The Effect of Rent Control on the Price of Rental Housing: Comment
评论了Marks（1984）使用特征价格法分析租金管制对住房特征隐含价格影响的研究，指出其因样本选择偏差（管制与非管制单元建于不同时期）导致结论不可靠。
In a recent issue of this journal, Marks (1984) uses a hedonic approach to analyze the effect of rent control on the implicit prices of various housing characteristics. Since a rent control policy may cause the implicit prices and quantities of housing characteristics to adjust marginally, a hedonic approach can be especially useful for disentangling these price and quantity effects. Separating his Vancouver sample of 3,885 rental units into rent-controlled (3,198 units) and uncontrolled (687 units) subsamples, Marks employs both an F-test for structural homogeneity between the two hedonic regression equations and a means replacement technique. The empirical findings appear to suggest that implicit prices are significantly higher in the uncontrolled subsample. Furthermore, the results appear to indicate that a rent control policy leads to a lower rent level. While the value of Marks's methodology is unquestionable, the validity of his empirical findings is suspect due to a severe sample selection bias. According to Marks (1984, 82), the majority of the uncontrolled units (98 percent) are not subject to rent control because all apartments built after 1974 were exempt from rent control for a period of five years so that such units would be exempt during 1978 the year being studied. As a result of this dichotomous distinction between controlled and uncontrolled units based on the year of construction, before or after 1974, an examination of Table 4 reveals that the mean age for a controlled unit is approximately 17 years; however, the mean age for an age-exempt uncontrolled unit is only about 4 months. In other words, the average controlled unit was built in 1961, whereas the construction date for a typical uncontrolled unit was some time in the period 1977-1978. In effect, Marks's study compares two samples of rental housing which were constructed at different points in time rather than at a similar or related point in time. Moreover, since the two housing samples reflect construction at different points in time, the implicit market prices for various housing characteristics may also be significantly different across the samples because the market conditions for various housing characteristics may have changed from one period to the next.