estate markets. To avoid the potential influence of leverage in the REIT data on the
dynamics, we restate the REIT return index to exclude the impact of leverage. Based on the
estimated vector autoregressive models, the results show that REIT returns lead direct real
estate returns regardless of the property sector. Hence, the property-type mix is not the reason
behind the observed lead-lag relationship between the markets as claimed in the extant
literature. Moreover, our findings suggest that the perceived lead-lag relations are not only
due to the slow adjustment of sellers’ reservation prices, but also due to the sluggish reaction
of demand in the direct real estate market.
The findings have several practical implications. In particular, REIT returns can be
used to predict direct real estate returns. The results also show that REIT returns have
predictable components. Furthermore, the positive lead-lag relations indicate that over the
longer horizon the co-movement between securitized and direct real estate markets is stronger
than the contemporaneous quarterly correlation coefficients suggest. That is, the securitized
real estate performance more closely tracts the direct real estate market performance in the
long run than indicated by the relatively short-term contemporaneous co-movement only.
Therefore, the benefits of including securitized real estate in a long-horizon multiple asset
portfolio containing already direct real estate are more limited than what typically reported
quarterly correlations would suggest.
Although the main findings are similar concerning all four sectors, the results also
show that there are some differences between the sectors regarding the return dynamics.
Therefore, it is reasonable to use sector level data when evaluating the return and price
dynamics in the REIT and direct property markets and when making forecasts concerning
future price development in the markets.
A review of the literature on the lead-lag relations between direct and securitized real
estate markets is presented in the next section. The third section describes the data used in the