A Semiparametric Panel Generalized Additive Model for Forecasting GDP Growth and Inflation in African Economies

Macroeconomic relationships in African economies are often nonlinear and heterogeneous, limiting the adequacy of conventional linear models for analysing and forecasting GDP growth and inflation. This study applies a semiparametric Panel Generalized Additive Model (Panel GAM) to quarterly panel data for 53 African economies over 2005Q1–2025Q4, estimating smooth nonlinear effects while accounting for temporal variation and country-specific heterogeneity. The estimated smooth terms indicate statistically significant nonlinear relationships for both outcomes. The model explains a larger share of the variation in the inverse hyperbolic sine-transformed Consumer Price Index than in GDP growth (adjusted R² = 0.671; deviance explained = 67.7% versus adjusted R² = 0.236; deviance explained = 25.1%), suggesting that the selected macroeconomic variables are more strongly associated with inflation dynamics than with economic growth during the study period. Out-of-sample results indicate that the framework can generate forecasts for both outcomes, while providing interpretable smooth functions. These results may inform policy analysis by highlighting the relevance of nonlinear responses and cross-country heterogeneity, particularly for inflation. The study contributes an interpretable semiparametric panel approach that complements Panel Vector Autoregressive and Panel Multi-Output Gaussian Process Regression models.