The non-significance in US dollar terms (column (2)) indicates the sensitivity of bond
prices to exchange rate movements. For individual countries, this is the case for all
countries except for the Netherlands where a coefficient of -4.49 (compared to -1.47 in
domestic currency) indicates that bilateral exchange rate movements with the US dollar
reinforce the negative relation. In this case it implies that higher economic growth for the
Netherlands is accompanied by an exchange rate depreciation vis-a-vis the United States,
thus leading to lower returns in US dollars than in domestic currency. Over a five year
horizon there is only marginally significant evidence (refer to Table 3). When estimated
without country fixed effects we find a positive coefficient (0.55) in terms of domestic
currency with a significance level of 5% (Table 2).
4 Explaining Country Heterogeneity
The analysis revealed substantial heterogeneity in cyclicality patterns across countries.
Consequently it is of interest to find as a second step explanations for the cross-country
variation in the estimations run so far. For this we employ the cross-sectional specification
βi = α + λZi + νi (7)
where ∕i are the set of estimated parameters from the country regressions above. Zi is a
set of control variables. It includes the domestic stock and bond market capitalisation (as
shares of GDP) and output per capita in natural log form (in PPP terms, taken from the
Penn World Tables).10 These control variables are chosen as indicators for the economic
and financial development of the countries included in the sample. Weighted least squares
estimation is used in order to take varying levels of accuracy for the (in the previous step)
obtained dependent variable into account.11
In Table 6 we see the results of this cross-sectional approach in order to find the
determinants of cyclicality in rates of capital gains. In both domestic currency and in US
dollars (columns (1) and (2)) we observe rather similar results for the ∕3is of the real rate
of capital gains on domestic equity markets.
We find GDP per capita to be positively significant (at the 1% level) for the cyclicality
of stock market capital gains. This allows the conclusion that a country’s pro-cyclicality
indicator is increasing with higher economic development. This suggests that higher de-
10We use average values by country for the explanatory variables over the period from 1975 to 2004.
Unreported robustness tests with average values over different time periods confirm the results.
11We weight by the (in the previous step) obtained t-statistics.