Nonlinearity in the efficacy of foreign aid and evidence of poverty traps

Research on the efficacy of foreign aid has been extensively conducted over the last three decades, but empirically the results remain inconclusive. Previous literature has relied upon linear growth model to estimate their models, methods inconsistent with the existence of poverty traps. These approaches have found that the effect of foreign is either negative, to positive, or even conditional on the policy environment in the country. More recent theoretical literature has started to point towards a nonlinear effect of foreign aid on the level of income. This nonlinear effect is consistent with a poverty trap model. To provide empirical support for this theory this paper relaxes the assumption of linearity in the model to determine if the efficacy of foreign aid is nonlinear. To accomplish this the relationship between foreign aid and the level of income is examined using a cross sectional data set of 72 countries that received Official Development Aid between 1985 and 2014 using a multivariate adaptive regressive splines (MARS) estimation. The MARS model is a nonlinear approach that systematically identifies both threshold levels and interaction terms between independent variables. Most importantly, the MARS model will detect if foreign aid has a statistically significant threshold effect. This finding would be consistent with the poverty trap hypothesis and also explain why there has been a lack of consensus so far. The results suggest that by using a nonlinear estimation we significantly improve, 18 to 44 percent reduction in the residual sum of squares, the fit of our growth model compared to an ordinary least squares estimation, as well as also detect a threshold in the effect of foreign aid that is consistent with a poverty trap hypothesis. These results are consistent when we analyze foreign aid as either a percent of gross national income—threshold at 4.6 percent—or per capita—threshold at $24.97. Below these thresholds the effect of foreign aid is either negative, or statistically zero, but when foreign aid is sufficiently large the effect is large and positive. We also detect significant interactions between foreign aid and both human capital and population growth. The policy implication being that if a country is going to experience a positive impact on their level of income the amount of foreign aid needs to be sufficiently large enough. Furthermore, the interactions suggest that some degree of selectivity should take place on which countries are best equipped to use aid.