Endogeneity Corrected Stochastic Production Frontier and Technical Efficiency

Abstract

A major econometric issue in estimating production parameters and technical efficiency is the possibility that some forces influencing production are only observed by the firm and not by the econometrician. Not only can this misspecification lead to a biased inference on the output elasticity of inputs, but it also provides a faulty measure of technical efficiency. We extend the Levinsohn and Petrin (2003) approach and provide an estimation algorithm to overcome the problem of endogenous input choice in stochastic production frontier estimation by generating consistent estimates of production parameters and technical efficiency. We apply the proposed method to a plant-level panel dataset from the Colombian food manufacturing sector for the period 1982-1998. This dataset provides the value of output and prices charged for each product, expenditures and prices paid for each material used, energy consumption in kilowatt per hour and energy prices, number of workers and payroll, and book values of capital stock. Empirical results find that the traditional stochastic production frontier tends to underestimate the output elasticity of capital and firm-level technical efficiency. The evidence in this research suggests that addressing the endogeneity issue matters in stochastic production frontier analysis.

Estimating Food Waste as Household Production Inefficiency

“…Maximum likelihood methods are studied in Kutlu () and Amsler, Prokhorov, and Schmidt (). Shee and Stefanou () consider using a proxy variable in a panel data estimation, and Tran and Tsionas () develop a copula approach without requiring external instruments. We follow the LIML method in Amsler, Prokhorov, and Schmidt () by first adding household average employment status as a proxy variable for physical activities to the right‐hand side of equation .…”

Feed use intensification and technical efficiency of dairy farms in New Zealand

“…Random shocks, which are observed by farmers but not econometricians, may affect both farm input choices and dairy productivity. Optimising farmers may adjust input levels in response to productivity shocks that they observe, but which are unobserved to the statistician, resulting in a simultaneity problem in the estimation of the production frontier (Shee and Stefanou ). Ignoring such endogeneity issues altogether could result in biased parameter estimates and incorrect statistical inference.…”

Decent rural employment and farm production efficiency: empirical evidence from Tanzania and Ethiopia

“…Second, some inputs in the right‐hand side of the estimation (Eq. ) might be endogenous (Kumbhakar, ; Kutlu, ; Shee and Stefanou, ; Tsionas et al., ). Land is owned by the state in both countries, and farmers are less likely to adjust farm land to productivity shocks.…”

“…If such an adjustment exists, the model is likely to suffer from endogeneity bias. To control for such unobserved effects, various econometric approaches are suggested often with the application of instrumental variables (IVs) (Kim and Kim, ; Kutlu, ; Shee and Stefanou, ; Wooldridge, ). As smallholder farms labor often consist of family labor for agricultural production activities, family size can be used as an instrument for labor input.…”