3regression line. The last column also reports the
value associated with the null hypothesis of a sign testthat countries in Latin America and the Caribbean are evenly distributed above and below the regressionline for the entire sample. For the first four indicators we mention, the noted patterns are stronglystatistically significant at conventional levels, with Latin America and the Caribbean faring well in thefirst indicator (political accountability), but poorly in the following three (government effectiveness, ruleof law and control of corruption).
Table 1. Governance and Per Capita Incomes in Latin America and the Caribbean
Position relative to OLS regression line
Measure of institutional quality
Below Above P value
Voice and accountability 8 18 0.05Political stability 11 15 0.43Government effectiveness 19 7 0.02Regulatory quality 10 16 0.24Rule of law 20 6 0.01Control of corruption 20 6 0.01Overall governance 16 10 0.24
a. The first two columns summarize the location of countries in Latin America and the Caribbean relative tothe regression line in a simple OLS regression of the indicated measure of governance on log per capitaGDP at purchasing power parity (PPP) in 1995. The final column reports the
value associated with a signtest of the hypothesis that the proportions of countries in Latin America and the Caribbean falling above andbelow the regression line are equal.
This paper provides an interpretation of the strong positive correlation between governance andper capita income shown in figure 1. The following section describes how we constructed the governanceindicators for 2000–01 used in figure 1 as part of an ongoing project to measure governance worldwide,drawing on a wide variety of sources of data on perceptions of governance and a consistent aggregationmethodology. In addition to allowing us to construct a set of governance indicators covering a very largesample of countries, our methodology allows the construction of measures of the precision of thesegovernance indicators. As we have found in previous work, the standard errors associated with estimatesof governance are large relative to the units in which governance is measured, suggesting that simplecountry rankings can be misleading and underscoring the need for caution in making precise comparisonsof the quality of governance across countries using this type of data.The next section turns to an interpretation of the positive correlations between per capita incomeand governance presented in figure 1. This correlation can, in principle, reflect some combination of (a)causal effects running from better governance to higher per capita income, (b) reverse causation orfeedback from higher per capita income to better governance, and (c) omitted variables which improveboth governance and per capita income. A recent body of literature identifies large causal effects runningfrom governance to per capita income, using deep historical determinants of institutional quality as