|
Many empirical studies have found that the quality of governance has an important effect on investment and on economic growth (Aron 2000). The measurement of governance quality and performance is an inexact science though, and a wide variety of sources have been used to examine linkages to investment. The most comprehensive and reliable source of information is the so-called "KK" indicators, which were first released by Kaufmann, Kraay, and Lobatón (1999a and b).
The newly released KK survey data (Kaufmann, Kraay, and Mastruzzi [KKM] 2006) updates indicators to 2005, and revises earlier estimates to take account of new information and the inclusion of a larger number of countries in later samples. Consistent and revised data are available for 1996, 1998, 2000, and for 2002-2005.
Drawing together information from over 200 different sources, the KK indicators measure governance performance in six separate dimensions: voice and accountability; political stability; government effectiveness; regulatory quality; the rule of law; and control of corruption. The KK scores are distributed normally around a mean of zero with a standard deviation of one. This means that virtually all scores lie in a range from -2.5 to +2.5, with larger values signifying a better score. Estimates of the accuracy of the indicators are also available and these suggest that accuracy has improved.
A particularly useful feature of the KK indicators is that they allow international comparisons on a consistent basis. In Figure 1.4.18 a comparison of ratings on each indicator for each country in 1996 and 2005 is shown. The vertical lines show the 90% confidence intervals constructed using estimated standard errors from 2005. It is striking that out of 30 comparisons, governance ratings have fallen in 22 cases. For all countries, other than Korea, raw scores fall in three or more dimensions. On regulatory and corruption indicators there is a general pattern of deterioration, with only two exceptions (Thailand on corruption, Korea on regulation). Korea stands out as having improved in four of six possible categories. If ranks are compared across time, ranks fall in 28 cases. In 1996, the crisis countries were in the top half of all countries in 21 out of the 30 indicator values, but by 2005 they ranked in the top half in 17 indicators.
But are these changes in governance performance statistically significant? Of the 22 cases where governance scores get worse, 15 differences lie outside a 90% confidence interval using 2005 standard errors. In 4 of the 8 cases of improvement, ratings lie outside the 90% confidence interval. T-tests of the statistical significance of differences suggest that 5 differences are significant at 90% (1 of which is an improvement). At a 75% confidence level, this number rises to 8.
A stricter measure still of the significance of differences is whether each of the paired scores lie outside each other's 90% (or 75%) confidence intervals. KKM (2005) observe that this test emulates a test for differences using a dynamic unobserved components estimator. At the 90% confidence interval, out of 30 comparisons, there are only 3 pairs that meet this criteria (including 1 improvement) and at the 75% interval this rises to 9 pairs (including 3 improvements).
KKM (2005) note that in the global sample, the proportions of changes in ratings that lie outside each other's 90% (75%) confidence interval is quite small. In the global sample, only 8.5% of changes qualify at the 90% confidence interval and 19.3% at 75% confidence interval. In the sample of crisis countries, the corresponding numbers are 10% and 30%. On balance, the KK measures suggest deterioration both absolutely and in terms of comparisons with other countries. However, it is possible that some of the observed changes may have occurred by chance.
|
Top of page