This paper investigates the determinants and impact of intra-regional trade within the three largest African trade blocs the Common Market for Eastern and Southern Africa (COMESA), the East African Community (EAC), and the Southern African Development Community (SADC) using Bayesian Model Averaging (BMA) on a panel of 26 member states from 1995 to 2022. Conventional gravity-model specifications are embedded in a model space that allows for parameter, functional-form, and multilateral-resistance uncertainty.

Posterior inclusion probabilities reveal that traditional gravity covariates (GDP, distance, adjacency, common language) remain dominant, but institutional quality, infrastructure indices, and preferential trade agreement (PTA) depth indicators exhibit high posterior support (> 0.85). Surprisingly, exchange rate volatility and tariffs display low inclusion probabilities once PTA depth is accounted for, suggesting that rules of origin and behind-the-border measures now drive trade costs.

Out-of-sample predictive checks show that BMA specifications reduce root-mean-square forecast errors by 11–18% relative to standard fixed-effects models. Counterfactual simulations indicate that a 10% improvement in PTA depth could raise intra-bloc trade by 6.4% for COMESA, 8.1% for EAC, and 5.7% for SADC, with larger gains accruing to landlocked members. Heterogeneity tests reveal that the trade enhancing effect of infrastructure is strongest where governance quality exceeds a threshold of –0.5 on the World Governance Indicators index.

Overall, the results provide robust evidence that deep integration rather than shallow tariff liberalisation offers the most promising route to expanding trade among COMESA, EAC, and SADC economies.