The Air Pollution and Property Value Debate: Some Empirical Evidence

Abstract

In recent issues of this REVIEW, Myrick Freeman and Kenneth Small have helped to clarify some confusion which grew out of a study of air pollution and property values by Ridker and Henning (1967). Ridker and Henning (hereafter referred to as R-H) performed a multiple regression of property values on air pollution, housing characteristics, accessibility, and neighborhood characteristics using cross-section data for the St. Louis metropolitan area. As Freeman (1974a) and Small (1975) point out, the early debate on the R-H study was largely sidetracked with the issue of whether one can predict the aggregate change in property values resulting from an improvement in urban air quality. A more appropriate focus for policy purposes is the question of whether property value data can be used to estimate households' willingness to pay for cleaner air. The R-H study calculated willingness to pay by first multiplying the air pollution coefficient obtained in their linear regression by the change in air pollution estimated for each household, and then summing over all St. Louis households. The air pollution coefficient was thus inteipreted as the average willingness to pay for air quality improvements for all St. Louis households. Freeman (1971) had correctly pointed out in an earlier note that the R-H procedure of calculating willingness to pay is inappropriate because a hedonic housing regression cannot in itself isolate demand and supply elements; and Freeman was not very sanguine about the possibilities of doing so. Without taking a position on whether the true willingness to pay for clean air can be determined empirically, both Freeman and Small put the earlier empirical work into perspective by showing that the pollution coefficient in the housing value regression does provide a correct measure of the 'mnarginal valuation placed by individual consumers on uniform changes in pollution levels (Small, 1975, p. 105). As a theoretical point, this result is reassuring, since it suggests that benefit estimates obtained from studies like that of Ridker and Henning may be reasonable if the air quality improvement is small. Unfortunately, however, the air quality improvements generated by the Clean Air Act Amendments of 1970 and 1977 are distinctly nonmarginal (pollution reductions are often predicted to be 50% or more). No empirical evidence has been available thus far to assess the magnitude of the bias associated with using marginal valuations to estimate the benefits from these nonmarginal pollution reductions.' This note attempts to fill this gap. Several authors have argued as a matter of principle that the willingness to pay for nonmarginal air quality improvements can be obtained from property value data if the proper conceptual procedure is used.2 Using the approach suggested by these papers and an empirical focus suggested by Rosen (1974), we have elsewhere provided estimates of households' willingness to pay for clean air as well as details concerning the nature of the data and possible specification biases (Harrison and Rubinfeld, 1978). In this paper we confine our discussion to the magnitude of the bias that results from using the air pollution coefficient in the hedonic housing price equation directly to estimate the willingness to pay of urban households for a nonmnarginal air quality improvement. Section II summarizes the model we use to estimate the demand for nonmarginal changes in air quality. In section III we distinguish three potential sources of bias in the R-H procedure, and assess the quantitative importance of each with data for the Boston Metropolitan area.