-10-
involves a bet that UIP does not hold over the investment period. Moreover, currency
substitution may enable international liquidity spillovers in a framework of flexible
exchange rates. These international adjustments of money holdings allow the
transmission of monetary shocks from one economy to another (via money demand) even
in a system of flexible exchange rates.
The time series for the G7 plus the euro zone are drawn from a variety of sources,
including various national statistical offices, central banks and the OECD. Most time
series are provided by professional databanks like Thomson Datastream, Feri and
Bloomberg. Since EMU series at a quarterly frequency are often available only for a
relatively short time-span, we partly rely on the Area-Wide Model database by Fagan,
Henry and Mestre (2001) who provide backdated time series. Hence, some caution is
warranted, as there are methodological differences across euro-zone countries in
collecting the data. Moreover, data availability on a national level becomes increasingly
scarce when moving back in time. Historical house prices for the EMU stem from Gros
(2007).
We estimate our baseline FAVAR model including the following variables for the
G7-countries plus the non G7-member euro-area countries: real GDP, four inflation
measures (CPI, PPI, import prices and GDP deflator), 3-month short-term interest rates,
broad monetary aggregates (typically, M2 or M3), two commodity price indices (HWWI
and CRB), house prices and share prices (national MSCI). A complete data list with
sources can be found in Table A7 in the appendix.
4. The FAVAR methodology
The empirical value added from our approach stems from the use of a Factor Augmented
VAR (FAVAR). A FAVAR is the combination of a standard VAR model with factor
analysis. Global variables were derived by using factor analysis. We regard this
procedure as superior compared to a simple aggregation of national variables for two
reasons. Simple sum aggregation implicitly assumes that the included national variables
are perfect substitutes. However, given differences in national measurement, this does not
need to be true. For example, the monetary aggregate M2 in the US is not a perfect