Problems with pooling in panel data analysis for developing countries : the case of aid and trade relationships

It is common practice in empirical development macroeconomics to use cross-country samples for econometric analyses. One issue that is rarely addressed in this literature is the appropriateness of pooling when panels are used. In particular, does it matter to the results if the countries exhibit different time series properties in respect of the relationship being observed? This is the issue addressed in this paper, taking the example of the "aid and trade" relationship. We also address the following questions: (a) Is there any support for the assertion that donors use aid to perpetuate their trade with recipients? (b) How important is trade in influencing aid allocation decisions of donors? The principal result is that pooling countries with different underlying time series properties (different inherent causality) is inappropriate and gives misleading results. In other words large samples are not necessarily the best samples. Using appropriate samples, we find no evidence that tied aid increases trade, although donors providing a higher share of aid tend to trade more with the recipient. In terms of aid allocation, donors appear to be concerned with relative aid and trade shares rather than absolute volumes.