Residential land prices prior to development

Journal of Real Estate Research, The, 1997 by Karl L Guntermann

The second measure of development expectations tested in this paper is the number of active subdivisions within one, two or three miles of each parcel. The significant though relatively small correlation (.25-.27) between POPFORE and the three subdivision variables (Exhibit 3) is one reason to prefer a specification with EMPFORE and the subdivision variables, since they are not significantly correlated. The subdivision variables are tested in a quadratic form since subdivision activity can be volatile in response to the local economic cycle.

The coefficient on the SUBD3MI variable in Exhibits 4 and 5 is highly significant, indicating that subdivision activity per se is one basis for forming expectations about future development. The same models testing the SUBD2MI variable produced results essentially similar to those presented in Exhibits 4 and 5. However, the SUBD1MI variable was generally insignificant (Appendix 2) when tested in those models. These results suggest that expectations are revised substantially as development activity changes within a two- to three-mile radius of each parcel and that land values reflect variations in expectations at those distances.

The net in-migration and construction employment variables (Exhibits 4 and 5) are significant and have the expected sign, providing support for the Capozza and Helsley (1989) growth hypothesis. However, the Phoenix population and total wage and salary employment variables, while significant, have negative regression coefficients. The negative coefficients occur in other specifications of the models (in both semi-log and log-linear forms) and for various lagged periods. The growth variables with the smallest correlations (Exhibit 3) are POPPHX, W&SEMP and CONSEMP These variables are combined in models presented in Exhibit 4, equations 5 and 6 and in Exhibit 5, equations 11 and 12. The coefficients for POPPHX and W&SEMP remain negative but their magnitude changes substantially while the coefficient for CONSEMP remains basically the same.

The net in-migration and construction employment variables were tested in the same models for lagged periods up to sixteen quarters with interesting results (Exhibit 6). The coefficients on the net migration variables remained significant for the entire period under either specification but the lagged construction employment variables (CONSLAG8 and LNCONLAG8) became statistically insignificant after eight quarters. The R2 of the models with the net in-migration variables peaked with lags of three and four quarters (log-linear) and then began to decline. The explanatory power of the models with the construction employment variables declined with each lagged quarter from the contemporaneous results in Exhibits 4 and 5.9 Exhibit 6 illustrates these results for selected lag periods.

The significance of these variables over extended lag periods is consistent with the Capozza-Helsley growth hypothesis in the sense that land values presumably reflect the capitalization of long-term expectations about growth and Phoenix (and Arizona) has had above-average growth rates for decades. The three- to four-quarter lag for the migration variable might well reflect the delay in reporting the migration estimates, which are not as directly measurable as other economic series. Construction employment, since it is more closely related to the real estate cycle, may not proxy for growth as well as the migration series, as evidenced by the steady decline in R2 for those models beginning with the first lagged period.