I use several variables to account for the relationship between selling price and
property structural characteristics. First, I expect a negative relation between age (AGE)
and price and a positive coefficient on age squared (AGE2). This expectation reflects the
frequently observed quadratic relation between price and age. A "vintage" effect is
sometimes observed, which is related to high prices for very old properties. I expect the
coefficients on SQFT, LANDSQFT, PARKING, FLOORS and UNITS to be positive.
The variable CONDITION, controls for building condition. I specify average
condition as the control group. With residential real estate, 79 percent of apartments are
reported to be in average condition, 14 percent of the apartment properties are
categorized by CoStar as being in above average condition, and 7 percent are labeled as
below average. In the office sample, 66 percent of the properties are classified to be in
average condition, 32 percent are above average, and only 3 percent are in below average
condition. Finally, with retail properties 70 percent are in average condition, 22 percent
are above average, and 8 percent are in below average condition.
I control for the effects of time by including dummies for each year in the sample
with 1999 as the base year for comparison. In the apartment regressions, I also determine
whether the use of the apartments is primarily as senior housing, subsidized housing, or
multifamily condominiums. The comparison group, which represents 98 percent of the
sample, is all other multifamily apartments.
Finally, I include dummy variables to control for differences across submarkets
within each major market. These submarkets are defined by CoStar. There are 405 unique
submarkets in the apartment sample. In the largest market, Los Angeles, 42 submarkets
are identified by CoStar. In the second largest market, New York, 30 different