the mortar that holds together physical capital, and the glue that binds social capital.

Corruption has long been a concern of students of international development, but until relatively recently, it was nearly impossible to undertake quantitative work on the topic. Since corruption by definition lacks transparency and is not officially tracked, it is hard to get quantitative evidence about its extent.

In the late 1990s, Transparency International (TI), a non-governmental organization headquartered in Berlin, began to systematically monitor corruption throughout the world. TI exploited a simple fact: There are tens of thousands of people who experience corrupt practices every day.

Many of these people work in multinational organizations (notably businesses), which puts them in a position to compare corruption among countries. From this pool of many people with experience and perceptions of corruption, TI has created the “Corruption Perception Index” (CPI), increasingly the go-to source for quantitative estimates of comparative corruption.

Every year since the late 1990s, TI has issued a report that gives each country that it covers a score, from zero (most corrupt) to ten (least corrupt). Each report is available on its website.

For reasons not entirely clear, TI says that its repeated cross-section surveys are unsuitable for historical work, although, of course, almost all economic and social data are drawn from repeated sampling rather than from data that follow the same sample over many years.

It is not clear why the CPI is less reliable for historical analysis than, for example, the consumer price index (that other CPI) or consumer sentiment. We created a spreadsheet with all 1,784 data points reported from 1998 through 2010, organized by country and by year, on which our analysis is based.

Of course, just because there are numbers that measure perceptions of corruption does not mean they are useful. As TI notes, over this 13-year period, no two years survey the same people in the same countries. Each year, multiple surveys, from varying sources, are compiled, and these differ across years. Therefore, one might conclude that data should not be compared across years.

But by the same token, in any given year, no person is expert on all 85, 133 or 180 countries covered by that year’s survey. We follow assumptions common among those who work with imperfect data from developing areas: First, we rely on TI to provide enough overlap in methods and sample to justify comparisons. Second, we assume that while there certainly is noise in the data, there is enough signal to justify quantitative applications. Finally, we assume it is possible to look statistically for some systematic changes in the meaning of the numbers, which is what we have done.

In this feature, we report on the changing landscape of perceived corruption in a subset of the 186 countries covered in one or more of these 13 surveys. We look at the 74 countries that appear in every survey since 1998, plus 79 additional countries that each have data for no fewer than six years between 1998 and 2010.

Before we summarize changes in perceived corruption, let us note some changes in the sample of countries covered by the survey. The CPI report of 1998 covered 85 countries. Scores ranged from 10 (the best possible score) to 1.4. The average score was 4.9 and the median 4.2. Over the subsequent years, all these numbers changed each year. The number of countries sampled grew, and the maximum score drifted downward — as did the average score, the median score and the variance among scores. No trend is evident in the minimum score.

More careful examination of the distribution of CPI scores year by year shows that between 1998 and 2010, the CPI score that demarcated every decile of the distribution fell. (Most of the decline occurred between 1998 and 2004.) That is, the tenth percentile of the distribution was reached at a lower CPI score, as was the 20th, and so on.

In other words, the population of scores was drifting lower each year, suggesting more and more corruption. However, further examination of the data reveals that this result is due to the addition of more and more countries to the survey — countries that were, on average, perceived as more corrupt than the countries covered in earlier surveys.

When we look at the evolution of perceived corruption in the 74 countries that have been surveyed every year since 1998, in the aggregate, there is no evidence to support a drift toward more corruption. For the countries with complete data, the average and the median CPI score remain virtually unchanged over the period, as do the decile break scores. Year to year, the distribution of scores of the sample of 74 countries is almost identical.

So we conclude that any trend in the Corruption Perception Index for a particular country is likely to result not from a general change in scaling of the CPI (analogous to grade inflation), but from changes in the actual perception of the extent of corruption in that country. (An exception: Countries at the very top of the distribution, with the least corruption, may exhibit a downward trend that simply reflects a change in scoring norms.)

Trends in individual countries

Because the 74 countries showed little change in the CPI when viewed in the aggregate, one might assume there is little change among the 74 countries when viewed individually. One might also reach this conclusion after looking at two or three adjacent years of data for any given country. It would be easy to conclude that corruption in individual countries changes little, year after year.

One would be wrong. In fact, there is a high degree of change in the perceived level of corruption in these countries. Even a period as short as 13 years reveals historical trends.

In order to evaluate trends in perceived corruption in individual countries, we looked at the linear trend line fitted to the CPI score. Of the 153 countries (the 74 with complete data plus 79 more with at least six years of data each), 68 showed no significant trend. Perceived corruption in 40 increased, while in 45 it decreased.

Figure 1 (below) illustrates with six examples, two each from the top, middle and bottom of the distribution. Three of these examples exhibit increasing corruption: the Philippines, Tunisia and the United Kingdom. The other three exhibit diminishing corruption: Indonesia, Turkey and Hong Kong. In all six countries, the direction of the trend (upward or downward) is significant at the 99% confidence level.

FIGURE 1. View a high-resolution version of this graph here.

Data Sources: Bernard Wasow; The Globalist Research Center; Transparency International

Table 1 (below) provides a taxonomy for all 153 countries with significant amounts of data. The columns in Table 1 show the average level of corruption in each of these countries over the period covered. The countries are grouped into quintiles — that is, into groups that each comprise a fifth of all the 186 countries with any data.

Thus, the “bottom quintile” includes the 37 countries that had the lowest average CPI scores (the highest average level of corruption). The rows break down the trend of corruption over the whole period. The trend can be downward, with corruption rising with a 99% confidence level (“deteriorating strongly”), a 95% confidence level (“deteriorating”) or a 90% confidence level (“deteriorating weakly”). If the linear trend is insignificant at the 90% level, the country falls into the “no trend” category. For countries that show diminishing corruption, that conclusion can similarly be strong, medium or weak.

TABLE 1. View a high-resolution version of this table here.

Data Sources: Bernard Wasow; The Globalist Research Center; Transparency International

Surprising trends

Some of these trends might be seen as rather surprising. We will note a few that struck us:

• India, which has shown significant improvement in perceived corruption, may soon pass China, which shows no trend in the level of corruption, in spite of rapid income growth.
• The shift of jurisdiction in Hong Kong from the United Kingdom to China has been followed by a steady improvement in the level of perceived corruption, with Hong Kong now perceived as less corrupt than the United Kingdom (as illustrated in Figure 1).
• A number of countries that one might dismiss as “hopelessly corrupt” because they have remained throughout near the bottom of the distribution, nevertheless have shown steady improvement. See Indonesia, Cameroon, Nigeria, Tanzania, Azerbaijan and Bangladesh.
• A number of countries that have enjoyed favorable reputations internationally have shown steady deterioration in the perceptions of corruption: the United Kingdom, Brazil and Israel are examples.
• Qatar has achieved a quite exceptionally rapid decline in perceived corruption. Since its entry to the survey in 2003, Qatar has passed not only every country above it in the ranking from the Mediterranean region (France, Spain, Cyprus and Israel), but the United States and United Kingdom as well.
• A number of the countries with relatively high and rising corruption have recently suffered political turmoil: Tunisia, Zimbabwe, Iran and Cote d’Ivoire. Other countries with similar conditions regarding corruption include Iraq, Venezuela, Sudan and the Philippines.
• Most of the countries that once made up the USSR show little evidence of leaving the bottom of the corruption perception universe: Russia, Ukraine, Kazakhstan, Uzbekistan, Armenia, Belarus, Moldova and Turkmenistan. The exceptions are the Baltic Republics and Georgia.