Why Run a Million Regressions? Endogenous Policy and Cross-Country Growth Empirics

This paper analyzes the link between long-run economic growth and policy that depends on important economic fundamentals. In simple growth regressions the measured policy effects would then generally be biased. Using a widely quoted theoretical model, the signs of the biases are derived. It is shown that the usually reported effects on growth of tax-rate variables related to GDP - the ratio of public to total investment and that of redistributive transfers to GDP - would generally be biased downwards. The signed biases provide a theory-based lens through which some empirical, yet theoretically puzzling, results can be viewed and subsequently discussed.