Discussion: The Implications of Homothetic Separability for Share Equation Price Elasticities
指出Kohler关于同质可分离效用（HSU）必然导致价格弹性估计有偏的结论是错误的，并推导了HSU下份额方程弹性的正确含义，同时通过蒙特卡洛实验揭示偏差源于支出份额方程估计而非HSU本身。
The assumption of homothetically separable utility (HSU) has been widely used in many recent studies of consumer demand. Atkinson (1979,1981), Caves and Christensen (1980), and Lau and Lillard (1979) maintain rather than test HSU in estimating price elasticities for electricity time-of-use pricing experiments, because of the nature of the data collected. Maintaining HSU when it cannot be supported by the structure of consumer preferences may lead to substantial biases in estimated elasticities. However, Kohler purports to show systematic biases for price elasticities estimated in an HSU system, even when it is correctly imposed, if estimation is carried out using a system of expenditure share equations. The purpose of this discussion is first to show that Kohler's proposition regarding HSU is incorrect, and then to derive the correct implications for elasticities computed from share equations under HSU. In Section 2 of this discussion, I show that Kohler's analytical conclusion that HSU guarantees biased price elasticities is erroneous. In Section 3, I examine Kohler's Monte Carlo results, showing that they are nonrobust and incorrectly characterize an HSU demand system. I then present the correctly specified Monte Carlo experiment, which raises some problems regarding Type I error in share equation estimation. The potential bias derives not from estimation of an HSU system, but rather from estimation of a system of expenditure share equations, where prices appear on both sides of the equation.