This paper by Ross Levine and David Renelt examines the robustness of empirical links between policy indicators and long-run average growth rates found in cross-country regressions. The authors use a variant of Leamer's extreme bounds analysis to test the sensitivity of coefficient estimates to changes in the list of explanatory variables. They find that while "policy" (broadly defined) appears to be significantly related to growth, there is no strong independent relationship between growth and most existing policy indicators. The paper highlights two main themes: (1) economic policy is related to long-run growth, but (2) the statistical relationships between growth and specific macroeconomic indicators are fragile, with small alterations in other explanatory variables overturning past results. The authors conclude that few findings can withstand slight alterations in the list of explanatory variables, suggesting that the confidence in the conclusions of cross-country growth regressions is limited. The paper also discusses the robustness of findings to different sets of included variables and the interpretation of policy indicators.This paper by Ross Levine and David Renelt examines the robustness of empirical links between policy indicators and long-run average growth rates found in cross-country regressions. The authors use a variant of Leamer's extreme bounds analysis to test the sensitivity of coefficient estimates to changes in the list of explanatory variables. They find that while "policy" (broadly defined) appears to be significantly related to growth, there is no strong independent relationship between growth and most existing policy indicators. The paper highlights two main themes: (1) economic policy is related to long-run growth, but (2) the statistical relationships between growth and specific macroeconomic indicators are fragile, with small alterations in other explanatory variables overturning past results. The authors conclude that few findings can withstand slight alterations in the list of explanatory variables, suggesting that the confidence in the conclusions of cross-country growth regressions is limited. The paper also discusses the robustness of findings to different sets of included variables and the interpretation of policy indicators.