NON-LINEAR DYNAMICS AND STRUCTURAL SHIFTS IN ETHIOPIA'S AGGREGATE PRODUCTION FUNCTION: A COMPARATIVE ANALYSIS USING GENERALIZED ADDITIVE MODELS AND STRUCTURAL BREAK APPROACHES

Abstract This study examines the non-linear and time-varying aspects of Ethiopia's aggregate production function, extending beyond the traditional Cobb-Douglas framework. We employ and compare two advanced modelling techniques: Generalized Additive Models (GAM) and Structural Break Analysis, using data spanning four decades (1984-2023). Our GAM results reveal significant non-linearities in the production function, particularly in the labor-output relationship, with labor productivity exhibiting an inverse U-shaped pattern. The capital-output relationship shows a more linear trend with slight diminishing returns at high levels. The Structural Break analysis identifies four critical junctures (1991, 1998, 2007, and 2014) that significantly alter the production function parameters, primarily manifesting as changes in total factor productivity. Both models outperform the classical Cobb-Douglas function, with the GAM demonstrating superior in-sample fit (adjusted R-squared of 0.994 vs. 0.9716 for the Structural Break model), while both show comparable out-of-sample predictive power. Our findings suggest the need for flexible labour market policies, continued emphasis on capital investment, and targeted interventions to enhance total factor productivity. This study contributes to the literature challenging simplistic growth models and offers a more sophisticated understanding of the production process in a rapidly evolving African economy. It also provides a methodological framework for analyzing non-linear and time-varying aspects of production functions, applicable to other developing economies.