Efficient Estimation of Structural Hedonic Systems
指出传统特征定价模型中边际价格作为工具变量的方法存在不一致性，提出在缺乏标准工具变量时仍可识别系统，并开发了高效估计量及规范检验方法。
When a product or job is not homogeneous but instead embodies several different characteristics, it is difficult to use price information about the product to learn anything about the structure of the supply and demand for the embodied characteristics. S. Rosen (1974) highlighted this problem and suggested that the estimated slope of the price/characteristic relation (the marginal price) could be used instead of prices in otherwise estimation of the supply and demand system. However, recently Brown and H. Rosen (1982), Brown (1983), and Mendelsohn (1985) have pointed out that this method is applicable only when the marginal price depends on variables not found in the supply and demand equations. In this paper, we show that since the hedonic system matches consumers with high demand for the characteristic to firms with low costs of producing it, exogenous variables from the supply equation are not valid instruments in the demand equation and vice versa. Therefore, even when the Brown/Rosen/Mendelsohn (BRM) condition holds, estimation of the identified system by simultaneous equations techniques will produce inconsistent estimates. We then show that the system can be identified despite the absence of the standard instruments, and develop an efficient estimator for the identified system. Since cross-equation constraints make estimation of hedonic systems particularly vulnerable to specification error, we also consider some specification tests. In this section, we consider a simple model with only one characteristic. The extension to several characteristics is straightforward. The basic model developed by Rosen can be described as follows. Firms with different technologies produce products which provide consumers with varying levels of some characteristic Z. Each consumer buys one unit of the good. The product is such that an amount of Z from two units of the good provides less utility than the same amount of Z embodied in a single unit (e.g., acceleration in a car, safety in a job). Consumers'