to 2006.
We employ the Datastream domestic and global equity price indices in order to calcu-
late annual rates of capital gains. These are available both in terms of domestic currency
and US dollars and have the advantage of including only pure equity prices (thus without
dividend payments). Hence these indices are appropriate in order to analyse the capital
gains channel of international investments. For bond markets, data are provided by the
Datastream benchmark indices on two-year and ten-year government bonds. These in-
dices are available both in domestic currency as well as in US dollars. Furthermore we
employ data provided by Datastream on domestic and global stock capitalisation, as well
as data on bond market capitalisation provided by the Bank of International Settlements
(BIS).
GDP (at constant prices) and CPI data for individual countries and the world economy
are retrieved from the IMF’s WEO database. Conventionally GDP growth rates measure
the average growth rate in a given year; however, this is not appropriate for our analysis.
As we are dealing with stock market rates of capital gains - which are essential year-
end to year-end rates - one has to apply the same logic to real rates of GDP growth.
Consequently we construct a year-end to year-end rate of GDP growth by considering
real GDP in the last quarter of a given year relative to the last quarter of the year before.
Thus we obtain a real GDP growth rate which is consistent with the other variables in our
analysis. In the same way we construct appropriate inflation rates in order to calculate
real rates of capital gains. Output per capita data are taken from the Penn World Tables.
Given the data availability and the empirical focus on cyclical factors, the data used
are at annual frequency.
3 The Cyclical Properties of Domestic Capital Mar-
kets
3.1 Equity Markets
As outlined above the analysis starts by considering the co-movement of domestic output
innovations (that is GDP growth rates) and the performance of domestic stock markets
measured by rates of capital gains. We consider regression specifications with both all
variables expressed in domestic currency (thus taking the perspective of a domestic in-
vestor in one of the sample’s countries) and all variables expressed in terms of US dollars
in order to have a common currency among all countries. This can be understood as