Academic journal article Real Estate Economics

Housing Price Volatility Changes and Their Effects

Academic journal article Real Estate Economics

Housing Price Volatility Changes and Their Effects

Article excerpt

Walter Dolde (*)

Dogan Tirtiroglu (**)

We examine significant volatility shifts in regional housing price changes, adapting a method of Haugen, Talmor and Torous (1991) independent of predefined sampling blocks. We identify 36 volatility events, most of which are purely regional, but three of which are national. We find significant associations of volatility events and economic conditions, especially national and regional income growth, inflation, and interest rates. During an initial adjustment period after a volatility shift, realized housing returns move opposite to volatility. We find evidence of significant interregional diffusion of volatility increases, but not of decreases. New insights on links between economic conditions and housing volatility and returns should be of value to household investors and mortgage investors.

Economic theory predicts that asset returns will change in response to changes in the factors affecting the decisions of buyers and sellers. Examples include taxes, demographics, the marginal productivity of capital, and Federal Reserve policy, to cite just a few. Theory predicts that such changes will apply not only to return means, but also to variances and covariances. Empirical studies of stock returns have documented such shifts, both in first and second moments. The nature of the relationship between stock return volatility and mean returns remains controversial and the subject of ongoing empirical research.

Less is known empirically about the existence, extent, and effects of volatility (1) changes in housing returns. Dolde and Tirtiroglu (1997) model and estimate the effects of innovations in housing prices on subsequent return means and variances. Using a GARCH model, they reject the null hypothesis of time-invariant volatility. They also find evidence of a significant positive relationship between conditional variance and returns. These GARCH estimates tell us that volatility is significantly related to lagged information or "news." (2) But they don't tell us whether there are significant volatility events: shifts to a new and much higher or lower level of volatility. That is our interest here.

It is a challenge to identify volatility shifts empirically. In particular, they may go undetected in return blocks with predetermined divisions. Suppose, for example, that volatility in some market tends to be higher during the summer. If researchers examine the volatility of monthly returns in calendar-year blocks, they will detect no deviations from constant volatility. Haugen, Talmor and Torous (1991) provide a method for identifying volatility shifts. Working with dally stock return data, they compute for each day the variance of returns in the 20 preceding days and in the 20 succeeding days. The ratio of the two variances has an F distribution under the null hypothesis of constant variance between the two 20-day blocks. Under the alternative hypothesis, variance may either increase or decrease. Thus the critical range includes both tails of the F distribution. By computing the variances for the blocks preceding and succeeding every trading day, Haugen et al. (1991) can identify all significant variance shifts, wherever they may lie relative to calendar divisions. Dates of significant volatility changes can be compared with the timing of news about important events to determine possible associations. One can also test for the existence of shifts in mean returns following identified volatility shifts.

In this paper, we adapt the nonparametric method of Haugen et al. (1991) to identify and test for significant volatility shifts in single-family housing returns. We perform our tests on monthly housing price indexes described in Chinloy, Cho and Megbolugbe (1997). The data cover January 1975 to December 1993 for four regions of the United States: Midwest, Northeast, South, and West.

We identify 12 statistically significant volatility increases and 24 volatility decreases. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.