The aid effectiveness literature contains about 100 papers that see aid as a treatment given to poor countries to generate development. 68 of these papers provide a total of 543 comparable estimates of the effect of aid on growth, which are the data of our meta-analysis. We consider two questions: (Q1) Are the estimates converging to a clear result over time as aid agencies gain experience, models become better and data accumulates? We find that the results do have a positive average, but it is small, insignificant and falling. (Q2) Can we identify the main factors that explain the large differences in the results? We find that much of the variation between studies can be attributed to publication outlet, institutional affiliation, data and specification differences. However, some of the difference between studies is real. In particular, the aid-growth effect is stronger for Asian countries. The meta-analysis indicates also the existence of indirect channels, which need to be further explored.