Measuring Governance Using Cross-Country Perceptions Data

This paper describes an ongoing project to measure governance using crosscountry perceptions data. The governance indicators measure six dimensions of governance and cover 209 countries and territories for 1996-2004. They are based on several hundred individual variables measuring perceptions of governance, drawn from 37 separate data sources constructed by 31 different organizations. We present the estimates of governance, and the margins of error capturing the range of likely values for each country. We show how these margins of error should be taken into account when considering cross-country differences and changes over time in governance. We find that in a number of countries the quality of governance improved significantly in the short term. Yet deteriorations also took place in some other countries, while in many there was little change. There has been no worldwide improvement in governance on average. We argue that perceptions-based data provide valuable insights relative to objective data on governance, and that individual objective measures of governance provide an incomplete picture of even the quite particular dimensions of governance that they are intended to measure. We also show that margins of error are not unique to perceptions based measures of governance, but are an important feature of all efforts to measure governance, including objective indicators. We also empirically investigate the importance of ideological biases in expert assessments of corruption and find little evidence that they are present. Governance indicators and per capita incomes are highly correlated across countries. Recent research shows that this correlation captures an important causal effect running from measures of governance such as these to per capita incomes. Critics of this view argue that the correlation captures substantial reverse causation from incomes to governance, and is tainted by "halo effects" where rich countries receive good ratings simply because they are rich. We review available evidence on these two critiques and find it to be lacking.