UNDP (2008)
A Users’ Guide to Measuring Corruption

This book is targeted at national stakeholders, donors and international actors involved in corruption measurement and anti-corruption programming. It explains the strengths and limitations of different measurement approaches, and provides practical guidance on how to use the indicators and data generated by corruption measurement tools to identify entry points for anti-corruption programming.

Uses and Abuses of Governance Indicators
Arndt, Christiane and Oman, Charles (2006)

This study (available for purchase from OECD) seeks to clarify current trends in the use and misuse of governance indicators as these indicators are applied to developing countries. It includes an in-depth analysis of the most carefully constructed and widely-used governance indicators, those produced by Daniel Kaufmann and his team at the World Bank Institute. The paper argues that composite perceptions-based indicators lack transparency and comparability over time, suffer from selection bias, and are not weel suited to help developing countries identify how effectively to improve the quality of local governance. Fact-based indicators are not necessarily more objective. The authors argue that governance indicators should be based on publicly-available data sets derived from facts, experiences and/or perceptions of diverse, clearly-defined population groups both within and outside the country in question

Measuring Corruption in Eastern Europe and Central Asia: A critique of the cross-country indicators
Knack, Stephen (2005)

This paper assesses corruption levels and trends among countries in the transition countries of Eastern Europe and Central Asia. One interesting finding was the lack of correlation between corruption in public procurement reported by firms and broader, perception-based indicators. In addition to an analysis of corruption measurement in the region, the author usefully includes a general "primer on corruption indicators" which outlines definitional and methodological differences between data sources. Composite indices are conceptually less precise than single sources. A major problems is that many give more weight to sources that correlate highly with each other. In the case of expert surveys, high correlation is a natural result of the fact that "experts" read the same analyses as well as each others' rankings. The author argues for broader use of firm, household and public official surveys to identify more specific corruption problems for programming purposes.