Past research on aid and growth is flawed because it typically examines the impact of aggregate aid on growth over a short period, usually four years, while significant portions of aid are unlikely to affect growth in such a brief time. We divide aid into three categories: (1) emergency and humanitarian aid (likely to be negatively correlated with growth); (2) aid that affects growth only over a long period of time, if at all, such as aid to support democracy, the environment, health, or education (likely to have no relationship to growth over four years); and (3) aid that plausibly could stimulate growth in four years, including budget and balance of payments support, investments in infrastructure, and aid for productive sectors such as agriculture and industry. Our focus is on the third group, which accounts for about 53% of all aid flows. We find a positive, causal relationship between this "short-impact" aid and economic growth (with diminishing returns) over a four-year period. The impact is large: at least two-to-three times larger than in studies using aggregate aid. Even at a conservatively high discount rate, at the mean a $1 increase in short-impact aid raises output (and income) by $1.64 in present value in the typical country. From a different perspective, we find that higher-than-average short-impact aid to sub-Saharan Africa raised per capita growth rates there by about half a percentage point over the growth that would have been achieved by average aid flows. The results are highly statistically significant and stand up to a demanding array of tests, including various specifications,endogeneity structures, and treatment of influential observations. The basic result does not depend crucially on a recipient's level of income or quality of institutions and policies; we find that short-impact aid causes growth, on average, regardless of these characteristics. However, we find some evidence that the impact on growth is somewhat larger in countries with stronger institutions or longer life expectancies (better health). We also find a significant negative relationship between debt repayments and growth. We make no statement on, and do not attempt to measure, any additional effect on growth from other categories of aid (e.g., emergency assistance or aid that might affect growth over a longer time period); four-year panel regressions are not an appropriate tool to examine those relationships
