San Diego and Riverside, indicates a price appreciation since 1999, all else equal, of 100

percent, 123 percent, and 131 percent, respectively.

 Submarket dummy variables are not reported in the regression outputs, but they

serve for achieving best possible control of relative location in the metropolitan market

and their inclusion improves significantly the fitness of all models.

 Next, I present a second model for the retail properties, OLS Model II, which is

identical in its specification to the second model used with the office properties sample.

The model adds to the model specification, given by equation (18) a third order

expansion of the latitude and longitude coordinates, which has the objective of effectively

controlling for the absolute location of the property. The third order expansion, rather

than a simple linear form, is entered in the equation to draw a price surface based on

location. The model specification is given by equation (23).

 Table 21 presents regression statistics from OLS Model II. Controlling for absolute

location improves the R-squares of market regressions by 0.5 to 1 percent. No significant

changes in the estimated coefficients are observed. With respect to the key variables of

interest results are re-confirmed. The coefficient on the variable representing a

replacement exchange, EXREPL, is positive and significant in 13 of the 15 regressions.

The magnitude of the estimated coefficients remains largely unchanged. This reconfirms

the evidence that buyers of replacement properties are paying significant price premiums

in the majority of the markets. Results with respect to relinquished exchanges, EXRELQ,

and a combination of relinquished and replacement exchanges, RELQ REPL, remain the

same.

 The coefficient on the variable representing a purchase by an out-of-state buyer,