-11-
substitute for M3 in the euro zone. In contrast, factors are latent variables which measure
the "underlying process". The rationale is that the behavior of several (national) variables
is driven by a few common (global) forces, the factors. Hence, the latter can provide an
exhaustive summary of the information included in data across countries and regions.
Moreover, and in contrast to simple aggregation, factor analysis allows the
distinction between common forces and idiosyncratic shocks, i.e. the amount in the
measured data which is not considered to be part of the underlying global forces.
Idiosyncratic components mean that the measured variables can include changes which
are exclusively the result of a national data-generating process. Instead, if global
variables are derived by aggregation, the distinction between idiosyncratic and global
shocks is blurred. Any idiosyncratic shock stemming from one (major) country will
inevitably influence the global aggregate and will therefore be counted as a common
shock across countries. Global and idiosyncratic shocks are presumed to have the same
influence, although idiosyncratic shocks should not influence the common movement in
an economic sense.
The use of the FAVAR methodology is especially appealing in the light of
spillovers and global shocks, since it allows both examining the interaction of global
variables and their effects on national variables. For example, it is possible to derive the
impact of a global liquidity shock on global GDP, global commodity, house and share
prices, global inflation, etc. Different types of shocks can therefore be put to some kind of
a "horse-race": Is a global liquidity shock more important for the global economy than a
global commodity price or global interest rate shock? At the same time, the response of
every national variable included in the respective global factor (national CPI, national
money supply, interest rate, share price, etc.) due to a global shock (liquidity, GDP,
commodity, etc.) can be examined.
The problem in expanding the external sector (in a VAR model) is that there is a
rapid increase in the number of parameters that need to be estimated with the addition of
each economy as well as the addition of each sector of the respective economy. The
established approach adopted to circumvent this problem of over-parameterization is to