Price-volume Correlation in the Housing Market: Causality and Co-movements (P.21)

To investigate the extent to which the price–volume correlation is explained by
the Granger causality between prices and trading volume and by the exogenous
shocks, we further decompose the “fitted” parts of both hpi,t and toi,t into three
components—the component explained by lagged prices, the component explained
by lagged trading volume, and the component explained by all other variables. We
then calculate the correlation between the same component of house prices and
trading volume, and thus have three correlations: the correlation between pricecaused
prices and trading volume (“Price-caused” in the table), the correlation
between trading volume-caused prices and trading volume (“Turnover-caused” in the
table), and the correlation between prices and trading volume caused by other
variables (“Co-movements” in the table).
Panel B of Table 6 reports the three types of price–volume correlations, and the
corresponding t-statistics of two-sided tests that the correlations follow distributions
with zero means. We have a few interesting findings. First, we find that the “comovement”
component of the price–volume correlation is statistically significant at
the 1% level under all specifications and for all subsamples. Moreover, the
correlation ranges from 0.529 to 0.589 for different specifications or subsamples,
which is much higher than the raw correlation and the “fitted” correlation. Second,
the “price-caused” component is statistically significant at the1% level but is
negative. The negative “price-caused” component of the price-volume correlation
seems to be caused by the positive effect of prices on future turnover and the
negative autocorrelation of prices at quarterly frequency. Third, the “turnovercaused”
component is significantly positive on average across MSAs, but varies

Next Page 22



Leave a Reply

You must be logged in to post a comment.