What can we learn about country performance from conditional comparisons across countries?

There have been many attempts to infer latent performance attributes of governments (or other institutions) from conditional comparisons that control for observed variables. Success in doing so could greatly improve government performance. This paper critically reviews the econometric foundations of the methods used. It is argued that latent heterogeneity remains a fundamental but unresolved identification problem. Locating a benchmark for measuring performance adds a further problem. Current methods do not yield a consistent estimate of even the mean latent performance attribute. An assessment of country performance by these methods could well be wildly wrong.