5 Empirical Results
The Granger causality test results are reported in Tables 3 through 6. In general, the Granger
causality test results are similar regardless of the property sector. Most importantly, in all the
cases the Granger causality tests show evidence of REIT returns leading TBI returns. That is,
REIT returns lead direct real estate returns even when controlling for the property type and for
leverage. Nevertheless, also the REIT prices seem to contain predictable elements. In
particular, REIT returns seem to be predictable either by changes in the risk premium
(apartments and retail property) or in the economic sentiment (industrial and office sectors).4
The fact that only the forward-looking variable regarding economic growth (i.e., the
sentiment) belongs in the office and industrial models indicates that both of these markets
embody notable forward-looking components concerning economic growth. With the
exception of the apartment sector, the inclusion of fundamentals in the models only slightly
increases the adjusted coefficient of determination for TBI returns.5 This, together with the
fact that in all the sectors REIT returns remain significant in the equation for TBI returns even
after including fundamentals in the models, emphasizes the predictive power of REIT returns
with respect to TBI returns.
[Tables 3-6 here]
The Granger causality tests examine the short-run predictability of and the lead-lag
relations between real estate returns. They do not give details about the reaction patterns and
speeds of the real estate prices to various shocks, however. Therefore, we derive impulse
responses of the real estate returns to shocks in the fundamentals and in the returns
themselves employing the same multiple variable VARs as in the Granger causality tests.