Price-volume Correlation in the Housing Market: Causality and Co-movements (P.12)

Fig. 3 Time series of some MSA and national level variables. This figure reports the 25%, median, and
75% percentiles of the quarterly total non-agricultural employment, average household income, the
unemployment rate across 114 MSAs, and the national average 30 year Fixed Rate Mortgage interest rate
(FRM rate). The sample period is from 1990:2 to 2002:2

the change in population to the change in the house price index over the sample
period. Holding constant the increases in population in a market, the greater is the
increase in house prices, the “tighter” is the market and the lower the long term
supply elasticity. Figure 4 plots the histogram of the long term supply elasticity
across MSAs.
We use AIC to choose the optimal lag order for endogenous variables, which is 3
for all specifications. We use only three quarterly dummies to avoid the multicollinearity
of the four dummies due to the within transformation. The model is
estimated with feasible GLS that allows for heteroskedasticity across MSAs. We
calculate t-statistics using heteroskedasticity-robust standard errors according to
Kezdi (2003). Kezdi (2003) shows that the robust standard deviations allow serial
correlation and heteroskedasticity of any kind, as well as unit roots and unequal
spacing. They also have good small sample properties.

Next Page 13



Leave a Reply

You must be logged in to post a comment.