Counting Chickens when they Hatch: Timing and the Effects of Aid on Growth

Abstract

Recent research yields widely divergent estimates of the cross-country relationship between foreign aid receipts and economic growth. We re-analyse data from the three most influential published aid‐growth studies, strictly conserving their regression specifications, with sensible assumptions about the timing of aid effects and without questionable instruments. All three research designs show that increases in aid have been followed on average by increases in investment and growth. The most plausible explanation is that aid causes some degree of growth in recipient countries, although the magnitude of this relationship is modest, varies greatly across recipients and diminishes at high levels of aid. Economists have spent decades debating, without resolution, the cross-country relationship between foreign aid receipts and economic growth. Some find that aid robustly causes positive economic growth on average. Others cannot distinguish the average effect from zero. Still others find an effect only in certain countries, such as those with good policies or governance. Wearied readers of this literature would be right to wonder what produces diverse findings from apparently the same aid and growth data. Here, we show that two traits of previous research help to explain why different studies reach different conclusions. Both traits relate to how these studies treat the timing of causal relationships between aid and growth. First, the most cited research has focused on measuring the effect of aggregate aid on contemporaneous growth, while many aid-funded projects can take a long time to influence growth. Funding for a new road might affect economic activity in short order, funding for a vaccination campaign might only affect growth decades later and humanitarian assistance may never affect growth. Second, because current growth is likely to affect current aid, these studies require a strategy to disentangle correlation from causation. They have tended to rely on instrumental variables but the instruments that have been used are of questionable validity and strength. When these issues are addressed, the divergence in empirical findings is greatly reduced. We show this by stepwise altering the research design of the three most influential papers in the aid and growth literature. We hold all else constant: we begin by