All the return indices, as expected, appear to be non-stationary in levels and stationary in
differences, and also the fundamental variables seem to be I(1) with the exception of the
inflation rate (see Table A1 in the Appendix). Therefore, the inflation rate is the only variable
that is included in the vector autoregressive models in levels. All the other variables are
differenced in the forthcoming analysis.
4 Econometric Methodology
We use vector autoregressive (VAR) models to study the dynamic interdependences between
securitized and direct real estate markets and the reactions of the real estate returns to shocks
in the fundamentals. For each of the four sectors, two separate VAR models are estimated: a
pairwise model that includes the securitized and direct real estate returns as the only
stochastic variables and a multiple variable model that includes the fundamentals as well.
The estimated models are specified as follows:
ΔXt = μ + Γ1ΔXt-1 + ... + Γk-1ΔXt-k+1 + Tt + εt (2)
where ΔXt is Xt - Xt-1, Xt is a 2+F-dimensional vector (where F is the number of fundamentals
included in the model) of the stochastic variable values in period t, μ is a 2+F-dimensional
vector of drift terms, Γi is a (2+F) x (2+F) matrix of coefficients for the lagged differences of
the stochastic variables at lag i, k is the number of lags included in the model and ε is a vector
of white noise error terms. Finally, Tt incorporates dummy variables that take the value of
one at time t and of zero otherwise. Dummy variables are incorporated in the models to cater
for some extreme movements in the real estate returns and thereby to fulfil the assumption of