clearly elucidates the moderate inflation which began around the mid-90s and has
persisted in the recent years of global excess liquidity. House prices have shown a
distinct appreciation especially in the last 5 years giving support, to some extent,
to the popular asset price inflation hypothesis in the real estate sector. Global
short-term interest rates were at a historically low level from 2002 to 2005, as the
monetary policy stance was extremely loose during this period.9
4.2 The VAR Methodology
The econometric framework employed is a vectorautoregressive model (VAR) which
allows us to model the impact of monetary shocks to the economy while taking care
of the feedback between the variables since all of them are treated as endogenous.10
Consider first the traditional reduced-form VAR model:
Γ(L)Yt=CDt+ut (1)
where Yt is the vector of the endogenous variables and Γ(L) is a matrix polynomial
in the lag operator L for which Γ(L) = I + Vp ∣ AiLi, so that we have p lags. Dt
is a vector with deterministic terms and the corresponding matrix of coefficients C,
and ut is the vector of the white noise residuals where serial correlation is excluded,
so that:
E(ut) = 0 (2)
l [ Σ : t = s
E(utus) = ʌ (3)
^ 0 : t = s
Since Σ is not a diagonal matrix, contemporaneous correlation is allowed for. In
order to model uncorrelated shocks, a transformation of the system is needed. Using
the Cholesky decomposition Σ = PP', taking the main diagonal of P to define the
diagonal matrix D and premultiplying (1) with A := DP-1 yields the structural
9 One might regard the deviation from a Taylor rate as a more accurate measure in this respect.
However, these numbers create a similar picture. See International Monetary Fund (2007), Chapter
1, Box 1.4.
10 Of course, one could model exogenous variables as well, but this option is not used here.
14