Estimating production functions through additive models based on regression splines

• DOI: 10.1016/j.ejor.2023.06.035
• Publication Date: 2023-06-24T18:45:22Z
• Authors: Victor J. España; Juan Aparicio; Xavier Barber; Miriam Esteve
• Keywords: Additive model; Overfitting; Production theory; Interpretability

This paper introduces a new methodology for the estimation of production functions satisfying some classical theory axioms, such as monotonicity and concavity, which is based upon adaptation an additive version machine learning technique known Multivariate Adaptive Regression Splines (MARS). The approach shares piece-wise linear shape estimator associated with Data Envelopment Analysis (DEA). However, able to surmount overfitting problems DEA by resorting generalized cross-validation. In this paper, computational experience was employed measure how well performs, showing that it can reduce mean squared error bias true function in comparison more recent Corrected Concave Non-Parametric Least Squares (C2NLS) methodology. We also show success depends on whether or not interactions among variables prevail degree non-additivity be estimated.