Aid, Policies, and Growth: Reply By CRAIG BURNSIDE AND DAVID DOLLAR*

dard regression techniques from the growth lit-erature to measure the effect of foreign aid on growth. The main finding in our paper was that the effect of foreign aid on growth depended on the macroeconomic policies of recipient coun-tries. In this issue, William Easterly et al. (2004), challenge the robustness of our result to new data. Before commenting on their findings it is useful to review the basis of our original findings. Our paper focused on three versions of a panel growth regression, estimated using data for 51 countries, and six four-year periods, from 1970 to 1993. These regressions may be sum-marized as: (1) git other terms 1ait error (2) git other terms 1ait 2aitpit error (3) git other terms 1ait 2aitpit 3ait 2 pit error. In the absence of interaction terms, as in (1), aid appeared to be ineffective: our estimate of 1 was insignificantly different from 0. When we estimated (2), 1 and 2 both appeared to be insignificant. However, when we estimated (3), we found that 2 was significantly greater than zero, and 3 was significantly less than zero. Given the typical values of ait and pit in our sample, these results implied that for countries with “good ” economic policies (high values of pit) the derivative of git with respect to ait was positive. We also concluded that the insignifi-cance of 2 in (2) and the significance of 3 in (3) were both the result of outliers for which ait pit was several standard deviations above the sample mean. If we reestimated (2) with these outliers excluded, we found that 2 was signif-icantly greater than zero. Our results were sim-ilar for OLS and 2SLS, and were robust to a number of unpublished changes in specification. We have many critics; to name just a few