Global Pricing of Equity

Jeff Diermeier, Bruno Solnik *

Version of October 5, 2000

Abstract

Global equity management has historically been structured primarily around country asset allocation.
This approach was supported by the usual observation that the country factor is the major source of
influence on stock price behavior and that the correlation between equity and currency is close to zero
and unstable, so country exposure matches currency exposure. This logic breaks down in a world
where companies work and compete on a global basis and are recognized as such by investors in their
pricing of securities. As corporations expand and diversify their international activities, the relative
importance of domestic factors for such corporations should decline. If a corporation is regarded as a
portfolio of international activities, its stock price should be influenced by international factors in
relation to the geographical breakdown of its activities. Similarly, the currency exposure should be
influenced by the geographical distribution of the firm's activities, rather than by the domicile of
incorporation or by the location where the stock is mainly traded. Under integrated or “global pricing”
the marketplace would reflect the value and changes to value of the foreign activities of the firm. In
essence a French firm with foreign activities could expect investors to value each stream of “national”
earnings at the relevant “national” discount rate adjusted for the firm's specific risk characteristics.
In this paper we present strong evidence in support of global pricing effects. We examine a
large cross section of security prices and we find that asset returns are significantly determined by
regional factors, currency factors, as well as domestic factors. Moreover, the sensitivity of individual
company returns to non-domestic factors is closely related to the extent of their international activities
as proxied by the relative importance of foreign sales to total sales. In our concluding section, we
review the implications of these findings for the asset management profession.

*
Jeff Diermeier is Chief Investment Officer of Brinson Partners Inc., 209 South LaSalle Street, Chicago IL 60604-1295
[diermeierj@brinson.com]. Bruno Solnik is Professor of Finance at HEC, 1 rue de la Libération, 78350 Jouy en Josas,
France [Solnik@hec.fr]. We are grateful to Stefano Cavaglia and Pierre Ruiz for helpful comments and assistance. Bruno
Solnik acknowledge support from the Fondation HEC.
 Global Pricing of Equity

1. Introduction

Investors continually search for better ways to describe the characteristics of their portfolios. For
those investing in global equities it is typical to create metrics to view the country, currency, industry
and other common factor attributes of the portfolio. To date actual management of these portfolios
has been fairly simple and is based upon the following observations.

• Country factors are the major source of influence on stock price behavior, and the correlation of
these country factors is rather weak. For example all French firms see their stock price strongly
influenced by French factors; for instance, Peugeot stock price behaves very much as that of any
other French firm rather than as that of other foreign car manufacturers.

• The correlation between equity and currency is close to zero and unstable, so country exposure
matches currency exposure. For example, the stock price of Peugeot does not systematically goes
up in case of a depreciation in the French franc (Peugeot is very French from a currency exposure
viewpoint and is treated as a franc, or euro, asset).

Both findings are grounded in extensive academic evidence based on data up to the mid-nineties. 1 As
a result, global equity management has historically been structured primarily around country asset
allocation. A two-step procedure is typically employed with the first step being country allocation and
the second selection of industries and stocks within country.

To support the application of this procedure, global portfolio data is typically arranged in a balance
sheet accounting style revealing benchmark, portfolio and active capitalization weights by country and
currency. Each security is presumed to be 100% allocated to the headquarters or trade listing location.
The full allocation of country and currency weight by home market makes sense in a world of highly
“segmented pricing” where all that matters is the location of its headquarters principal trading 2 . A
French Franc firm listed in Paris is valued as a French asset irrespective of the degree of its
international activities.

This logic breaks down in a world where companies work and compete on a global basis and are
recognized as such by investors in their pricing of securities. As corporations expand and diversify
their international activities, the relative importance of domestic factors for such corporations should

1
See references: Lombard, Roulet and Solnik (1999), Beckers, Connor, and Curds (1996), Drummen and Zimmermann
(1992), Griffin and Karolyi (1998), Griffin and Stulz (2000), Heston, and Rouwenhorst (1994), Rouwenhorst (1999).

2
See for example Froot and Dabora (1999).
decline. For example a German firm (like Daimler Benz), which acquires a US firm (e.g. Chrysler)
should see its stock price influenced by US factors in proportion to the value of the US component. If
a corporation is regarded as a portfolio of international activities, its stock price should be influenced
by international factors in relation to the geographical breakdown of its activities. Similarly, the
currency exposure should be influenced by the geographical distribution of the firm's activities, rather
than by the domicile of incorporation or by the location where the stock is mainly traded. This
globalization applies not only to the few well-known multinational corporations 3 , but also, to a lesser
extent, to a large number of firms which have progressively stepped up their global activities by
increased exports, foreign direct investment or M&A 4 .

Under integrated or “global pricing” the marketplace would reflect the value and changes to value of
the foreign activities of the firm. In essence a French firm with foreign activities could expect
investors to value each stream of “national” earnings at the relevant “national” discount rate adjusted
for the firm's specific risk characteristics.

In this paper we present strong evidence in support of global pricing effects. We examine a large
cross section of security prices and we find that asset returns are significantly determined by regional
factors, currency factors, as well as domestic factors. Moreover, the sensitivity of individual company
returns to non-domestic factors is closely related to the extent of their international activities as
proxied by the relative importance of foreign sales to total sales. In our concluding section, we review
the implications of these findings for the asset management profession.

2. Methodology

Figure 1 illustrates alternative methods of classifying the "nationality" of a firm. First, one could use
the firm's primary listing location, its headquarters location, or the primary location of its
shareholders. This very primitive but prominently used classification approach fails to recognize that
share prices reflect information that is publicly and widely available to all international investors.
Second, one could focus on share price movements and statistically estimate the response of
individual securities to domestic market factors and international (or regional) market factors. These
inferences say little about the underlying story behind the response and many not be stable and
persistent. Third, one could examine fundamental accounting and economic data as for instance in
UNCTAD's World Investment Report where the transnationality of a company is estimated by the
proportion of foreign sales, foreign employees and foreign assets it contains. Although intuitively

3
Needless to say, many would claim that a corporation, especially a MNC, is a complex organization, which cannot be
simply equaled to the sum of its national components. An investigation of these alleged organizational benefits is left for
future research.

4
Cross-border M&A have increased at a dramatic speed in the past ten years. IDC reports that cross-border M&A
averaged an annual rate of $40 billion in 1989 to 1993, $160 billion in 1994 to 1998 over $500 billion in 1999 and over $1
trillion in 2000.

2
appealing such fundamental data needs to be related to observed pricing relationships in the
marketplace to be useful.

Figure 1
Measuring Country Loadings of Individual Stocks

Firm Listing,
Headquarters,
Shareholder
Nationality

Fundemental Statistical
Data Estimation

3
Box: Mathematical Derivations

Let's denote Vi, the value of a firm i expressed in its domestic currency (currency i). We decompose
this value as the sum of the present value of distributable earnings derived in each country k. Let's
denote Vi,k the value of the distributable earnings derived in country k (expressed in currency k), it is
the value of a "hypothetical" firm whose distributable earnings are influenced by the country-k factor.
Let's further note Sk the exchange rate of currency k (in units of domestic currency i):
Vi = Σ k Sk Vi,k (1)

Assume a simple one-country-factor model for a purely-domestic firm, where the expected elasticity
(beta) to country factors is equal to one. The return of the country-k component, Ri,k, can be written
as:
Ri,k = α i,k + Ik + ε i,k (2)

Where Ik is the country k index return, measured in currency k.

Assume no correlation between currency and country factor risks, then the currency return, Ck , adds
to the local-currency stock market return and the total return on firm i is given by:
Ri = Σ ϖ i,k Ri,k + Σ ϖ i,k Ck (3)

Where ϖ i,k = Sk Vi,k/Vi is the relative importance of a country k in the total value of a firm.

Hence:
Ri = α ι+ Σ ϖ i,k Ik + Σ ϖ i,k Ck + ει (4)

The exposure to each country factor should be equal to the relative importance of the country in the
economic activity of the firm.

If a firm engages in corporate currency hedging the correlation between stock return and currency
return is not zero so:
R= α ι + Σ ϖ i,k Ik + Σ ψ i,k Ck + ει (5)

Where ψ k < ϖk .

4
This paper uses the last two estimation approaches to prove that global pricing exists, which confirms
the problematic nature of the conventional approach. It involves a basic model that makes use of
statistical estimation and goes on to compare those estimates to fundamental business data. It
demonstrates how fundamental data is rationally priced in the statistical exposures we estimate.

The exact derivation of the model is given in the Box under the assumptions of global financial
integration. We make two simplifying assumptions. First, the value of a purely-domestic firm is
influenced by a single country factor (that could be correlated with other country factors). Second,
and in the absence of additional information, we expect the sensitivity (beta) of a purely-domestic firm
to be one relative to the domestic factor.

To test this "global pricing" story we first compute statistical estimates of factor exposures from stock
market data. Do individual stock returns load up on other country index returns? Second we compare
the factor exposures to fundamental measures of international activity. In other words, we investigate
whether the degree of internationalization of the activities of a firm has the expected influence on the
relative importance of domestic and international market factors in explaining equity returns. To
estimate factor exposures, we conduct for each firm a time-series regressions of the stock returns on
the domestic market factor, the foreign country factors, and the foreign currency factors as shown in
eq (5). Instead of using every single foreign country as factor, we group them in three regional
factors5 : Europe, Asia and North America. We also use the three leading currencies associated with
each region: euro, yen and U.S. dollar. The domestic factor is estimated using a domestic index
cleansed of multinational firms.

R i = α i + βi I dom + ∑ γ i, reg I reg + ∑ δ i, regC reg + ei

reg reg (6)

where the coefficients β i , γi,reg , and δ i,reg are the exposures to the various factors.

The net international exposure is simply measured as the sum of the three regional exposures. The net
currency exposure is simply measured as the sum of the three foreign currency exposures. Note that
the domestic factor is partly correlated with international factors, so by “net international (re currency)
exposure” we mean the international exposure beyond that already reflected in the domestic factor.
The correlation between our domestic factors and foreign currency is found to be rather weak. We do
not report it here, but as expected it is well below the correlation of the usual market capitalization
weighted country indices with foreign currencies.

Next, we perform a cross-sectional regression between the estimated international exposures and the
extent of international activity for all firms listed in a given country. Rather than looking separately at

5
Our approach allows to take into account the fact that some companies focus only on some regional markets (e.g. Europe
for British firms) but does not pose the degrees-of-freedom problem that would be encountered if all country factors had
been used.

5
individual firms, we compare the estimated exposures to the degree of internationalization of the
firm's activities by a cross-sectional regression within each country of domicile.

For each type of exposure (domestic, international and currency) estimate we run the cross-sectional
regression:

Expoi = λ0 + λ1 Fi + ei (7)

where:

Expoi is the exposure of stock i to a specific factor (e.g. domestic, international, currency)

Fi is the degree of foreign activity of firm i proxied by the foreign-sales ratio (foreign/total).

When looking at the domestic factor, we expect to find the intercept λ0 to be equal to 1 and the slope
λ1 to be negative. On average, the domestic beta of a purely-domestic firm should be equal to one. On
the other hand, a multinational firm with little domestic activities should have zero sensitivity to the
domestic factor. For international factors, we expect the intercept λ0 to be equal to 0 and the slope λ1
to be positive. A purely-domestic firm should have no international market exposure (λ0 ), beyond that
already present in the domestic factor. The exposure to international factors should increase with the
extent of international activities. Actually we could even argue that the slope λ1 should be -1 for the
domestic exposure and +1 for the international exposure, under the global portfolio approach. 6 On
the other hand, we should expect to find all slopes λ1 to be zero if the "location of trade" story (market
segmentation) prevails.

3. Data

Stock Prices

For the sample of firms described below we use weekly returns (Friday close) from July 1989 to
January 1999. Dividend adjusted returns are obtained from Datastream. The association of a company
name (as given in the various data sources on foreign activities) and the stock quotation code required
a very careful examination. We excluded firms that had a major corporate event (e.g. major cross-
border acquisition) over the period.

6
The slope could differ from unity for various practical reasons. As is apparent later, the domestic factors are proxied by
the return on a portfolio of firms with mostly-domestic activities. However, some of these firms have a limited amount of
international activities, so the domestic factor proxy is not fully domestic. Another reason is that our indicator of
international activity is an imperfect one. Finally, the two-step econometric procedure use is sensitive to measurement
error in the variables. An alternative econometric procedure would be to use a multivariate estimation for all firms where
each firm's exposure is directly constrained to be of the form λ0 + λ1 Fi . However, this procedure suffers from serious
error-in-measurement problems and results are more difficult to interpret as they do not rely on the risk exposure measures
traditionally used.

6
Statistical Estimates

The extent of global stock pricing is assessed by statistical inference from regressing stock returns of a
firm on a set of domestic and international markets factors.

• Domestic factor: In past literature, the domestic factor is summarized by the return on the
national stock index. A market capitalization weighted national stock index, however, can
largely be determined by multinational firms, which tend to be the largest-capitalized firms in
any country. For example the top six Dutch multinational firms represent more than 60% of the
total Dutch market capitalization. This makes it difficult to separate the influence of truly
domestic factors from that of international factors.

To circumvent this problem, we create a domestic index, which is different from the national
index7 . We construct a portfolio of national firms with mostly-domestic activities. This is an
equally-weighted portfolio of all firms in our sample with a percentage of domestic revenues at
least equal to a threshold. The determination of this threshold is arbitrary and country
dependent so as to create a diversified index. The national threshold selected is given in Table
1. We conducted numerous experiments to check that our results are not sensitive to the
threshold level.

Table 1: Universe of Firms

Country Number of Firms Threshold of Domestic Operations

for inclusion in Domestic Index
France 89 75%
Germany 85 85%
Italy 44 85%
Japan 208 99%
Netherlands 39 65%
Switzerland 86 60%
UK 200 99%
USA 462 99%
Total 1213

7
A similar "portfolio" approach is used in most factor models. Lombard, Roulet and Solnik (1999) followed the same
route.

7
 • International Market factors: The value of a firm should be influenced by the economic
conditions in the countries in which it operates. In this study we use regional stock indices as
international market factors. The regions are Europe, North America and Asia. Our choice is a
compromise between using either one single world factor or every single country factor. The
net international market exposure is equal to the sum of the regional market exposures. All
regional indices are recalculated to exclude the country of nationality of the firm studied (for
example, we use the European index ex UK, for British firms). For US and Japanese firms we
only use two foreign regions, as these countries are the dominant market component of their
region. All regional indices are measured in the region's major currency: the U.S. dollar for
North America, the Japanese yen for Asia and the ECU (the predecessor of the Euro) for
Europe. Hence currency effects will appear in the exposure to currency factors mentioned
below.

• Currency factors: Currencies should affect stock pricing to the extent that international factors
do. But the influence of currency factors could extend beyond that of geographical factors.
Furthermore, many firms adopt an active approach to currency risk management. Hence we
included as international factors the exchange rate of the domestic currency with that of each of
the regions used above. These are the most important currencies in world trade. For example,
the currency factors for British firms are the British pound value of one ECU (the predecessor
of the Euro), one Yen and one U.S. dollar. The net foreign currency exposure is equal to the
sum of the three.

Fundamental Business Data

Our requirements for fundamental business data should have several properties. The data itself should
have an intuitive relationship with the notion of foreign activity. The sample of firms included should
be diverse in terms of foreign activity from purely domestic to heavily multinational firms. An
exhaustive sample of firms is not needed to test for global pricing so quality of data takes precedence
over number of firms.

Data availability and quality severely constrain this study. Efforts to construct a clean database turned
out to be a substantial and time consuming undertaking, even though the raw data was supplied by
well-known vendors. This suggests that for global market analysis to achieve maturity, firms will
need to be required to present foreign fundamental data in a more standardized, rigorous fashion.

We selected data on the percentage of domestic and foreign sales as our measure of foreign activity.
This data is generally taken from annual reports and made available via Datastream, Worldscope and
Factset. Financial institutions such as Nomura and Morgan Stanley also provided some data 8 . This
data varies in quality and consistency in part because of different national accounting standards and
methods. Different sources report markedly different foreign revenue ratios for numerous firms. For

8
Morgan Stanley reports the breakdown of sales and other indicators by region, without specifying the domestic
component.

8
example one source reported Boeing as having only 1% of foreign revenues. This can be due to
different handling of “foreign sales”, ”export sales” and “foreign revenues.” Some financial firms
report insignificant foreign sales at the holding company level without consolidating foreign sales of
subsidiaries. Some European firms report European sales as “domestic”, whereas others report own
country sales as domestic. Many firms simply do not report the geographical breakdown of their
activity in any reliable way.

Our primary sources for European, Japanese and US firms were Datastream, Nomura and Morgan
Stanley respectively. Where available the most comprehensive measure of foreign revenues was used
involving foreign and export sales. The domestic foreign sales ratio for 1997 (available at the end of
1998 and cleaned in 1999) was used over the entire period of the study. This 1997 indicator overstates
the true extent of international activities for some of the firms over the entire period of the study.
Other measures of activity, namely earnings and assets, were less available but were used for some
robustness checks which did not find major differences in conclusions. The resulting database
contained 1213 firms from seven countries (see Table 1). We included countries for which we had a
sufficient number of firms to conduct cross-sectional tests. 9

4. Empirical results

Estimating the Factor Loadings

For all firms in our sample, we estimate the exposures as stated in equation (6) and compute the net
international exposure by summing the exposures in the three regions. We find persuasive evidence
for global pricing.

For example, SmithKline Beecham is a British firm with only 8% of its sales in the UK. Hence its
stock price exposure to the domestic (UK) market should only be 8% while the risk exposures to
foreign market factors should amount to 92%. Similarly, it should be valued as a mostly non-sterling
asset because most of the revenues come from abroad. The foreign currency exposure of the stock
price measured in British sterling should be large and similar to the foreign market exposure unless
currency hedging or similar behavior takes place. In the time-series regression estimates of exposures,
we find a domestic market exposure of 17%. The net exposure to international markets is 94% and
the net foreign currency exposure is 46%. The estimated exposures for SmithKline Beecham are
reported in Table 2. To be sure that the meaning of these exposures is fully understood, let’s take the
example of the exposure, or sensitivity, to the American stock market, that is equal to 0.55. 10 On the
average, the stock price of SmithKline Beecham (measured in sterling) goes up by 0.55% when the
U.S. stock market goes up by 1%.

9
We could not get information for a large number of Swiss firms, but we report in Tables 3, 4 and 5 comparable results
found by Lombard, Roulet and Solnik (1999).

10
Actually, SmithKline Beecham report that American sales amount to about 50% of its total sales.

9
Table 2: Exposures for SmithKline Beecham
As estimated in equation (6); Standard errors in parentheses

Domestic (β ) Foreign market (γs) Foreign Currency (δ s)

Asia EuropeX America Total Asia Europe America Total

0.17 0.08 0.31 0.55 0.94 0.08 0.11 0.27 0.46

2
(0.09) (0.06) (0.10) (0.09) (0.09) (0.16) (0.11) R =0.22

The lower foreign currency exposure suggests that SmithKline Beecham engages in various forms of
corporate currency hedging. The foreign currency exposure of 0.46 means that SmithKline Beecham
is not fully a sterling asset. When sterling drops against all currencies, the stock price of SmithKline
Beecham expressed in sterling tends to go up by 0.46 times the depreciation. Thus an American
investor buying this stock is only partly exposed to sterling. Conversely, a British investor owning the
stock is strongly exposed to foreign currencies.

International Factor Exposure and Foreign Sales

Our theoretical model suggests that factor exposures ought to be related to the extent of firms'
international activities. Our proxy for firms' international activities is the ratio of foreign sales to total
sales. Thus if one was to examine a portfolio of securities, one should find that firms with a low
exposure to their domestic factor should have a high proportion of foreign sales. Similarly, firms with
a high exposure to foreign (market and currency) factors should have a high proportion of foreign
sales. The currency hedging activities of firms would be reflected in differing foreign market and
currency exposures. These hypotheses were tested via cross sectional regressions where the portfolio
of securities we consider are the constituents of each individual country we examine.

Results for the cross sectional regressions of domestic exposures on foreign sales ratios are presented
in Table 3. Results for the regressions of net international market exposure (sums of regional
exposures, Σγ) on foreign sales ratios are presented in Table 4. Results for the regressions of net
foreign currency exposure (sum of regional currency exposures, Σδ) on foreign sales ratios are
presented in Table 5.

10
Table 3: Domestic Exposure
For each country, the first line gives the estimated coefficients from equation (7):
Expoi = λ0 + λ1 Fi + ei
Where the dependant variable is the domestic exposure and the explanatory variable
is the foreign sales ratio. Standard errors appear in parentheses.

Country Intercept (λ0 ) Slope (λ1 ) R2 Number

of Firms

U.K. 0.96 -0.28 4% 200

(0.04) (0.07)

Germany 1.03 -0.32 4% 85

(0.07) (0.15)

France 0.79 -0.53 17% 89

(0.07) (0.12)

Netherlands 1.07 -0.79 41% 39

(0.10) (0.15)

Italy 1.01 -0.32 13% 44

(0.06) (0.12)

Switzerland 0.97 -0.51 29% 86

(0.07) (0.09)

Japan 1.06 -0.56 10% 208

(0.03) (0.12)

U.S.A. 0.93 0.26 3% 462

(0.02) (0.09)

11
Table 4: International Market Exposure
For each country, the first line gives the estimated coefficients from equation (7):
Expoi = λ0 + λ1 Fi + ei
Where the dependant variable is the international market exposure and the
explanatory variable is the foreign sales ratio. Standard errors appear in parentheses.

Country Intercept (λ0 ) Slope (λ1 ) R2 Number

of Firms

U.K. 0.01 0.35 22% 200

(0.02) (0.05)

Germany -0.04 0.73 31% 85

(0.06) (0.12)

France 0.13 0.61 13% 89

(0.10) (0.16)

Netherlands -0.12 0.79 49% 39

(0.09) (0.13)

Italy -0.03 0.49 40% 44

(0.04) (0.09)

Switzerland -0.08 0.55 32% 86

(0.07) (0.09)

Japan -0.03 0.59 22% 208

(0.02) (0.08)

U.S.A. 0.03 0.13 2% 462

(0.01) (0.04)

12
Table 5: Foreign Currency Exposure
For each country, the first line gives the estimated coefficients from equation (7):
Expoi = λ0 + λ1 Fi + ei
Where the dependant variable is the foreign currency exposure and the explanatory
variable is the foreign sales ratio. Standard errors appear in parentheses.

Country Intercept (λ0 ) Slope (λ0 ) R2 Number

of Firms

U.K. -0.05 0.29 17% 200

(0.02) (0.04)

Germany -0.06 0.38 9% 85

(0.06) (0.12)

France 0.08 0.40 6% 89

(0.10) (0.17)

Netherlands -0.13 0.46 19% 39

(0.10) (0.15)

Italy 0.02 0.01 0% 44

(0.07) (0.13)

Switzerland 0.02 0.08 0% 86

(0.06) (0.08)

Japan -0.03 0.56 24% 208

(0.02) (0.07)

U.S.A. 0.00 0.16 3% 462

(0.01) (0.05)

13
To illustrate the results, let's take the portfolio of Dutch firms. In the regression between domestic
exposure and the degree of international activity, the intercept is 1.07 (with a standard error of 0.10)
and the slope -0.79 (with a standard error of 0.15). In other words, a purely-domestic firm is expected
to have a domestic beta of 1.07 (not significantly different from 1 at the 1% confidence level). As the
degree of internationalization of Dutch firms increase, their domestic exposure decreases with a slope
of -0.79 (not significantly different from -1 at the 1% confidence level). For example, a firm which
derives 50% of its revenue outside of the Netherlands is expected to have a domestic exposure of only
0.67. Conversely, the international exposure is close to 0 for purely-domestic Dutch firms (-0.12 and
not significantly different from zero at the 1% level), and increases with the degree of international
activity (slope +0.79, and not significantly different from 1 at the 1% confidence level). A similar
conclusion is reached for currency exposure which increases with the degree of international activity
(slope of 0.46). Note however that the currency slope of 0.46 is well below the international market
slope of 0.79, suggesting that Dutch firms engage in various forms of currency hedging. Buying
stocks of Dutch multinationals cannot be regarded as a pure euro investment. A visual presentation of
the results for international market exposure is given in Figure 2.

Figure 2: Regression between International Market Exposure and the degree of

International Activity.

1.2
International Market Exposure

1.0
0.8
0.6
0.4
0.2
0.0
-0.2 0% 20% 40% 60% 80% 100%
-0.4
Foreign Revenue Ratio

14
General conclusions can be drawn from the Tables. We will focus on Tables 4 and 5 which reflect the
extent of international pricing. As can be seen in Table 4, the slopes are positive, and statistically
significantly different from zero, for all countries indicating that foreign sales activity indeed
translates into international stock pricing. The adjusted R-squares are quite large except for the U.S.
where the relation is weak. The evidence for European and Japanese firms is quite strong. Even
though our indicator of international activity is rather crude and static (1997 foreign sales ratio) it is
effective in discriminating across firms. 11 In the USA, multinationals are only slightly more exposed
to international market factors than domestic firms. This puzzling result confirms the findings of
Lombard, Roulet and Solnik (1999). One explanation advanced is that the US economy is a large and
open economy, where firms from all over the world compete, making it more international than the
other countries. Still, a US firm with extensive operation in a given region should be more affected by
a sudden recession in that region than the typical domestic US firm.

Looking at Table 5, we see that the slopes for foreign currency exposure are always smaller than for
international market exposure (except the US, where they are almost identical). This result suggests
that firms engage in various forms of corporate currency hedging. This is most pronounced in
Switzerland and Italy. Indeed, Swiss firms are seldom listed abroad and are known for attempting to
smooth earnings reported in Swiss francs. On the other hand, Japanese firms are most sensitive to
foreign currency factors, indicating that the stock price of Japanese export firms react positively, and
strongly, to a drop in the value of the yen. Similar evidence is found in a recent working paper by
Bodnar, Dumas and Marston (2000).

We conducted several robustness checks that we do not report here for sake of brevity. We checked
the sensitivity of the results to the construction of the domestic indexes by varying the thresholds for
inclusion, excluding the firms that make up the domestic indexes from the analysis, and using other
proxies for the domestic indexes (including indexes constructed from "local" firms as defined by the
Financial Times). While the coefficients changed slightly, conclusions remained unaffected. When
available we tried other proxies for the degree of international activities (profit, assets 12 ); the sample
of firms was much smaller and biased towards multinational firms but conclusions remained the same.
A joint multivariate estimation (as described in footnote 6) pointed in the same direction.

Finally we split the period in two 5-year sub-periods. We did not find an increase in the extent of
global pricing (the slopes λ1 ). This result would suggest that the world financial market was already
integrated in the early nineties (at least for developed markets). What is changing is that corporations
are becoming more global in their business activities through increased exports and cross-border
mergers and acquisitions.

11
There are many econometric reasons for expecting a downward bias in the estimated slope. Our domestic factor is not
purely domestic for many countries, as it includes firms with some international operations. Hence, the slope should not be
unitary. Second our indicator of international activity is not a perfect one; measurement error introduce a downward bias.
Our indicator applies to international activity at the end of the period; on the average firms were less international at the
start of the period. So we should not be surprised to fin a slope lesser than one.

12
Actually the three indicators are strongly correlated (correlation above 0.9), as already found in Rugman (1976).

15
4. Conclusions and Implications for Asset Management

The main conclusions of this research can be summarized as follows:

• There is strong evidence that firms are priced globally with the market taking into consideration
the portfolio of international and domestic value contained within a company's aggregate value.

• It would be incorrect to assume that the location of a firm's headquarters or its trading post
captures the major determinant of its stock price behavior.

• A relationship exists between the degree of domestic/international firms stock exposure as inferred
from return data and its domestic/international sales. The greater the proportion of international
sales the greater the likelihood that the stock responds to foreign stock price movement. The
results are less pronounced for US firms.

• Foreign stock market exposures exceed foreign currency exposures suggesting that some
smoothing or hedging activity takes place. However, international firms see their stock price
strongly exposed to foreign currencies. As a result, fully hedging back to home currency the
accounting currency exposure typically estimated on the part of investors should result in
systematic over-hedging.

• Improved estimation will require more standardized and detailed reporting requirements on the
part of the corporate community. The difficulties we faced in obtaining reliable data for this study
shows that there is a long way to go.

These findings suggest that conventional methods of describing markets and currency allocations are
problematic and biased. It is clear that the market looks to the underlying fundamentals of where
business activity takes place when pricing assets.

Thus heavily home biased portfolios, isolated to some fragment of a global capital market, appear
more and more to be a randomly undiversified portfolio approach with no incremental return
expectation to offset the idiosyncratic risk. For example, a Swiss institutional investor focusing on
Swiss stocks would end up with some global exposure because some Swiss securities are not truly
“domestic”. In Table 6 we report the capitalization weighted average exposures of all constituents
(both the "local" companies and the "multinational" companies) of this market. The loading on the
domestic Swiss index is only 0.45 suggesting a far from unitary response to the Swiss factor. The
aggregate loading of 0.60 on foreign markets reflects the transnationality of companies (like Nestle,
Novartis, Roche, and UBS) that happen to be domiciled in Switzerland. It should thus be clear that
our bottom up analysis critically stands in contrast with traditional accounting based portfolio
measures of market exposure. Furthermore, our analysis suggests that a fund manager who swapped
Alusuisse (with a domestic exposure of 1.04) for Roche (with a domestic exposure of 0.04) would
significantly alter the Swiss exposure of his portfolio; traditional measures would show the Swiss
exposure to be unchanged. While a diversified Swiss portfolio afford some global exposure, it carries

16
a lot of idiosyncratic risk. It is highly weighted towards five or six names and several major industries
(especially in the MTT sector) are simply absent. Furthermore, there is little reason why a Swiss firm
would necessarily be the best worldwide investment opportunity in each of the present industries.

Table 6: Exposures for the Swiss Stock Market

Domestic (β ) Foreign market (γs) Foreign Currency (δ s)

Asia Europe America Total Asia Europe America Total

0.45 0.07 0.38 0.15 0.60 0.10 0.10 -0.07 0.13

Conventional asset allocation methods may lead to currency over hedging. Applying fixed hedge
ratios to all stocks of a foreign country or region is not appropriate.

The simple paradigm of country allocation and the allocation within country breaks down when the
companies themselves are global. Analysis of the individual firm and its diversity becomes critical.
Cavaglia, Brightman, and Aked (2000) find that global industry factors have been rising in importance
relative to country factors; thus, they suggest that industry factors may capture the principal diversity
amongst companies. Future research will need to examine the interplay of traditional and nationally
based style factors - value, growth, size.

Finally, we note that it has also become fashionable to deal with the complexity of global asset
management by retaining the country asset allocation approach while introducing a new asset class
"multinationals". Our results suggest that this is an inadequate shortcut. The world is not split in two
types of firms: local and multinational. Rather it is a matter of degree with a seemingly intuitive
progressive relationship. In other words, the more international a firm is, the more it is exposed to
international factors. 13 We do not find that big multinationals constitute a separate asset class.

13
Additional empirical evidence in support of this conclusion can be found in Cavaglia and Aked (1999). Their study uses
foreign sales and earnings data to corroborate evidence of global pricing for the 200 largest multinationals.

17
 References

Beckers, S., Connor, G. and R. Curds, "National versus Global Influences on Equity Returns",
Financial Analysts Journal, March/April 1996.

Bodnar, G. M., Marston R., and B. Dumas, "Pass-Through and Exposure", Working Paper,
Wharton School, 2000.

Cavaglia, S., Brightman, C. and M Aked, "On the Increasing Importance of Industry Factors",
Financial Analysts Journal, forthcoming 2000.

Cavaglia, S. and M Aked, "The Global 200: A Separate Asset Class?”, Unpublished manuscript,
March 1999.

Drummen, M. and H. Zimmermann, "The Structure of European Stock Returns", Financial

Analysts Journal, July-August 1992.

Froot, K. A. and E. Dabora, "How Are Stock Prices Affected by the Location of Trade?", Journal
of Financial Economics, 53, 1999, pp. 182-216.

Griffin, J. M. and G. A. Karolyi, "Another Look at the Role of the Industrial Structure of Markets
for International Diversification Strategies", Journal of Financial Economics, vol 50, N° 2, 351-
374, December 1998.

Griffin, J. M. and R. Stulz, "International Competition and Exchange Rate Shocks: A Cross-
Country Industry Analysis of Stock returns", Review of Financial Studies, Forthcoming 2000.

Heston, S. L., and K. G. Rouwenhorst, "Does Industrial Structure Explain the Benefits of
International Diversification?", Journal of Financial Economics, August 1994.

Jorion, P., "The Exchange-Rate Exposure of U.S. Multinationals", Journal of Business, 1990, Vol.
63, 3.

Lombard, T., J. Roulet and B. Solnik, “Pricing of Domestic versus Multinational Companies”,
Financial Analysts Journal, Vol 55, N° 2, 35-50, March/April 1999.

Napolinato, G., and A. J. Cohen "S&P 500 Foreign Currency Exposure", Goldman Sachs, U.S.
Research (Accounting/Portfolio Strategy), September 8, 1997.

Rouwenhorst; K. G. “European Equity markets and the EMU”, Financial Analysts Journal, vol 55,
N° 3, 57-64, May/June 1999.

Rugman, A. letter to the editor, Financial Analysts Journal, March/April 1976.

18