J. Int. Financ. Markets Inst.

Money 83 (2023) 101729

Contents lists available at ScienceDirect

Journal of International Financial Markets,

Institutions & Money
journal homepage: [Link]/locate/intfin

What explains the benefits of international

portfolio diversification?
Najah Attig a, Omrane Guedhami b, *, Gregory Nazaire c, Oumar Sy d
a
Dalhousie University, 6100 University Av, PO BOX 15000, Halifax, NS B3H 4R2, Canada
b
University of South Carolina, 1014 Greene Street, Columbia, SC 29208, USA
c
Jointly Appointed Lecturer, Cox Institute of Agricultural Technology, Dalhousie University, 50 Pictou Rd, Truro, NS B2N 2R8, Canada
d
Dalhousie University, 6100 University Av, PO BOX 15000, Halifax, NS B3H 4R2, Canada

A R T I C L E I N F O A B S T R A C T

JEL Classification: This study uses a large sample of international stocks to examine the sources of the benefits of
G11 international portfolio diversification. It finds not only that international diversification out­
G12 performed industrial diversification over the past 27 years, but also that the gains from inter­
G15
national diversification derive primarily from mitigating market, political, and inflation risks.
F30
Economic risk appears important for investors giving more weight to smaller countries, while
F37
F65 financial risk appears more important to funds limited to large countries. Risks related to the
quality of the legal and credit environments seem less important for international diversification.
Keywords:
International portfolio diversification
Industrial diversification
Market risk
Political risk
Economic risk
Developed markets
Emerging markets

1. Introduction

International portfolio diversification has been accepted as an essential risk reduction tool since the early work of Markowitz
(1952) and continues to receive growing support from academics and practitioners alike. This is not surprising, given that the different
institutional arrangements and diverse political, cultural, financial, and economic environments in which international capital markets
operate result in low country intercorrelations and offer investors real benefits from diversifying their portfolios internationally.
What is somewhat surprising is the paucity of research devoted to explaining why international portfolio diversification is bene­
ficial. Heston and Rouwenhorst (1994) consider this lack of understanding of the sources of international portfolio diversification
benefits as one of the still open questions in the literature, noting: “Although we identify large country-specific effects, the remaining
puzzle is to determine the sources behind these effects” (p.5). They further conclude that their “findings do not identify the origin of
these strong independent movements. Different explanations may attribute them to various forms of economic policy; we leave their

* Corresponding author.
E-mail addresses: [Link]@[Link] (N. Attig), [Link]@[Link] (O. Guedhami), [Link]@[Link] (G. Nazaire), oumar.
sy@[Link] (O. Sy).

[Link]
Received 18 February 2022; Accepted 28 December 2022
Available online 31 December 2022
1042-4431/© 2023 Elsevier B.V. All rights reserved.
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

consideration for future research” (p.26). Fresh empirical evidence along the lines of Heston and Rouwenhorst’s insights seems
warranted, especially because two competing predictions inform our understanding of the relevance of international diversification. It
is possible that, under the assumption that global factors became more important, country indexes will be more correlated and the
benefits of international diversification will be lower. Alternatively, globalization reversals will lead to further segmentation from the
global market, and the benefits of international diversification are likely to be larger. This prediction builds on the view that trends
towards globalization tend to retrench following major financial crises (Milesi-Ferretti and Tille, 2011; Doidge, Karolyi, and Stulz,
2020; Gormsen and Koijen, 2020).
Understanding which of the two predictions dominated over the past 27 years is an important question that can only be answered
empirically. We add to the literature by examining whether international diversification is still relevant despite the long-term trends
towards globalization and integration of international markets. Understanding this topic will provide us with a more complete picture
of the relevance of international diversification benefits during market crashes in the most recent quarter-century, a period that has
witnessed major financial crises such as the 2007–2009 global financial crisis and the COVID-19 pandemic.
The point of departure for our second contribution is the lack of convincing evidence on the local risks behind the benefits of
international portfolio diversification. Cosset and Suret (1995) provide one of the first pieces of evidence on the benefits of diversifying
a particular local risk. They examine political risk and find that significant diversification gains can be achieved by broadening the
investment opportunity set to include politically risky countries. Further, De Santis, Gerard, and Hillion (1999) show that investors are
rewarded for their exposure to currency risk. Our study adds to the literature by considering six other local risks that may determine
the benefits of international diversification: market risk, economic risk, financial risk, inflation risk, legal risk, and credit risk. This
investigation is crucial because a better understanding of the major local sources of risk inherent in international investing allows the
design and implementation of more effective risk management strategies.
To address the two questions, we rely on the two-step approach of Campa and Fernandes (2006).1 In the first step, we estimate the
pure country effects using the dummy variables framework of Heston and Rouwenhorst (1994). In the second step, we regress the
absolute pure country effects on the seven measures of risk. In this latter stage, we differ from Campa and Fernandes on two important
points. First, we focus on risk measures instead of economic and financial integration and development measures. While the market
integration and development variables can significantly influence the benefits of international portfolio diversification, they can in no
way be viewed as the risks that are diversified when investors geographically spread their investments across several markets. While
the characteristic variables considered by Campa and Fernandes can be correlated with the local risk factors driving international stock
returns, we believe that directly identifying these risks is more useful for implementing top-down international diversification stra­
tegies. Second, to reduce the likelihood that spurious regression bias determines the regression outcome, we use monthly absolute
deviations of pure country effects as regressors instead of yearly absolute deviations. This is important because we find that for all 48
countries considered in this study, their yearly series of absolute pure country effects contain a unit root, which makes them unreliable
in standard regressions. Further, as argued by Rossana and Seater (1995), the temporal aggregation of economic time-series data can
lead to substantial information losses.
We draw on a large sample of nearly 42,000 stocks from 25 developed markets and 23 emerging markets. We find that for January
1995 to December 2021, international diversification provided a much more effective risk-reduction tool than industrial diversifi­
cation. The results also show that the benefits of international diversification primarily stem from the mitigation of market, political,
and inflation risks. Economic risk appears significant for global investors giving more weight to small countries, while financial risk
seems more important to investors giving more weight to large and mature countries. The risks linked to the quality of the legal and
credit environments appear less important for international portfolio diversification.
The paper proceeds as follows. In Section 2, we review the related literature and identify research gaps. In Section 3, we describe
the data and provide summary statistics. Section 4 presents the methods we use to (i) measure the benefits of international diversi­
fication and (ii) determine the sources of international diversification gains. Section 5 discusses the benefits of international and in­
dustrial diversifications, while Section 6 presents the results on the sources of international diversification benefits. Section 7
concludes.

2. Related literature and research gaps

Pioneered by the early studies of Grubel (1968), Levy and Sarnat (1970), and Solnik (1974), a sizable literature has since sought to
identify the benefits of international diversification.2 Heston and Rouwenhorst (1994) and Griffin and Karolyi (1998) recommend
diversifying internationally based on the evidence of low correlations between national stock markets.3 However, studies such as Lin

1
The analysis of Cosset and Suret relies on mean–variance optimization procedures based on 36 country indexes. While such quadratic opti­
mization techniques are appropriate for a few indexes, the method is not suitable for a large cross-section of stocks. Bae, Elkamhi, and Simutin
(2019, p.2579) also argue that “relying on equity indices to assess diversification benefits understates diversification gains.”.
2
On the corporate level, Gande, Schenzler, and Senbet (2009) find that global diversification enhances firm value, while Lang and Stulz (1994)
find that firm industrial diversification destroys value. Lewis (2000) explains why estimates of international diversification gains differ depending
on whether researchers use stock market data or consumption data.
3
Gilmore and McManus’ (2002) results suggest that US investors can obtain benefits from international diversification into three Central Eu­
ropean markets. Bekaert and Urias (1996) find significant diversification benefits for the UK country funds, but not for the US fund and attribute this
variance to differences in portfolio holdings in the United States versus the United Kingdom.

2
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

et al. (1994), Longin and Solnik (1995), Freimann (1998), Goetzmann et al. (2005), and Hong et al. (2007) point to a marked increase
in cross-country correlations, a growth often attributed to the globalization and integration of international markets. Further, Longin
and Solnik (2001) and others find that country intercorrelations are asymmetric, as they tend to increase dramatically during difficult
times, when investors urgently need to see the benefits of diversification materialize.4 This asymmetry cast doubt on the real benefits of
international portfolio diversification in hard times (e.g., Solnik et al., 1996; Butler and Joaquin, 2002; Chua et al., 2009; Ilmanen and
Kizer, 2012; Page and Panariello, 2018).5
Another thrust of work shows that industrial diversification had become at least as important as international diversification to­
wards the turn of the millennium (e.g., Baca, Garbe, and Weiss, 2000; Cavaglia, Brightman, and Aked, 2000; L’Her, Sy, and Tnani,
2002; Phylaktis and Xia, 2006). These results sparked a debate over whether the surge in industrial effects in the late 1990 s was
permanent or temporary. Brooks and Del Negro (2004) argue that this surge is primarily due to a transient shock in the technology,
media, and telecommunications (TMT) industries at the heart of the IT bubble at the turn of the millennium, but succeeding literature
finds only mixed results (e.g., Ferreira and Gama, 2005; Baele and Inghelbrecht, 2009; Bekaert, Hodrick, and Zhang, 2009).
A related burgeoning body of literature suggests that the global financial crisis has retrenched the trends towards the globalization
and integration of international financial markets (Milesi-Ferretti and Tille, 2011; Doidge et al., 2020; Akbari, Carrieri, and Mal­
khozov, 2022). For instance, the COVID-19 outbreak not only led to the collapse and recovery of stock markets around the world but
also caused an unprecedented pause in the global economy to control the spread of the virus (e.g., Gormsen and Koijen, 2020; Davis
et al., 2021), which may cast further doubt on the benefits of global diversification.
Attig and Sy (2021), however, demonstrate that international diversification continues to outperform industrial diversification
deep in the post-millennium period, especially when investors consider emerging markets.6 Attig and Sy conclude that the benefits of
diversification persist through bad times, demonstrating their countercyclical nature and proving their value when investors need
them most.7 Along the same lines, Viceira and Wang (2022) conclude that diversification gains have not declined for long-horizon
investors (despite the secular increase in global stock correlations). De Santis and Sarno (2008) suggest that there are gains to
holding diversified portfolios despite that the strengthened comovement of returns across countries. Li, Sarkar, and Wang (2003)
conclude that integration of world equity markets reduces, but does not eliminate, the diversification benefits of investing in emerging
markets, even after subjecting portfolios to constraints such as short selling. They also show that the international diversification
benefits remain substantial for US equity investors. Relatedly, Fletcher and Marshall (2005) provide evidence on the benefits of moving
from a domestic strategy to an international strategy for UK investors (even in the presence of short-selling restrictions).8
Although the literature reviewed above indicates that the benefits of international diversification are compelling, more remains to
be known about the determinants of these benefits. For instance, it is not clear a priori whether international diversification is still
relevant. For instance, one could argue that the gradual removal of barriers to international trade and investment entails that not only
will purely local firms be more impacted by external competition, but also that multinational corporations themselves are likely to
consolidate their international activities, making them like internationally diversified portfolios of foreign assets. In this context,
global factors will be more important, country indexes will be more correlated, and the benefits of international diversification will be
lower. Alternatively, since greater political, financial, and economic integration renders markets more vulnerable to global shocks
(Lehkonen, 2015), it would not be too surprising that trends towards globalization may retrench following major financial crises
(Milesi-Ferretti and Tille, 2011; Doidge et al., 2020; Gormsen and Koijen, 2020). Under this alternative view, because the past quarter-
century has witnessed major financial crises such as the 2007–2009 global financial crisis and the COVID-19 pandemic, the associated
globalization reversals will likely lead to further segmentation from the global market, and the benefits of international diversification
are likely to be larger. We address this question by looking at the benefits of diversification in the most recent 27 years. Equally
important, as stated at the outset, extant literature is still devoid of convincing evidence on local risks behind the benefits of inter­
national portfolio diversification. We attempt also to fill this important void.

3. Data and descriptive statistics

We rely on monthly data from Thomson Reuters Datastream to conduct our study. We consider 48 markets that have sufficient data
to be considered. Of these, 25 are classified as developed markets (DMs) based on the FTSE classification. The other 23 are emerging
markets (EMs).
For each country, we consider only the main exchange, which typically accounts for most traded securities. This restriction pre­
vents cross-listed shares from being considered several times. However, for a few countries, stocks are listed not only on one dominant
exchange but on several local exchanges. Those are China (Shanghai and Shenzhen stock exchanges), Germany (Frankfurt, Stuttgart,

4
See also King and Wadhwani (1990), Bertero and Mayer (1990), Erb, Harvey, and Viskanta (1996), King et al. (1995), De Santis and Gerard
(1997), Ang and Chen (2002), Campbell et al. (2002), and Das and Uppal (2004).
5
Other studies point to the presence of home country bias. This is because investors tend to invest largely in their home country as they prefer
trading and investing in the familiar, and thus ignoring the benefits of international diversification (e.g., Huberman 2001; Grinblatt and Keloharju,
2001).
6
Driessen and Laeven (2007) show that investors in developing countries benefit the most from investing abroad.
7
Flavin and Panopoulou (2009) suggest that market stable linkages favor international diversification strategies even during turbulent periods.
8
Importantly, Fletcher, Paudyal, and Santoso (2019) provide evidence on diversification gains in bond markets, even in the presence of no short-
selling constraints.

3
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Hamburg, and Munich stock exchanges), India (Bombay and National stock exchanges), Malaysia (Kuala Lumpur and MESDAQ stock
exchanges), Russia (Russian Trading System and Moscow stock exchanges), and the United States (NYSE, Nasdaq, and Amex). For
these countries, we consider stocks listed in multiple local exchanges.
We focus on stocks by dropping the non-equity instruments such as preferred stocks, warrants, trusts, GDRs, and other non-common
equity. To minimize the potential for errors in the database, we used the screening procedure of Ince and Porter (2006). In particular,
we dropped: (i) stocks that do not have any industry classification based on FTSE Russell’s Level 3 Industry Classification Benchmark or
have missing information, (ii) stocks with less than 12 monthly returns available, (iii) observations with trailing zero-returns, missing
return, or missing size, (iv) observations with lagged return index (RI) less than 0.01, (v) observations whose price is higher than one
million, (vi) observations with an end-of-previous-month share price lower than $1, (vii) returns falling out of the 0.1 % and 99.9 %
percentile ranges in each month, and (viii) returns above 300 % that reverse within a month.9 With this screening process, we end up
with a sample of 41,989 stocks (about 6.6 million monthly observations) from January 1995 to December 2021.10 The sample includes
total monthly dollar-denominated returns, size (market capitalization), and a description of the stocks’ country of incorporation and
industry classification.
Table 1 presents summary statistics describing the countries and industries covered. The second column shows that while about
half of the countries have a maximum number of 20 industries present, some countries lack some industries and Colombia and Ireland
have the least representation with 13 and 14 industries, respectively. Media, which is present in only 41 countries, is the least prevalent
industry, while Foods, Construction Materials, and Industrial Goods & Services are the only industries represented in all 48 countries.
With 37.57 % of global capitalization, the US is the most capitalized, followed by Japan (11.79 %) and China (6.17 %). At the other
extreme, Bangladesh (0.04 %), Kenya (0.03 %), and Sri Lanka (0.02 %) are the least capitalized. Industrial Goods & Services is the
largest industry, with 12.09 % of the global capitalization, while Media is the smallest with 1.77 % of the global capitalization.
The value-weighted average monthly country returns for 1995–2021 are highest for Indonesia (3.51 %) and Russia (2.33 %) and
lowest for Japan (0.37 %) and Greece (0.47 %). The differences in industry’s returns are less pronounced, going from 0.67 % for
Utilities to 1.36 % for Technology. Countries and industries also differ considerably in terms of volatility, measured by the standard
deviation of the value-weighted monthly country or industry index. The volatility of the country indexes ranges from 4.57 % for
Switzerland to 20.09 % for Indonesia, while those of the industry indexes go from 3.52 % for Personal Care to 7.15 % for Basic
Resources.

4. Empirical design

4.1. Measuring the benefits of diversification

When it comes to measuring the benefits of international and industrial portfolio diversification, the dummy variable model of
Heston and Rouwenhorst (1994) is the method of choice, as it provides the pure country and industry effects, whose variability directly
determines the benefits of international and industrial diversifications.
To obtain the pure effects, we estimate the following cross-sectional regression:
∑
48 ∑
20
Rjt = αt + γ ct Ccj + λit Iij + ∊jt (1)
c=1 i=1

where Rjt is the return of stock j for month t, Ccj is a dummy variable that takes one if stock j belongs to country c (c = 1to48) and zero
otherwise, Iij is a dummy variable that takes one if stock j belongs to industry i (i = 1to20) and zero otherwise, ∊jt is a zero-mean
residual, and αt , γct , and λit are parameters to be estimated. For completeness, we briefly summarize the methodology below.
To be able to estimate (1) without encountering a problem of perfect multicollinearity, we impose the following restrictions:
∑
48
ωcgt− 1 γct = 0 (2)
c=1

∑
20
ωigt− 1 λit = 0 (3)
i=1

where ωcgt− 1 and ωigt− 1 denote the capitalization weights of country c and industry i in the global market portfolio at time t − 1. To

9
That is, if Rt > 300% or Rt− 1 > 300% and (1 + Rt )(1 + Rt− 1 ) − 1 < 50%, we set Rt and Rt− 1 to missing.
10
While the Datastream database begins in the early 1970s, most emerging markets don’t have decent coverage until the mid-1990s. Given this
data limitation, we begin the study in 1995 to allow for sufficient heterogeneity across all seven risk measures considered.

4
N. Attig et al.
Table 1
Summary Descriptive Statistics for Countries and Industries. This table reports summary descriptive statistics by country and industry. The statistics are the number n of industries in each country or
of countries in each industry, the average monthly weight relative to the global capitalization, the value-weighted average monthly index return, and the associated standard deviation (SD). The sample
period is January 1995 to December 2021.
Developed markets Emerging markets Industries

n Weight Mean SD n Weight Mean SD n Weight Mean SD

Australia 20 1.91 1.02 6.22 Argentina 18 0.10 0.88 10.94 Technology 44 9.07 1.36 6.92
Austria 17 0.19 0.88 6.81 Bangladesh 18 0.04 1.44 9.82 Telecommunications 47 6.53 0.70 5.54
Belgium 18 0.56 0.77 6.42 Brazil 20 1.54 1.32 10.80 Media 41 1.77 0.69 5.61
Canada 20 2.67 1.16 5.71 Chile 19 0.39 0.59 6.80 Health Care 47 9.01 1.05 3.98
Denmark 17 0.48 1.20 5.49 China 20 6.17 1.33 8.54 Banks 47 10.97 0.80 6.10
Finland 19 0.62 1.20 8.07 Colombia 13 0.13 1.23 9.94 Financial Services 47 4.60 0.93 5.63
France 20 4.09 0.99 6.16 Egypt 20 0.09 0.52 7.72 Insurance 43 3.97 0.86 6.10
Germany 20 3.43 0.80 6.58 Greece 18 0.17 0.47 11.02 Real Estate 45 2.08 0.79 5.35
Hong Kong 20 2.82 0.90 6.54 India 20 2.06 1.10 8.92 Automobiles & Parts 42 2.92 0.81 5.80
Ireland 14 0.21 0.88 7.75 Indonesia 20 0.29 3.51 20.09 Consumer Products 47 3.50 0.86 5.23
Israel 20 0.25 1.02 6.39 Kenya 18 0.03 1.00 6.64 Retail 42 3.52 1.00 5.00
5

Italy 20 1.31 0.73 6.81 Malaysia 20 0.72 0.60 7.94 Travel and Leisure 45 2.13 0.77 5.67

Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Japan 20 11.79 0.37 5.21 Mexico 18 0.64 0.84 7.52 Foods 48 5.44 0.92 3.86
Netherlands 20 1.17 0.97 7.01 Morocco 17 0.08 0.94 4.65 Personal Care 46 2.72 0.86 3.52
New Zealand 17 0.10 0.97 6.03 Pakistan 19 0.09 0.90 8.83 Construction Materials 48 2.19 0.68 5.76
Norway 20 0.44 1.13 7.48 Peru 17 0.12 0.87 5.94 Industrial Goods 48 12.09 0.85 5.21
Poland 20 0.21 0.99 9.89 Philippines 18 0.25 0.58 7.68 Basic Resources 47 3.69 0.68 7.15
Portugal 17 0.15 0.71 6.80 Russia 20 0.93 2.33 13.37 Chemicals 43 2.24 0.85 5.57
Singapore 20 0.85 0.70 6.50 South Africa 19 0.74 0.90 7.98 Energy 44 6.92 0.93 6.49
South Korea 20 1.46 1.54 11.04 Sri Lanka 19 0.02 0.81 7.36 Utilities 47 4.61 0.67 3.99
Spain 20 1.30 0.91 7.16 Taiwan 20 1.65 0.81 7.11
Sweden 19 1.10 1.27 7.43 Thailand 20 0.52 0.74 9.56
Switzerland 19 2.21 0.99 4.57 Turkey 20 0.34 1.59 13.61
UK 20 5.99 0.75 5.13
US 20 37.57 1.13 4.94
Across all 48 countries Across all 20 industries
n Weight Mean SD n Weight Mean SD
Equally weighted mean 18.92 2.08 1.03 7.94 45.40 5.00 0.85 5.42
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

reduce the turnover of country and industry portfolios and limit the impact of trading costs, we allow the weights to change once per
year based on the market capitalization at the end (December) of the previous year.
We estimate (1) each month subject to restrictions (2) and (3) using weighted least squares (WLS). With these restrictions, αt is the
value-weighted return of the global market portfolio, γ ct is the pure country effect, and λit is the pure industry effect. The larger the
volatility of the pure country (industry) effects, the larger the benefits of international (industrial) diversification.11

4.2. Explaining the benefits of diversification

International portfolio diversification makes sense only when markets are somewhat segmented from the global portfolio, that is,
when nontrivial portions of international stock returns react to purely local factors. The question is then, precisely which local factors
determine the deviations from global returns?
To ascertain which local risks are responsible for the diversification benefits, we run the following pooled time-series cross-
sectional regression12:
|γct | = η + ΘRISK ct− 1 + ΦXct− 1 + ect (4)

where |γct | denotes the absolute pure country effect at time t, η is the intercept, Θ ∈ R7 and Φ ∈ R3 are regression parameter vectors,
RISK is a set of seven local risk measures, X is a set of three control variables, and ect is the residual term.
Campa and Fernandes (2006) examine five measures of international economic and financial integration and development (market
turnover, country openness, financial integration, industrial concentration, and country development). Although they can have an
impact on risk and thus influence the benefits of international diversification, these five variables cannot be directly interpreted as the
local diversifiable risks generating the diversification benefits. Therefore, for one or several of these variables to be useful in the risk
management strategy, further work needs to be done to determine if and how the priced variables relate to the local risks behind the
international diversification benefits.
The main advantage of our approach is that it allows us to directly evaluate which local risks determine the absolute pure country
effects. This helps not only to identify local risks that are important, but also to design and implement more effective international
diversification strategies. For example, if a risk measure—say political risk—significantly affects the absolute country effects in (4),
this implies that a top-down international diversification strategy will be ineffective if it only considers countries with high levels of
political risk. In other words, for the strategy to reduce risk appropriately, the political risk inherent in stock returns must be mitigated
by carefully allocating money among countries with different levels of political risk.
There are a few technical details worth noting. We estimate (4) with standard errors clustered at the country level to correct for
within-country residuals correlation. We also have year fixed effects to account for unobserved time trend that has homogeneous
impacts on all countries. Following Rajan and Zingales (1998) and La Porta and Shleifer (2008), we do not include country fixed effects
because some of our independent variables such as legal risk and credit risk are time-invariant. Finally, we estimate (4) using both OLS
and WLS. In the WLS estimations, to recognize the size differences among the countries, we weight each absolute country effect using
ωcgt− 1 .
Equation (4) uses monthly absolute pure country effects as the dependent variable, whereas Campa and Fernandes (2006)
aggregate this variable at the annual frequency to perform their main analysis.13 While they see this as “an estimation choice,” we
believe that the two specifications are different for one important reason: nonstationarity. In unreported results, we cannot reject the
null of nonstationarity for all 48 yearly absolute deviation series in augmented Dickey and Fuller (1979) and Phillips and Perron (1988)
tests. In contrast, none of the monthly series is deemed nonstationary using the same tests. Besides nonstationarity, Rossana and Seater
(1995) argue that aggregation to annual frequency may lead to substantial information losses.14 These issues motivate our use of
absolute monthly effects in all our regressions.
As local risks, we consider market, political, economic, financial, inflation, legal, and credit risks. Of these, only political risk has
been studied as a potential determinant of international diversification gains (Cosset and Suret, 1995), but not in recent data. To
measure market risk, we estimate a GARCH process (Bollerslev, 1986) to obtain a time series of the conditional variance associated
with each country’s value-weighted market index. The market risk for each month is the squared root of the estimated conditional

11
Caution should be exercised in interpreting our finding since the Heston–Rouwenhorst approach does not account for market frictions such as
transaction costs and short-selling constraints. Several related studies using alternative approaches such as mean–variance spanning tests show that
frictions can alter the benefits of international diversification (e.g., De Roon, Nijman, and Werker, 2001). We refer the reader to Fletcher and
Marshall (2005) who provide evidence on the benefits of international diversification (for UK investors), even in the presence of short-selling
restrictions.
12
In our model, the dependent variable is generated from a prior regression. However, the errors in the generated regressand do not trigger a
serious econometric problem. As Greene (2000, p.376) notes: “As long as the regressor is measured properly, measurement error on the dependent
variable can be absorbed in the disturbance of the regression and ignored.”.
13
It should be noted that the authors also presented in their robustness section the results obtained with monthly data and found them to be
consistent with those obtained with annual data.
14
Temporal aggregation usually involves averaging and can therefore reduce noise in the data. However, it can also eliminate the underlying
characteristics of the data, resulting in substantial information loss.

6
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

variance.15 If investors price risk at the local level (e.g., Griffin, 2002), we would expect this variable to explain a significant portion of
countries’ pure effects. As in Bekaert, Harvey, Lundblad, and Siegel (2016), we use the Political Risk Index (PRI) from the International
Country Risk Guide (ICRG) as our proxy of political risk. PRI reflects political risk (Bekaert, Harvey, Lundblad, and Siegel, 2014)
because it assigns risk points to 12 political variables, including government stability, internal conflicts, external conflicts, and reli­
gious and ethnic tensions.
Similarly, we rely on ICRG’s Economic Risk Index (ERI) and Financial Risk Index (FRI) regarding economic and financial risks. ERI
aggregates five economic risk components related to GDP, budget deficit, and current account balance. FRI combines information on
five financial risk variables related to debt, current account, liquidity, and exchange rate. We measure inflation risk by the World
Bank’s annual growth rate of the GDP implicit deflator, which captures the rate of price change in the economy as a whole. Our in­
dicator of legal risk is Djankov, La Porta, Lopez-de-Silanes, and Shleifer’s (2008) anti-self-dealing index, which measures the level of
protection of minority shareholders against self-dealing by the controlling shareholder. We use the index of creditors’ rights from
Djankov, McLiesh, and Shleifer (2007) to measure credit risk. This index reflects the strength of creditors’ recourses against defaulting
debtors in different jurisdictions.
When trying to ascertain which risks are being internationally diversified, it is important to account for country differences in
trading activity, market integration, and industrial structure, which may affect diversification benefits. Therefore, following the
literature (e.g., Campa and Fernandes, 2006), we control for turnover ratio, openness, and industrial concentration. Turnover ratio is
the annualized ratio of the value of domestic shares traded to their market capitalization. Openness, or more precisely trade openness,
is the ratio of the sum of exports and imports of measured goods and services to gross domestic product. Finally, we measure a
country’s industrial concentration as the sum of the squared differences between the weights of each industry in the country and the
∑ ( )2
global portfolio: ICct = 20 i=1 ωict − ωigt . The Appendix defines in more detail the variables and provides their sources.
Table 2 shows summary statistics for the seven measures of risk. With 48 countries and 324 monthly observations, our largest
possible sample is 15,552 country-month observations. For most of the variables, we were able to get close to the maximum coverage.
Turnover ratio has the most missing values, but even so, it covers over 81 % of the maximum sample. The mean of all variables lies
between the lower and upper quartiles, suggesting the absence of large outliers. The last two columns show the time-series cross-
sectional means of each variable for the emerging and developed markets subsamples. Developed markets appear less risky than
emerging markets on all measures.

5. The benefits of diversification

Table 3 shows the average and standard deviation of the pure country and industry effects for the period January 1995 to December
2021. The average monthly pure country effects range from − 0.51 % for Japan to 2.64 % for Indonesia. This is more than four times the
monthly average pure industry effects range, from − 0.20 % for Media and Basic Resources to 0.51 % for Technology. The differences
between the country and industry effects are even more striking when considering their volatilities. The monthly standard deviations
of the pure country effects range from 1.83 % for the US to 18.46 % for Indonesia, a distribution that is more than six times the range of
the standard deviations of the pure industries effects, from 1.29 % to 3.98 % for Industrial Goods and Technology, respectively. As a
result, as seen in the figures at the bottom of Table 3, the pure country effects are on average more volatile than pure industry effects on
both the equal- and value-weighted basis. This evidence indicates that international diversification has been superior to industrial
diversification over the past quarter century.
To investigate the time variation in the benefits of international and industrial diversifications, we compute the capitalization-
weighted rolling average standard deviation of the pure country and industry effects using the following formulas:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ t
√ ∑
∑48 √ (γ cτ − γct )2
√
c
SDt = ωcgt− 1 τ=t− 23 (5)
c=1
23

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ t
√ ∑
√ (λ − λit )2
∑
20 √τ=t− 23 iτ
SDit = ωigt− 1 (6)
i=1
23

∑ ∑
where γ ct = 24− 1 tτ=t− 23 γ cτ and λit = 24− 1 tτ=t− 23 λiτ are the average country and industry effects in the 24-month window,
respectively.
Fig. 1 represents the evolution of the standard deviations of the pure country and industry effects from December 1996 to December
2021. The plot leaves no doubt that international diversification is a more effective risk management tool than industrial diversifi­
cation. Consistent with Baca, Garbe, and Weiss (2000), Cavaglia, Brightman, and Aked (2000), and L’Her, Sy, and Tnani (2002),
industrial diversification emerges as better than international diversification around the turn of the millennium. Also consistent with
Brooks and Del Negro (2004), we see that this domination of industrial diversification was temporary: international diversification
dominates in the post-IT bubble period.

15
We thank a referee for suggesting this method of measuring market risk.

7
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Table 2
Summary Statistics on the Determinants of Diversification. This table presents the summary statistics (number of observations (N), mean,
standard deviation, minimum, lower quartile, median, upper quartile, and maximum) of the seven measures of risk and control variables used in
regression (4): market risk, political risk, economic risk, financial risk, inflation risk, legal risk, and credit risk, as well as the countries turnover ratio,
trade openness, and industrial concentration. The variables are described in the Appendix. In the last three columns, we report each variable’s mean
for the subsamples of developed markets (DMs) and emerging markets (EMs), and the p-value associated with the mean difference test. The sample
period is January 1995 to December 2021.
Full universe Subsamples

Variable N Mean Standard Deviation Min Median Max DMs EMs Difference
Mean Mean p-value

Market risk 15,552 7.60 3.95 3.13 6.53 63.06 6.46 8.83 [0.000]
Political risk 13,932 73.61 12.34 42.00 76.00 97.00 82.88 62.96 [0.000]
Economic risk 13,932 38.11 4.61 17.50 38.50 50.00 40.58 35.27 [0.000]
Financial risk 13,932 38.78 4.56 17.50 38.50 50.00 39.32 38.16 [0.000]
Inflation risk 15,228 5.02 9.95 − 6.01 2.54 144.01 1.90 8.57 [0.000]
Legal risk 15,228 0.51 0.23 0.17 0.46 1.00 0.54 0.48 [0.000]
Credit risk 15,228 1.92 1.10 0.00 2.00 4.00 2.17 1.64 [0.000]
Turnover ratio 12,672 62.42 64.12 0.00 43.50 694.43 73.90 49.00 [0.000]
Openness 13,884 76.55 60.10 15.61 60.51 441.60 92.29 58.38 [0.000]
Concentration 15,552 0.12 0.10 0.01 0.09 0.83 0.11 0.12 [0.000]

Table 3
Pure Country and Industry Effects. The table gives the pure country and industry effects’ mean and standard deviation (SD). All figures are in
percent per month. The pure effects are based on the WLS estimation of (1) subject to restrictions (2) and (3). The sample period is January 1995 to
December 2021 (324 months).
Developed markets Emerging markets Industries

Mean SD Mean SD Mean SD

Australia 0.31 3.21 Argentina 0.10 9.28 Technology 0.51 3.98

Austria 0.19 3.73 Bangladesh 0.67 10.78 Telecommunications − 0.16 3.10
Belgium − 0.06 3.68 Brazil 0.55 8.42 Media − 0.20 2.43
Canada 0.41 2.60 Chile − 0.18 5.24 Health Care 0.14 2.74
Denmark 0.33 3.64 China 0.51 8.22 Banks − 0.04 2.71
Finland 0.52 5.00 Colombia 0.46 8.35 Financial Services 0.07 1.92
France 0.15 2.96 Egypt − 0.22 7.52 Insurance − 0.04 2.76
Germany − 0.07 3.18 Greece − 0.36 8.74 Real Estate 0.04 2.64
Hong Kong 0.13 4.33 India 0.32 7.00 Automobiles & Parts 0.22 3.15
Ireland 0.05 5.38 Indonesia 2.64 18.46 Consumer Products 0.16 1.79
Israel 0.20 5.03 Kenya 0.22 6.39 Retail 0.16 2.64
Italy − 0.05 4.22 Malaysia − 0.19 7.23 Travel and Leisure − 0.03 2.61
Japan − 0.51 4.09 Mexico 0.09 5.20 Foods 0.06 2.85
Netherlands 0.10 3.76 Morocco 0.17 5.44 Personal Care − 0.05 3.07
New Zealand 0.15 4.32 Pakistan 0.12 9.00 Construction Materials − 0.03 2.00
Norway 0.41 4.16 Peru 0.06 5.43 Industrial Goods 0.08 1.29
Poland 0.26 7.21 Philippines − 0.18 6.39 Basic Resources − 0.20 3.57
Portugal 0.05 4.71 Russia 1.57 11.61 Chemicals 0.18 2.07
Singapore − 0.11 4.44 South Africa 0.09 5.26 Energy − 0.09 3.61
South Korea 0.65 8.22 Sri Lanka − 0.02 7.61 Utilities − 0.13 2.95
Spain 0.16 4.19 Taiwan − 0.12 5.43
Sweden 0.50 3.96 Thailand − 0.04 7.69
Switzerland 0.09 2.90 Turkey 0.84 11.87
UK − 0.02 2.24
US 0.25 1.83

Across all 48 countries Across all 20 industries

Mean SD Mean SD
Equally weighted mean 0.23 6.03 0.03 2.69
Value weighted mean 0.00 3.69 0.00 2.71

Fig. 2 also shows the within-region pure country effects, using the approach of Brooks and Del Negro (2005).16 Specifically, we
augment regression (1) by adding four regional dummy variables (North America, Developed Europe, Developed Asia, and EMs) and
impose an additional restriction that the weighted regional effects (the coefficients on the regional dummy variables) must average to

16
See also the analysis of Driessen and Laeven (2007), which confirms that investors in less developed countries benefits largely from investing
outside the region of the home country.

8
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Fig. 1. Evolution of Country and Industry Effects. The figure plots time-series of the capitalization-weighted average standard deviation of the
pure country effects (solid blue line), within-region pure country effects (dotted blue line), and pure industry effects (thin red line) using 24-month
rolling windows. The plots are from December 1996 to December 2021 (301 windows). (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

Fig. 2. Economic Benefits of International Diversification in terms of Variance, Value-at-risk, Expected Shortfall, and Sharpe Ratio. This
figure shows in Panel A the portfolio variance as the number of stocks in the portfolio increases, expressed as a percentage of a typical stock
variance, and in Panels B, C, and D the 5% Value-at-Risk (VaR), the 1% Expected Shortfall (ES), and the Sharpe ratio, respectively. The top (solid
red) line represents global portfolios that diversify across both countries and industries. The bottom (light gray) line represents portfolios that
diversify across industries within a single country. The middle (blue) lines are portfolios that diversify across 48 countries or 25 developed markets
(DMs) or 23 emerging markets (EMs) but within a single industry. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

9
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

zero. In doing so, the resulting pure country effects capture the within-region country effects, which track the risk-reduction gains
associated with international diversification within the average region. Consistent with Brooks and Del Negro (2005), who find that
regional risk diversification accounts for about half of the benefits of international diversification, we find that the dominance of
international diversification over industrial diversification is not assured when regional risk is not reduced. For instance, while full
international diversification dominates industrial diversification in 259 of the 301 24-month windows (86.05 % of cases), industrial
diversification dominates within-region international diversification in 180 windows (59.80 % of cases). Therefore, for the greater
benefits of international diversification to materialize, regional risk must be diversified by carefully allocating invested money across
various regions.
To economically verify whether the higher country effects translate into higher benefits for international diversification, following
the tradition set by Solnik (1974), we compute the proportion of a typical international stock variance that can be reduced by forming
portfolios of n (n = 1,⋯,75) stocks. For each n, we randomly draw 10,000 portfolios from a particular group and compute for each the
equally weighted average monthly excess returns (relative to the one-month T-Bill return). Using these excess return series for 2005 to
2020, we first compute the portfolios’ variances and average them across the 10,000 simulations.
Panel A of Fig. 2 shows that randomly combining stocks into globally diversified portfolios (across both countries and industries)
reduces the variance to about 6.74 % of a typical international stock variance. This risk reduction is very close to the 7.08 % global
limit obtained by Griffin and Karolyi (1998) for 1992 to 1995, confirming that diversification gains have not disappeared in the more
recent period we cover. Interestingly, a similar level of risk reduction (up to 7.67 % of the average stock variance) can be achieved
through international diversification across the 48 countries but within a single industry.
In comparison, the risk reduction limit reached by industrial diversification within a single country is much more modest, at 15.80
% of the variance of the typical international stock. The performance of industrial diversification is still lower than those obtained by
pure international diversifications across DMs and EMs, which can reduce the portfolio variances to less than 10 % of the average stock
variance.
As a measure of risk, the variance is often criticized because it considers both positive and negative movements in stock prices as
risky. However, we find no significant change in the five diversification strategies’ pecking order when considering two measures of
downside risk instead of variance. Panel B of Fig. 2, which shows how the portfolio’s 95 % confidence level Value-at-Risk (VaR) varies
with the number n of stocks in the portfolio, illustrates well the risk-curbing benefits of global and international diversification
strategies. For the five strategies considered, the maximum expected loss decreases as n increases. At the limit, the VaR achieved by
global diversification is − 7.81 %, which is the lowest among the five strategies. At the other extreme, the VaR achieved with pure
industrial diversification, which tops out at − 10.96 %, provides less downside protection than any of the three international di­
versifications considered.
The main weakness of VaR is that it does not consider the magnitude of losses beyond the VaR level (Yamai and Yoshiba, 2005). As
an alternative measure of downside risk, we consider the Expected Shortfall (ES) proposed by Artzner, Delbaen, Eber, and Heath
(1997), which accounts for losses beyond the VaR level. Panel C of Fig. 2 shows how the portfolio ES (at the 99 % confidence level)
varies with n. We presented the 99 % confidence level results to explore the risk of losing very big, but we can advance that similar
results hold when we consider the 95 % confidence level. Like VaR, the risk of losing as measured by ES decreases monotonically as
diversification increases through n. Global diversification generates the lowest conditional expected loss with a limit of − 16.93 %.
Although an international diversification in 25 DMs offers slightly less downside protection than international diversifications in 48
countries or 23 EMs, the three geographical strategies stand out as being better than industrial diversification.
In view of the results obtained on the three risk measures, we conclude that over 1995 to 2021, international diversification is a
more effective tool for reducing risks than industrial diversification. But is international diversification better than industrial diver­
sification from a performance viewpoint? Researchers such as Driessen and Laeven (2007), Li et al. (2003), and Fletcher and Marshall
(2005), among others, have used Sharpe ratios to evaluate the economic gains from international diversification strategies. As visible
in Panel D of Fig. 2, which depicts how the portfolio Sharpe ratio develops as the number of stocks in the portfolio increases, the greater
effectiveness of international diversification translates into better performance. While a global strategy that diversifies across both
countries and industries offers at the limit a Sharpe ratio of 0.24, a geographical strategy that spans 48 countries but within a single
industry comes very close with a limit Sharpe ratio of 0.23. In contrast, the limit Sharpe ratio obtained by a typical portfolio that is
industrially diversified but confined in a single country is lower, at 0.19, confirming the superior performance of international
diversification.17

6. The sources of diversification benefits

Table 4 presents the results from regression (4). We consider three specifications for each of the two estimation methods. To
minimize the risk of data mining, we give more weight to relations that are robust across several specifications.
The estimated coefficient on market risk is positive and significant at the 1 % level in all six specifications, which indicates that
greater market risk amplifies the magnitude of the benefits from international diversification. This result makes sense as it is well
known that stock returns react primarily to local market news; this is probably why standard asset-pricing models such as the CAPM of

17
A caveat is worthy of note. All our results are based on monthly returns. Yet, the benefits of global equity diversification may depend on the
investment horizon (Viceira and Wang 2022). With this in mind, the higher benefits of international diversification documented here can be seen as
lower bounds when longer return horizons are considered.

10
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Sharpe (1964) and Lintner (1965) consider market risk as the key determinant of the cross-section of expected asset returns. Further,
back-of-the-envelope calculations indicate that market risk carries the most economic significance among all regressors in any of the
specifications. A one-standard-deviation increase in market risk (see Table 2) leads to no less than a 1.20 % increase in the absolute
country effects. Market risk is also the most important contributor to international diversification benefits when considered alone, with
a coefficient of determination of about 25.6 %.
It is important to note that a high score on ICRG’s political risk measure indicates low political risk. Therefore, as expected, political
risk always loads negatively and significantly at least at the 5 % level. The negative sign means that greater political risk leads to a
higher segmentation from the global benchmark. Alone, political risk explains about 15.8 % of the absolute effects’ variations, second
only to market risk. The significance of political risk in our regressions confirms Cosset and Suret (1995), who find that including high-
political-risk countries in the investment opportunity set improves the risk-return trade-off significantly. The result is also consistent
with Frijns, Tourani-Rad, and Indriawan (2012), who find that political crises tend to decrease the integration of emerging markets.
The economic risk measure considered takes a high value when the economic risk is low. Given this, the statistically significant (at
the 5 % level) negative coefficients consistently obtained in two OLS regressions suggest that investors require a significant premium to
bear local economic risk. This result agrees with the finding of Campa and Fernandes (2006) that higher economic development is
associated with lower absolute country effects. However, we cannot find any significance for economic risk in the WLS estimations.
Our reading of this result is that economic risk is more important for the absolute effects of small countries and that it remains marginal
for investors who seek to reflect country size in their holdings.
In an ideal framework where local and foreign investors face the same set of investment opportunities (i.e., PPP is maintained),
Grauer, Litzenberger, and Stehle’s (1976) version of the international CAPM will hold. However, if PPP does not stand, inflation risk
should be priced beside global market risk (Adler and Dumas, 1983). Confirming the theoretical predictions, inflation risk is an
important dimension of international diversification in all our specifications. Therefore, we conclude that inflation risk is an important
part of the key risks that are mitigated when investors diversify their investments across international markets. The effect of inflation
risk appears stronger in the WLS regression, which is consistent with Bekaert and Wang (2010), who find that inflation risk is easier to
hedge in emerging markets than in developed markets.
We do not distinguish any significant effect of financial risk in the multivariate OLS estimations, which is surprising not only
because FRI aggregates macroeconomic variables such as foreign debt and liquidity, which should be crucial for smaller and less
mature countries, but also because Erb et al. (1996) find that FRI is the ICRG index which is the most informative about expected future
returns. Further, the effect of financial risk is significant in the three WLS estimations, but the coefficients are also positive. This
evidence suggests that financial risk is primarily relevant for larger markets and that better financial conditions translate into higher
absolute country effects. While this result may seem counterintuitive at first glance, it is nonetheless consistent with the stylized facts
documented in the literature. For example, Milesi-Ferretti and Tille (2011) find that the financial fallout from the global financial crisis
is a highly heterogeneous phenomenon, with capital flows retrenchment (a form of financial risk) shorter-lived in emerging economies
than in mature and industrialized economies. Doidge et al. (2020) also argue that financial globalization has increased among EMs
after the crisis but has gone the wrong way among DMs, noting in their summary: “The valuation gap for firms from developed markets
increases by 31 % after the global financial crisis—a reversal in financial globalization—while the gap for firms from emerging markets
(excluding China) stays stable.”.
None of the coefficients associated with credit and legal risks is significant in determining absolute pure country effects. The lack of
significance appears, at first glance, surprising, given the importance of investors’ legal protection and its effects on capital markets (La
Porta, Lopez-De-Silanes, Shleifer, and Vishny 1997, 1998; Shleifer and Vishny, 1997; Stulz, 2005). Arguably, the non-significance of
legal and credit risks can be attributed, at least partially, to the potential noisiness of the metrics. We take caution in concluding that
the perils underlying (i) the legal protection of minority shareholders and (ii) the recourses that creditors have against defaulting
debtors are not the local risks that allow international diversification to be beneficial.18
Adding the three control variables did not change any of our conclusions. We find no significant turnover effect on the absolute pure
country returns. This lack of significance is not due to the inclusion of risk measures, given that turnover remains insignificant even if it
is alone. Therefore, we conclude that turnover is not a key determinant of international diversification benefits.
Openness appears significant at the 5 % level in two OLS estimations. However, the sign of the coefficients in the two regressions is
different. While the negative coefficients in the univariate regression suggest that greater openness is associated with lower country
shocks, hence lower international diversification benefits, the reliably positive coefficient in the multivariate regression is less intuitive
because it implies that openness leads to higher absolute country effects and higher diversification benefits.19 Given the lack of
robustness of the openness effects across the specifications, we conclude that this variable is not a key driver of the benefits of in­
ternational diversification. This finding somewhat supports Stulz (2005), who provides an excellent framework for understanding why
agency problems can render full market integration elusive despite the dramatic reduction in explicit barriers to international
investment.

18
The lack of significance of these two risk measures is not caused by their common covariation with ICRG’s risk measures, which account for
many facets of the quality of a country’s institutions and corporate governance mechanisms. This is so because the measures of legal and credit risks
are not significant in the univariate regressions reported in columns (1) and (4).
19
Greater openness and integration can increase vulnerability to external shocks (Claessens, Dell’Ariccia, Igan, and Laeven, 2010; Blanchard, Das,
and Faruqee, 2010) and tends to reverse in the aftermath of bad shocks (e.g., Doidge et al., 2020; Akbari et al., 2022), which could explain the
positive and significant estimate.

11
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Table 4
Source of the Gains from International Diversification.
The dependent variable in all regressions is the monthly absolute pure country effects:
|γct | = η + ΘRISK ct− 1 + ΦXct− 1 + ect (4)

where RISK is a set of seven local risk measures (market, political, economic, financial, inflation, legal, and credit) and X is a set of three control
variables (turnover ratio, openness, and industrial concentration). The variables are described in the Appendix. Panel A shows the OLS results, while
Panel B shows the WLS results. The WLS estimation weights each observation by the country’s capitalization weight in the global market portfolio. In
both methods, we include year fixed effects and use t-statistics (in parentheses) based on standard errors clustered at the country level. Columns 1 and
4 show the slope coefficients obtained when one regressor is considered alone. The sample period is January 1995 to December 2021. ***, **, and *
indicate significance at the 1 %, 5 %, and 10 % levels, respectively.
A. Ordinary Least Squares (OLS) B. Weighted Least Squares (WLS)

Univariate Multivariate Univariate Multivariate

(1) (2) (3) (4) (5) (6)

Market risk 0.611*** 0.512*** 0.513*** 0.494*** 0.311*** 0.303***

(8.31) (5.55) (5.55) (8.46) (5.02) (4.98)
Political risk − 0.132*** − 0.051*** − 0.055*** − 0.128*** − 0.065*** − 0.063***
(-4.76) (-5.91) (-5.73) (-7.88) (-4.94) (-5.44)
Economic risk − 0.340*** − 0.047 − 0.080** − 0.068 − 0.023 − 0.038
(-4.99) (-1.46) (-2.09) (-0.65) (-0.44) (-0.77)
Financial risk − 0.193*** − 0.004 − 0.002 0.115* 0.055 0.066**
(-2.76) (-0.20) (-0.08) (1.91) (1.62) (2.18)
Inflation risk 0.137*** 0.033*** 0.029** 0.133*** 0.065*** 0.050**
(4.03) (2.84) (2.38) (6.06) (5.81) (2.53)
Legal risk − 0.806 0.633 0.384 − 0.499 − 0.197 − 0.001
(-0.85) (1.67) (1.00) (-0.42) (-0.24) (-0.00)
Credit risk − 0.276 − 0.001 0.046 0.367 0.280 0.185
(-1.43) (-0.01) (0.69) (1.25) (1.43) (1.58)
Turnover ratio − 0.000 0.000 − 0.004 0.000
(-0.03) (0.02) (-0.65) (0.35)
Openness − 0.007** 0.003** 0.007 0.000
(-2.39) (1.96) (1.38) (0.12)
Concentration 4.258 2.745*** 8.317** 5.153***
(1.60) (3.81) (2.60) (4.97)
Constant 5.678*** 6.623*** 4.250** 4.117
(4.46) (4.17) (2.04) (1.57)
Year fixed effects YES YES YES YES YES YES
Observations 13,608 11,256 13,608 11,256
R-squared 0.302 0.316 0.255 0.262

Prior studies (e.g., Roll, 1992; Campa and Fernandes, 2006) predict a relation between industrial concentration on the benefits of
international diversification. This is so because more specialized countries, in terms of industrial organization, tend to be more
impacted by the global shocks affecting the industries on which their economies rely. Consistent with this view, the coefficient on
industrial concentration is of the expected positive sign and significant in three of the four specifications. Given this, we conclude that
industrial concentration can affect the benefits of international diversification.

7. Conclusion

This research contributes to the international diversification literature by addressing two important research questions. First, is
international diversification still the best risk-reduction tool (compared to industrial diversification) over the past quarter century? The
1995–2021 period is interesting because it includes four major crises (Asian crisis, IT bubble burst, global financial crisis, and the
COVID-19 pandemic) that could affect the relative benefits of international and industrial diversifications because globalization and
integration trends are known to retrench during crises (Milesi-Ferretti and Tille, 2011; Doidge et al., 2020; Gormsen and Koijen, 2020).
Looking at this diversification benefits during period of crises is important to both regulators seeking to how better protected investors
and practitioners seeking to better manage their portfolios. Second, what local risks are behind the benefits of international diversi­
fication? From a practical point of view, a better understanding of the mitigable sources of risks inherent in international investing
enables the design and implementation of more effective risk management tools. We consider seven potential candidates: market risk,
political risk, economic risk, financial risk, inflation risk, legal risk, and credit risk. Of these, only political risk has been considered a
potential international diversification determinant in the published literature (e.g., Cosset and Suret, 1995; De Santis et al., 1999).
Based on a large sample of nearly 42,000 stocks, we find that international diversification still dominates industrial diversification
over the past quarter century. This domination persists after the IT bubble burst and is not altered during the global financial crisis or
the more recent COVID-19 pandemic. Further, we find that the benefits of international diversification come primarily from the
mitigation of market, political, and inflation risks. Economic risk appears to be important only for global investors who bet heavily on
small markets while, on the other hand, financial risk is only important for global investors that place more weight on large markets.

12
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Risks linked to the quality of legal and credit institutions appear to be less important for diversification.
Therefore, in their risk management and asset allocation strategies, investors need to ensure that the international markets they
select are sufficiently heterogeneous with respect to the local risks deemed important. By underscoring the importance of international
diversification in mitigating portfolio risk, our evidence points to the relevance of liberalizing international financial markets and
introducing policies and regulations in developing markets to enable investors from all corners of the world to benefit from inter­
national diversification. As suggested by Driessen and Laeven (2007), the introduction of globally oriented mutual funds may foster
diversification benefits. However, our fresh evidence underscores the need to design these funds so that market, political, and inflation
risks, which appear to be significant, are well-diversified. For funds limited to developed markets, the financial risks inherent in in­
ternational equities need to be further diversified, while emerging market investors should pay more attention to further diversifying
the economic risks inherent in these less mature and developing markets.

CRediT authorship contribution statement

Najah Attig: Conceptualization, Methodology, Writing – original draft, Formal analysis. Omrane Guedhami: Conceptualization,
Writing – review & editing, Writing – original draft. Gregory Nazaire: Conceptualization, Writing – review & editing, Writing –
original draft. Oumar Sy: Conceptualization, Data curation, Methodology, Writing – original draft, Formal analysis.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Data availability

The authors do not have permission to share data.

Acknowledgement

None.

Appendix. Variable definitions and data sources

Variable Definition Source

A. Risk measures
Market risk We use the conditional standard deviation of the value-weighted market returns, using a GARCH process. Datastream
Political risk We measure political risk by using the Political Risk Index (PRI) provided by the International Country Risk Guide ICRG ([Link]
(ICRG). The PRI assesses the political risk by assigning risk points to 12 political risk components, including [Link])
government stability, internal conflict, external conflict, military in politics, religious tensions, ethnic tensions,
and bureaucracy quality. The PRI ranges from 0 to 100.
Economic risk We use ICRG’s Economic Risk Index (ERI) to measure economic risk. The index aggregates five economic risk As above
components: GDP per capita, GDP growth, inflation, budget deficit, and current account balance. The ERI ranges
from 0 to 50.
Financial risk We measure financial risk through ICRG’s Financial Risk Index (FRI). It aggregates five financial risk variables: As above
foreign debt, debt service, current account, international liquidity, and exchange rate stability. The FRI ranges
from 0 to 50.
Inflation risk Measured by the GDP implicit deflator’s annual growth rate and shows the rate of price change in the economy as a [Link]
whole. The GDP implicit deflator is the ratio of GDP in current local currency to GDP in constant local currency.
Legal risk We measure the legal risk faced in a country by relying on the measure of the legal protection of minority Djankov et al. (2008)
shareholders proposed by Djankov et al. (2008). We use the anti-self-dealing index, which measures the strength of
minority shareholder protections against self-dealing by the controlling shareholder. The index ranges from 0 % to
100 %, where higher values indicate that a country imposes stronger controls on self-dealing transactions,
providing more protection to minority shareholders.
Credit risk We use Djankov et al.’s (2007) creditor right index to measure credit risk. It reflects the legal rights of creditors Djankov et al. (2007)
against defaulting debtors in different jurisdictions. The index ranges from 0 to 4, with higher scores indicating
strong creditor rights or lower risks.
B. Control variables
Turnover Turnover Ratio is the value of domestic shares traded divided by their market capitalization and is annualized by [Link]
ratio multiplying the monthly average by 12.
Openness Openness is the ratio of the sum of exports and imports of goods and services measured to gross domestic product. [Link]
The data are collected from World Development Indicators.
Concentration Like the Herfindahl–Hirschman index, the level of industrial concentration in a country is computed by taking the Datastream
sum of the squared differences between the industries’ weights in the country and the corresponding weights in the
global portfolio.

13
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

References

Adler, M., Dumas, B., 1983. International portfolio choice and corporation finance: A synthesis. J. Financ. 38, 925–984.
Akbari, A., Carrieri, F., Malkhozov, A., 2022. Can cross-border funding frictions explain financial integration reversals? Rev. Financ. Stud. 35, 394–437.
Ang, A., Chen, J., 2002. Asymmetric correlations of equity portfolios. J. Financ. Econ. 63, 443–494.
Artzner, P., Delbaen, F., Eber, J.M., Heath, D., 1997. Thinking coherently. Risk 10, 68–71.
Attig, N., O. Sy., 2023. Diversification during hard times. Financial Analysts Journal, forthcoming.
Baca, S.P., Garbe, B.L., Weiss, R.A., 2000. The rise of sector effects in major equity markets. Financ. Anal. J. 56, 34–40.
Bae, J.B., Elkamhi, R., Simutin, M., 2019. The best of both worlds: Accessing emerging economies via developed markets. J. Financ. 74, 2579–2617.
Baele, L., Inghelbrecht, K., 2009. Time-varying integration and international diversification strategies. J. Empir. Financ. 16, 368–387.
Bekaert, G., Harvey, C.R., Lundblad, C.T., Siegel, S., 2014. Political risk spreads. J. Int. Bus. Stud. 45, 471–493.
Bekaert, G., Wang, X., 2010. Inflation risk and the inflation risk premium. Econ. Policy 64, 755–806.
Bekaert, G., Hodrick, R.J., Zhang, X., 2009. International stock return comovements. J. Financ. 64, 2591–2626.
Bekaert, G., Harvey, C.R., Lundblad, C.T., Siegel, S., 2016. Political risk and international valuation. Finance 37, 1–23.
Bekaert, G., Urias, M.S., 1996. Diversification, integration and emerging market closed-end funds. J. Financ. 51, 835–869.
Bertero, E., Mayer, C., 1990. Structure and performance: Global interdependence of stock markets around the crash of October 1987. Eur. Econ. Rev. 34, 1155–1180.
Blanchard, O.J., Das, M., Faruqee, H., 2010. The initial impact of the crisis on emerging market countries. Brook. Pap. Econ. Act. 41, 263–307.
Bollerslev, T., 1986. Generalized autoregressive conditional heteroskedasticity. J. Econ. 31, 307–327.
Brooks, R., Del Negro, M., 2004. The rise in comovement across national stock markets: Market integration or IT bubble? J. Empir. Financ. 11, 659–680.
Brooks, R., Del Negro, M., 2005. Country versus region effects in international stock returns. J. Portf. Manag. 31, 67–72.
Butler, K.C., Joaquin, D.C., 2002. Are the gains from international portfolio diversification exaggerated? The influence of downside risk in bear markets. J. Int. Money
Financ. 21, 981–1011.
Campa, J.M., Fernandes, N., 2006. Sources of gains from international portfolio diversification. J. Empir. Financ. 13, 417–443.
Campbell, R., Koedijk, K., Kofman, P., 2002. Increased correlation in bear markets. Financial Analysts Journal 58, 87–94.
Cavaglia, S., Brightman, C., Aked, M., 2000. The increasing importance of industry effects. Financ. Anal. J. 56, 41–54.
Chua, D.B., Kritzman, M., Page, S., 2009. The myth of diversification. J. Portf. Manag. 36, 26–35.
Claessens, S., Dell’Ariccia, G., Igan, D., Laeven, L., 2010. Lessons and policy implications from the global financial crisis. Econ. Policy 62, 269–293.
Cosset, J.C., Suret, J.M., 1995. Political risk and the benefits of international portfolio diversification. J. Int. Bus. Stud. 26, 301–318.
Das, S.R., Uppal, R., 2004. Systemic risk and international portfolio choice. J. Financ. 59, 2809–2834.
Davis, S.J., Liu, D., Simon, X., 2021. Stock prices, lockdowns, and economic activity in the time of coronavirus. NBER Working Paper # 28320.
De Roon, F.A., Nijman, T.E., Werker, B.J., 2001. Testing for mean-variance spanning with short sales constraints and transaction costs: The case of emerging markets.
J. Financ. 56, 721–742.
De Santis, G., Gerard, B., Hillion, P., 1999. International portfolio management, currency risk and the euro. Working paper, Anderson Graduate School of
Management, University of California at Los Angeles.
De Santis, G., Gerard, B., 1997. International asset pricing and portfolio diversification with time-varying risk. J. Financ. 52, 1881–1912.
De Santis, R.A., Sarno, L., 2008. Assessing the benefits of international portfolio diversification in bonds and stocks. ECB Working Paper No 883.
Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive time series with a unit root. J. Am. Stat. Assoc. 74, 427–431.
Djankov, S., McLiesh, C., Shleifer, A., 2007. Private credit in 129 countries. J. Financ. Econ. 12, 77–99.
Djankov, S., La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 2008. The law and economics of self-dealing. J. Financ. Econ. 88, 430–465.
Doidge, D., Karolyi, G.A., Stulz, R.M., 2020. Is financial globalization in reverse after the 2008 global financial crisis? Evidence from corporate valuations. NBER
Working Paper #27022.
Driessen, J., Laeven, L., 2007. International portfolio diversification benefits: Cross-country evidence from a local perspective. J. Bank. Financ. 31, 1693–1712.
Erb, B.C., Harvey, C.R., Viskanta, E.T., 1996. Political risk, economic risk, and financial risk. Financ. Anal. J. 52, 29–46.
Ferreira, M.A., Gama, P.M., 2005. Have world, country, and industry risks changed over time? An investigation of the volatility of developed stock markets. J. Financ.
Quant. Anal. 40, 195–222.
Flavin, T.J., Panopoulou, E., 2009. On the robustness of international portfolio diversification benefits to regime-switching volatility. J. Int. Finan. Markets. Inst.
Money 19, 140–156.
Fletcher, J., Marshall, A., 2005. An empirical examination of the benefits of international diversification. J. Int. Finan. Markets. Inst. Money 15, 455–468.
Fletcher, J., Paudyal, K., Santoso, T., 2019. Exploring the benefits of international government bond portfolio diversification strategies. Eur. J. Financ. 25, 1–15.
Freimann, E., 1998. Economic integration and country allocation in Europe. Financ. Anal. J. 54, 32–41.
Frijns, B., Tourani-Rad, A., Indriawan, I., 2012. Political crises and the stock market integration of emerging markets. J. Bank. Financ. 36, 644–653.
Gande, A., Schenzler, C., Senbet, L.W., 2009. Valuation effects of global diversification. J. Int. Bus. Stud. 40, 1515–1532.
Gilmore, C.G., McManus, G.M., 2002. International portfolio diversification: US and Central European equity markets. Emerg. Mark. Rev. 3, 69–83.
Goetzmann, W., Li, L., Rouwenhorst, K., 2005. Long-term global market correlations. J. Bus. 78, 1–38.
Gormsen, N.J., Koijen, R.S.J., 2020. Coronavirus: Impact on stock prices and growth expectations. Rev Asset Pric Stud 10, 574–597.
Grauer, F.L.A., Litzenberger, R.H., Stehle, R.E., 1976. Sharing rules and equilibrium in an international capital market under uncertainty. J. Financ. Econ. 3, 233–256.
Greene, W.H., 2000. Econometric Analysis, Fourth Edition. Upper Saddle River, NJ: Prentice-Hall.
Griffin, J.M., 2002. Are the Fama and French factors global or country specific? Rev. Financ. Stud. 15, 783–803.
Griffin, J.M., Karolyi, G.A., 1998. Another look at the role of the industrial structure of markets for international diversification strategies. J. Financ. Econ. 50,
351–373.
Grinblatt, M., Keloharju, M., 2001. Distance, language, and culture bias: The role of investor sophistication. J. Financ. 56, 1053–1073.
Grubel, H., 1968. International diversified portfolio: Welfare gains and capital flows. Am. Econ. Rev. 58, 1299–1314.
Heston, S.L., Rouwenhorst, K.G., 1994. Does industrial structure explain the benefits of international diversification? J. Financ. Econ. 36, 3–27.
Hong, Y., Tu, J., Zhou, G., 2007. Asymmetries in stock returns: Statistical tests and economic evaluation. Rev. Financ. Stud. 20, 1547–1581.
Huberman, G., 2001. Familiarity breeds investment. Rev. Financ. Stud. 14, 659–680.
Ilmanen, A., Kizer, J., 2012. The death of diversification has been greatly exaggerated. J. Portf. Manag. 38, 15–27.
Ince, O.S., Porter, R.B., 2006. Individual equity return data from Thomson Datastream: Handle with care! J. Financ. Res. 29, 463–479.
King, M.A., Wadhwani, S., 1990. Transmission of volatility between stock markets. Rev. Financ. Stud. 3, 5–33.
King, M.A., Sentana, E., Wadhwani, S., 1995. Volatility and links between national stock markets. Econometrica 78, 901–934.
L’Her, J.F., Sy, O., Tnani, Y., 2002. Country, industry, and risk factor loadings in portfolio management. J. Portf. Manag. 28, 70–79.
La Porta, R., Lopez-De-Silanes, F., Shleifer, A., Vishny, R.W., 1997. Legal determinants of external finance. J. Financ. 52, 1131–1150.
La Porta, R., Lopez-De-Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and finance. J. Polit. Econ. 106, 1113–1155.
La Porta, R., Shleifer, A., 2008. The unofficial economy and economic development. Brook. Pap. Econ. Act. 47, 123–135.
Lang, L.H.P., Stulz, R.M., 1994. Tobin’s q, corporate diversification and firm performance. J. Polit. Econ. 102, 1248–1280.
Lehkonen, H., 2015. Stock market integration and the global financial crisis. Eur. Finan. Rev. 19, 2039–2094.
Levy, H., Sarnat, M., 1970. International diversification of investment portfolios. Am. Econ. Rev. 60, 668–675.
Lewis, K.K., 2000. Why do stocks and consumption imply such different gains from international risk sharing? J. Int. Econ. 52, 1–35.
Li, K., Sarkar, A., Wang, Z., 2003. Diversification benefits of emerging markets subject to portfolio constraints. J. Empir. Financ. 10, 57–80.
Lin, W.L., Engle, R.F., Ito, T., 1994. Do bulls and bears move across borders? International transmission of stock returns and volatility. Rev. Financ. Stud. 7, 507–538.

14
N. Attig et al. Journal of International Financial Markets, Institutions & Money 83 (2023) 101729

Lintner, J., 1965. The valuation of risk assets and the selection of risky investments in stock portfolio and capital budgets. Rev. Econ. Stat. 47, 13–37.
Longin, F., Solnik, B., 1995. Is the correlation in international equity returns constant: 1960–1990? J. Int. Money Financ. 14, 3–26.
Longin, F., Solnik, B., 2001. Extreme correlation of international equity markets. J. Financ. 56 (2), 649–676.
Markowitz, H., 1952. Portfolio selection. J. Financ. 7, 77–91.
Milesi-Ferretti, G.-M., Tille, C., 2011. The great retrenchment: International capital flows during the global financial crisis. Econ. Policy 26, 287–342.
Page, P., Panariello, R.A., 2018. When diversification fails. Financ. Anal. J. 74, 19–32.
Phillips, P.C.B., Perron, P., 1988. Testing for a unit root in time series regression. Biometrika 75, 335–346.
Phylaktis, K., Xia, L., 2006. Equity market comovement and contagion: A sectoral perspective. Financ. Manag. 36, 381–409.
Rajan, R.G., Zingales, L., 1998. Financial dependence and growth. Am. Econ. Rev. 88, 559–586.
Roll, R., 1992. Industrial structure and the comparative behavior of international stock market indices. J. Financ. 47, 3–42.
Rossana, J., Seater, J.J., 1995. Temporal aggregation and economic time series. J. Bus. Econ. Stat. 13, 441–451.
Sharpe, W., 1964. Capital asset prices: A theory of market equilibrium under conditions of risk. J. Financ. 19, 425–442.
Shleifer, A., Vishny, R., 1997. A survey of corporate governance. J. Financ. 52, 737–783.
Solnik, B., 1974. Why not diversify internationally rather than domestically? Financ. Anal. J. 30, 48–54.
Solnik, B., Bourcrelle, C., Le Fur, Y., 1996. International market correlation and volatility. Financ. Anal. J. 52, 17–34.
Stulz, R.M., 2005. The limits of financial globalization. J. Financ. 60, 1595–1638.
Viceira, L.M., Wang, Z., 2022. Global portfolio diversification for long-horizon investors. NBER Working Paper w24646. [Link]
Yamai, Y., Yoshiba, T., 2005. Value-at-risk versus expected shortfall: A practical perspective. J. Bank. Financ. 29, 997–1015.

15