Selectivity Bias in Urban Land Value Functions

Existing studies of land values in urban areas can be grouped into four broad categories: (1) those that combine several land uses and estimate a single function without distinguishing the type of land use (Glickman 1979; Kau and Sirmans 1979; Mills 1969; Yeates 1965); (2) those that combine several land uses and estimate a single function but include dummy variables for zoning (Chicoine 1981; Hushak 1975; Hushak and Sadr 1979; Jud 1980; Mark and Goldberg 1986); (3) those that estimate separate functions for each of several types of land use (Adams et al. 1968; Clonts 1970; Dunford et al. 1985; McDonald 1981a, 1981b; Mills and Song 1979; Peiser 1987); and (4) those that estimate functions for one type of land use (Downing 1973; McDonald and Bowman 1979). The first approach is deficient because land value functions are expected to vary by land use. The second approach allows land value functions to vary by use but assumes that the response of land values to variables such as distance from the city center is independent of the zoning category. The third and fourth approaches are, in general, preferable to the first two because they allow land value functions to vary in a general way by land