Revisiting the effect of regional integration on African trade : evidence from meta-analysis and gravity model

Abstract

Two main shortcomings flaw the estimation of gravity model in previous studies that examined the trade-creating effects of African regional Trade Agreements (RTAs). First, these studies fail to account for the multilateral resistance term (MRT). This omission makes the estimates from standard gravity model bias and inconsistent. Second, there is a significant proportion of zero trade flows, however, these studies also fail to account for them properly. They use either the Tobit model or replace zero flows with arbitrary small values. Apart from these problems, they also exhibit considerable heterogeneity in the RTA effects on trade. In this paper, a meta-analysis of previous empirical studies is conducted to derive a combined effect size and also explain heterogeneity in RTA effects. In addition, I use the gravity model to compare the trade-creating effect of the main African RTAs. Using the gravity model, I compare the estimation methods of previous studies to the Poisson pseudo-maximum-likelihood estimator that tackles the zero flows. From the meta-analysis, I find a general positive effect of African RTAs of about 27-32% after correcting for publication bias. The source of upward bias is not limited to publication selection as the RTA effects tend to be significantly overestimated when zero flows and MRT are not controlled for properly. A comparative assessment of the RTAs shows a striking heterogeneity.