4 Empirical analysis
4.1 Data description and aggregation issues
In our analysis, we use quarterly time series from 1984Q1 to 2006Q4 for the
United States (US), the Euro area, Japan, United Kingdom (UK), Korea, Australia,
Switzerland, Sweden, Norway and Denmark, so that our analysis covers 72,2% of the
world GDP in 2006 and presumably a considerably larger share of global financial
markets.4 For the aforementioned countries, we gather real GDP (Y), the GDP de-
flator (P), a short term money market rate (IS), a broad monetary aggregate (M),
and, as asset prices, a house price index (HPI) and the MSCI World price index
(MSW). The monetary aggregate is M2 for the US, M3 for the Euro Area, M2 plus
cash deposits for Japan, M4 for the UK and mostly M3 for the other countries. The
data stem from the IMF, the BIS the ECB and the OECD are collected seasonally
adjusted where available and otherwise applied to the X12-ARIMA procedure.5
In the next step, we aggregate the country series to obtain global series consider-
ing the principles mentioned by Beyer, Doornik and Hendry (2000) and employing
the same method as used by Giese and Tuxen (2007) in the same context. First,
we calculate variable weights for each country by using PPP exchange rates to con-
vert nominal GDP into a single currency.6 The weight of a country i in period t is
therefore:
wi,t =
BIPi,t ePPPi,t
BIPagg,t
Secondly, we take the growth rates of the variable in domestic currency and aggregate
these to global growth rates by using the weights calculated above:
11
gagg,t = ∑Wi,t * gi,t
i=1
Aggregate levels can now be obtained by choosing an initial value (e.g. 100) and
4Own calculations based on IMF data.
5For the delivery of the house price data, we would like to thank Mark Weth and Sebastian
Schich from the Deutsche Bundesbank who collected house price data in their project to ”‘demo-
graphic changes and real house prices”’.
6 1999 is our base year for the PPP exchange rates.
11