A Reply to “A Replication of “Counting Chickens When They Hatch†(Economic Journal 2012)â€

The regressions in Clemens et al. (2012) are fully replicable with open-access data and code. Roodman (2015) alters the regression specifications in that paper by adding twice-lagged aid, after which he cannot reject the null hypothesis of a zero effect of aid on growth. We show, with Roodman’s data and code, that his altered specifications have very low power to reject the null–roughly 0.1 to 0.2. In other words, there is an 80-90% chance that Roodman’s altered regressions fail to reject the null by construction. This renders the exercise uninformative about the robustness of the findings in Clemens et al. (2012) or, more generally, about the effect of aid on growth.