Mitigating the Effects of Multicollinearity Using Exact and Stochastic Restrictions: The Case of an Aggregate Agricultural Production Function in Thailand: Comment

Abstract

Mittelhammer, Young, Tasanasanta, and Donnelly in estimating an aggregate agricultural production function use a variety of techniques to deal with multicollinearity present in their sample, including exact and stochastically restricted least squares and principal components regression.' They claim that consideration of these techniques allowed mitigation of a serious multicollinearity problem present in their data and permitted more precise parameter estimates. The authors also hold that two of the techniques they considered generally outperformed OLS in terms of risk and overall reasonableness ... (p. 199). While the empirical evidence Mittelhammer et al. present is convincing and the product of sound econometrics, we feel that some important limitations of the estimation techniques they employed as alternatives to ordinary least squares (OLS) were not sufficiently stressed. Furthermore, we will provide estimators that are unambiguously superior to least squares under a variety of standard loss functions.