Price-volume Correlation in the Housing Market: Causality and Co-movements (P.4)

variables affect both prices and volume in housing markets. We test Granger
causality from prices to trading volume (Stein 1995 theory) and from trading volume
to prices (Wheaton 1990 theory), to compare and contrast the Wheaton (1990) and
Stein (1995) theories. In this step, we also estimate three alternative specifications of
the VAR model. Specifically, we separate positive changes in house prices from
negative changes to test an asymmetric relation between house prices and volume
implied by Stein (1995), which is due to that equity constraints are binding in falling
markets. Further, housing markets are well known for being heterogeneous,
particularly in terms of supply elasticity. Therefore, we break down our sample
into two groups of MSAs, with above and below median supply elasticity
respectively, and estimate and Granger causality tests for each group.
Second, we empirically analyze determinants of prices and trading volume in
housing markets and investigate the existence and magnitude of the co-movements
between prices and trading volume, as well as to what extent they help explain the
price–volume correlation. For each specification of the VAR model, we decompose
the changes in prices and volume respectively into two components: the fitted part
(explained by our VAR model) and the residual, and investigate if our model
captures the price volume correlation in the data. Further, we decompose the fitted
values into three components: a price-caused component (explained by lagged
prices), a trading volume-caused component (explained by lagged trading volume),
and a co-movement component (explained by exogenous changes in the economy),
and study how each component helps explain the fitted price–volume correlation.
Finally, we use impulse response functions to describe the responses of prices and
trading volume to shocks.
This paper provides original insights into the determinants of prices and trading
volume in the housing market. We find that both house prices and trading volume
are significantly affected by changes in the labor market, which include changes in
total non-agricultural employment, average household income, and the unemployment
rate. The housing market is also significantly affected by the level and trend of
mortgage rates (we use the national average interest rate for 30year fixed rate
mortgages). When the mortgage rate is high and when it is falling, both home prices
and trading volume are low. Interestingly, the stock market performance also has a
statistically significant effect on house prices. When the S&P500 index is high
(level) or when it shows a down turn (trend), home prices tend to be low and trading
volume tend to be high.
We find strong evidence that home prices Granger cause trading volume.
Moreover, it is the decreases in prices, not the increases, that affect future trading
volume, which is direct evidence of support for Stein (1995). We also find some
evidence that trading volume Granger causes home prices, which, derives mainly
from markets with low supply elasticity. This appears to indicate that in markets
where supply can easily adjust, trading volume does not seem to affect future prices.
The fact that trading volume more significantly affects prices in supply constrained
markets seems consistent with Wheaton (1990). Overall, we find supporting
evidence for both Stein (1995) and Wheaton (1990).
We find a statistically significant positive price–volume correlation in the housing
market, and this correlation seems to be explained by co-movements of house prices
and trading volume instead of the causal relations between prices and volume.

Next Page 5



Leave a Reply

You must be logged in to post a comment.