Hedonic price index estimation under mean‐independence of time dummies from quality characteristics

摘要: Summary. We estimate hedonic price indices (HPI) for rental offices in Tokyo the period 1985–1991. take a partially linear regression (PLR) model, x (year dummies) and nonparametric z (office quality characteristics), as our main model; usual model is used well. Since consists of year dummies, linearity not restriction PLR only that no interaction between z. For HPI are estimated -consistently with two-stage procedure. data, turns out to be (almost) mean-independent This implies least squares estimation (LSE) models misspecified function still consistent. The mean-independence also leads an efficiency result that, under heteroskedasticity unknown form, estimator at efficient any LSE specifying (rightly or wrongly) part In addition these, several interesting practical lessons noted doing estimation. First, cross validation (CV) literature can fail if ignored. Second, high order kernels make CV criterion ill behaved. Third, product work well spherically symmetric kernels. Fourth, specification tests may poorly due sample splitting problem outliers data choosing more than one bandwidth; this regard, test suggested by Stute (1997) et al. (1998) recommended.