Development aid and its impact on poverty reduction in developing countries : a dynamic panel data approach

Abstract

Foreign aid has been used on the one hand by donors as an important international relations policy tool and on the other hand by developing countries as a source of funds for development. Since its inception in the 1940s, foreign aid has been one of the most researched topics in development economics. This study adds to this growing aid effectiveness literature, with a particular focus on the under-researched relationship between foreign aid and extreme poverty. The main empirical assessment is based on a sample of 120 developing countries from 1981 to 2013. The study had two main objectives, namely: (i) to estimate the impact of foreign aid on poverty reduction and (ii) to examine the direction of causality between foreign aid and poverty in developing countries. From these two broad objectives, there are six specific objectives, which include to: (i) examine the overall impact of foreign aid (total official development assistance) on extreme poverty, (ii) investigate the impact of different proxies of foreign aid on the three proxies of extreme poverty, (iii) assess whether political freedom (democracy) or economic freedom enhances the effectiveness of foreign aid, (iv) compare the impact of foreign aid on extreme poverty by developing country income groups, and (v) examine the direction of causality between extreme poverty and foreign aid. To achieve these objectives, the study employed two main dynamic panel data econometric estimation methods, namely the systemgeneralised method of moments (SGMM) technique and the panel vector error correction model (VECM) Granger causality framework. While the SGMM was used to assess the impact of foreign aid on extreme poverty, the panel VECM Granger causality was used to examine the direction of causality between foreign aid poverty. The SGMM was used because of its ability to deal with endogeneity by controlling for simultaneity and unobserved heterogeneity, whereas the panel VECM was preferred because the variables were stationary and cointegrated.