19.3. THE INTERNATIONALCAPM 731
sign. Personally I would perhaps even entirely omit the exposure terms: given
the uncertainties surrounding the risk premia and the exposures, one might just
work with the world-market term in the i-capm, and simply widen the scope of
the sensitivity analysis that should be part and parcel of every capital-budgeting
exercise:
E( ̃rj−r)≈βj,w;..E( ̃rw−r). (19.38)
The only surviving difference with the standardcapmwould then be the use of a
world market as benchmark, and the multiple beta.^13
19.3.6 Empirical Tests of the InternationalCAPM
In this chapter, we are suggesting that you replace your familiar single-marketCAPM
equation by a more complicated version, Equation [19.36] or [19.38]. The first issue
is whether one of the basic assumptions of the international model, the absence of
controls on capital flows, is reasonable. Second, are the empirical data compatible
with the InternationalCAPMand, if so, can we also reject the single-country view of
the world?
Let us first examine the effect of direct controls on foreign investment. The
controls may either limit foreign investment into a country or restrict domestic
residents from investing abroad. Restrictions on foreign investment into a country
may be imposed in different ways—in the form of a limit on the fraction of equity that
can be held by foreigners or a restriction on the types of industries in which foreigners
can invest. Historical details on the type and magnitude of these restrictions can
be found in Eun and Janakiramanan (1986, Table 1). There may also be domestic
controls on how much a resident can invest abroad. For example, Japanese insurance
companies could not invest more than 30 percent of their portfolio in foreign assets
at the time, and only 30 percent of Spanish pension funds could be invested abroad.
Two questions need to be answered. One, if these restrictions exist, do they have a
significant impact on the choice of the optimal portfolio and hence, potentially, on
asset pricing? Two, how significant are these constraints today?
Bonser-Neal, Brauer, Neal and Wheatley (1990) examine whether the restrictions
on investing abroad are binding. They look at closed-end country funds and find
that these mutual funds trade at premia relative to their net asset values, indicating
(^13) The need to still use a multivariate regression even in the truncated model follows from the fact
that our basic model is Equation [19.33], not Equation [19.23]. Equation [19.33] simplifies to the
univariate equation, [19.23], only if either the prices of exchange covariance risk,ηk, are all zero,
or the covariances between asset returns and exchange rate changes themselves are zero. The first
case requires very special utility functions (withλ= 1), and the second case cannot possibly be
true for all assets and home currencies simultaneously. Thus, we do need the multivariate model.
Moreover, although the risk premium for exchange risk can be small it is unlikely to be exactly zero.
That is, we use the one-factor world model merely as an approximation. If we would, in addition,
use a univariate beta, we would introduce another (unnecessary) error to the approximation.