Estimation of Industrial Heterogeneity from Maximum Entropy and Zonotopes Using the Enterprise Surveys

Abstract

This study introduces a novel framework for estimating industrial heterogeneity by integrating maximum entropy (ME) estimation of production functions with Zonotope-based measures. Traditional production function estimations often rely on restrictive parametric models, failing to capture firm behavior under uncertainty. This research addresses these limitations by applying Hang K. Ryu's ME method to estimate production functions using World Bank Enterprise Survey (WBES) data from Bangladesh, Colombia, Egypt, and India. The study normalizes entropy values to quantify heterogeneity and compares these measures with a Zonotope-based Gini index. Results demonstrate the ME method's superiority in capturing nuanced, functional heterogeneity often missed by traditional techniques. Furthermore, the study incorporates a "Tangent Against Input Axes" method to dynamically assess technical change within industries. By integrating information theory with production economics, this unified framework quantifies structural and functional differences across industries using firm-level data, advancing both methodological and empirical understanding of heterogeneity. A numerical simulation confirms the ME regression functions can approximate actual industrial heterogeneity. The research also highlights the superior ability of the ME method to provide a precise and economically meaningful measure of industry heterogeneity, particularly for longitudinal analyses.