-28-
and 9 plot average returns and the share of equity in the external portfolio over 1997-2001
for a cross-section of countries. In both cases, the figures show a strongly positive relation
between the equity share and the average return—a larger equity share is associated with a
higher return. Second, in terms of data properties, we record in Figures 10 and 11 that
returns are substantially more variable than yields for both foreign assets and liabilities,
providing the example of the United States—plots for other countries are similar.
In term of time-series behavior, Tables 5A and 5B report fixed-effects regressions over
1983-2001 for the specification given in equation (9), for foreign assets and foreign
liabilities respectively.33 Columns (1) and (2) of Table 5A consider the returns on portfolio
equity foreign assets. The MSCI world return index explains the dynamics of these returns
quite well. Adjusting for geographical differences in overseas investment patterns (column
2) does not improve performance. A possible explanation is that the geographical weights
are based on end-1997 positions, and therefore this index may be compromised due to
time-varying portfolio weights; in addition, foreign investors may hold equity baskets in a
given country that differ in composition from its broad market index.
Columns (3)-(4) repeat these exercises for returns on FDI assets at book value, and
columns (5)-(6) for returns on FDI at market value. Again, the explanatory power of the
MSCI index is as good as the geographically-weighted index—as expected, both track
returns on FDI at market value much better than returns at book value. Column (7) shows
that the weighted foreign bond return explains about ¼ of the overall variance in debt asset
returns. Both weighted interest rate measures track the yield on foreign debt assets quite
closely (columns 8 and 9).
33 The countries are Australia, Canada, Finland, Germany, Italy, Japan, the Netherlands,
Portugal, Spain, Sweden, Switzerland, the United States, and the United Kingdom. Canada
was dropped from the equity regressions, since it measures its foreign equity assets only at
book value. Other countries were excluded due to data limitations.