A1595
Title: Stochastic Frontier Panel Model with Endogeneity in Inputs and Environmental Variables
Authors: Yuanlu Suo - Capital University of Economics and Business (China) [presenting]
Taining Wang - Capital University of Economics and Business (China)
Subal Kumbhakar - State University of New York at Binghamton (United States)
Abstract: We propose a semiparametric panel stochastic frontier model that accommodates endogenous inputs and environmental variables with fixed effects. Environmental variables enter the frontier parametrically by modifying input elasticities and affect inefficiency nonparametrically through unknown nonnegative functions. We address input endogeneity internally, without external instrumental variables (IVs), by modeling firms' input choices as outcomes of profit maximization. This yields a system of nonlinear endogenous share equations with cross-equation dependence that identifies the frontier parameters. We address endogeneity in environmental variables externally through a flexible control-function approach, where residuals from semiparametric IV regressions enter the model via unknown correction functions. These two strategies eliminate endogeneity without imposing distributional assumptions or relying on linear IV systems. We develop a two-step semiparametric estimator, where the first step employs a semiparametric three-stage least squares estimator for the frontier parameters, and the second step applies a constrained kernel-backfitting estimator for the inefficiency functions under nonnegativity constraints. We show that both estimators are consistent and asymptotically efficient under large-n, fixed-T asymptotics, and we demonstrate their strong finite-sample performance through simulations.
