SHOULD INTERNATIONAL PORTFOLIOS BE PRIMARILY

Transcription

1 SHOULD INTERNATIONAL PORTFOLIOS BE PRIMARILY DIVERSIFIED OVER COUNTRIES OR OVER INDUSTRIES? July 2007 Tilburg University Bachelor Thesis Finance S Supervisor: Dr. Lieven Baele GGH van der Loo.pdf

2 Table of contents ABSTRACT INTRODUCTION DEVELOPMENT OF RELATIVE IMPORTANCE OVER TIME STRUCTURAL FACTORS Integration of financial markets COMOVEMENT: SENTIMENTAL VERSUS FUNDAMENTAL VIEW METHOD DESCRIPTION CONCLUSION...20 REFERENCE LIST

3 Abstract Researchers argue a lot about the relative importance of country versus industry factors in international portfolio management. This paper gives an overview of the historical development of the relative importance and of the factors which are driving this development. Furthermore, it gives a view on the comovement of stock returns. Researchers agree that diversification over countries was the best way to diversify your international portfolio until the nineties. The most important factor shifting this importance is the integration of international financial markets. 2

4 1 Introduction There exists a lot of empirical evidence on the relative importance of industrial and country factors on the returns and risks of international portfolios. Lots of academics concluded that till about the turn of the century industry factors became more and more important. They explain that this is caused by structural factors which will be presented in chapter 3. The following papers belong to this category. Baca et al. (2000) believe that the relative importance of industrial sector effects compared to country effects has shifted upwards. In the world s largest equity markets, they state, the impact of industrial effects is now almost equal to the impact of country effects. There sample consisted of total monthly returns for the Datastream Global Equity Indexes from 1979 to Within the seven largest countries they studied 10 industry sectors. In the research of Brooks and Del Negro (2002), the results show that the industry shocks in the past had a less considerable effect on stock prices than had country shocks. They examined 42 developed and emerging countries and 38 industries listed in the level 4 Datastream Global Equity Industries from 1986 to Campa and Fernandes (2006) investigated the country and industry specific factors in international portfolio returns using data of 48 countries over the last 30 years. In this sample were 19 emerging markets and 27 developed countries and the data that was used was divided over 36 industries. They found that the industry factors have significantly increased from the 1970s till 2000, but went downwards after this year. The country factors, in contrary, stay relatively stable for the last thirty years. In the equity market, Carrieri, Errunza, and Sarkissian (2004) found that in developed markets diversifying over industries has become of main importance, while in emerging markets, diversifying over countries still should be the predominant asset allocation. They used a sample of monthly equity market returns over the period 1976 to 2003 from 17 OECD countries and 10 local industries. Also Cavaglia, Brightman and Aked (2000) agree with this view and provide evidence that from 1995 till 2000 diversification across industries was a better way to reduce risk than diversification by countries. For this research they used a dataset consisting of data of 21 developed equity markets from 1986 to To measure the performance of a portfolio consisting of securities of the same industry in one country they used 36 industry-level national total returns. In the research by Diermeier and Solnik (2001), conducted on security prices, they found that regional factors and currency factors have a large effect on asset returns besides the effects of domestic factors. Also, they found that the effect on companies by non-domestic factors is 3

5 strongly related to the amount of the company s international activities. They measured this importance by comparing foreign sales to total sales. They used a sample of Japan, United States and Europe based firms and their weekly returns from 1989 to Eiling, Gerard and De Roon (2004) state that from 1992 till 1998 the best risk-return ratios were earned, while undertaking a country-based diversification strategy. From 1998, industryfactors and country-factors are equally important in the diversification strategy. They used the Datastream monthly returns on eleven country indices and ten EMU-zone industry indices from 1990 to Ferreira and Gama (2005) indicate in their paper that from the late nineties, the correlations between local industries have declined. So, it becomes more and more important to diversify over industries.the dataset they used constists of daily total return indices and market capitalizations for 38 Datastream industries from 1973 to developed markets were included in their study. Griffin and Karolyi (1997) found evidence that little variation in country index returns can be explained by the industrial composition of that country. They used Dow Jones World Stock Index data over the period December April This Index is composed of 66 industries and 45 sub-industries divided over 25 countries. In the paper of Heston and Rouwenhorst (1994), they explain the correlation structure for 12 European countries between 1978 and 1992 by comparing the industry structures of the countries. They conclude that country-effects are more relevant than industry-effects and so make diversification across countries a far more effective tool for diversification. In his research, Rouwenhorst (1999) stated that no evidence is available for the disappearance of the differences between the equity returns of countries of the European Monetary Union. Baele and Inghelbrecht (2006) found in their research that diversification over industries has become substantially more important, related to the historical belief that diversification over countries is the most effective way to reduce risk. Baele and Inghelbrecht used data, consisting of weekly total return indices and market capitalizations from 1973 to They covered 4 regions, 21 countries and 21 global industries. Other academics found that the growing relative importance of industry effects was caused by only temporary effects. Bekaert et al. (2006) found in their research that the increasing relative importance of industrial effects to country effects was a temporary phenomenon. In earlier years the importance of industrial effects grew to the same level as the importance of country effects, but it seems not to be a structural movement. Bekaert et al. studied the comovements between the weekly returns on country-industry portfolios and country-style portfolios for 26 4

6 industries, 23 developed countries and 9 styles from 1980 to The styles are defined, based on book-to-market ratios (BM-ratio) and the size of the firms. In later research, by using the same data, Brooks and Del Negro (2004) suggest that from the mid-1990s till the stock market bubble in 2003, the most effective way to reduce risk in portfolios is to diversify over industries. After the bubble, the best way to diversify again seems to be over countries. Many researches share the opinion that till the beginning of the nineties, diversification across countries is the best way to reduce risk and to maximize returns. After the beginning of the nineties diversification over industries gained importance till about the year After 2000 the industry factors lost effect and became as important as the country factors. But, some researchers like Ferreira and Gama (2005) state that even after 2000 diversification over industries is the most effective way to reduce risk and maximize returns. Furthermore, Carrieri et al. (2004), found that the common belief that after 2000 industry- and country effects are of the same importance, not holds for emerging countries. They found that for emerging countries, diversifying over countries still is the most effective way for diversification. The remaining of this paper is organised as follows: Chapter 2 will give an overview of the historical development of the relative importance of country- versus industry effects. Chapter 3 describes the structural factors, which are responsible for the development of the relative importance of the country- and industry effects. Chapter 4 gives a explanation of the differences between the sentimental versus fundamental views on comovement of asset returns. Chapter 5 describes the methodology used by Heston and Rouwenhorst (1994) in their paper on the influence of industry and country structure on the correlation and volatility structure for 12 European countries. In Chapter 6 the conclusions of this paper will be presented. 5

7 2 Development of relative importance over time Baca et al. (2000) question whether the relative importance of industry effects during the collapse of the TMT-bubble was a structural or a temporary phenomenon. They found evidence for the fact that during the collapse of the bubble investor should have diversified their portfolio over industries. They computed that in August 1998 the pure sector returns lied between -6,29% and 14,27%. In contrary, the pure country returns lied between -2.24% and 5,80%. This means that in the period of the collapse it was far more efficient to diversify over industries as it was to diversify over countries. Baele and Inghelbrecht (2006) agree with Baca et al. (2000) and found that the convergence of industry and geographical diversification is not a pure result of the TMT bubble. Baele and Inghelbrecht (2006) also found that structural and temporary factors both have effects on market betas. Structural effects cause a strong increase in both regional and global market betas, while temporary factors cause the time variation in these betas. The same results came up for effects in industries. Baele and Inghelbrecht (2006) also found that from 1976 till 2006 the average country-specific risk had been higher than average industry-specific risk. They disagree with a lot of researchers like Bekaert et al. (2006) who say that the difference in relative effects of country- and industryfactors has diminished. As said, Bekaert et al. (2006) disagree with Baele and Inghelbrecht (2006). They examined the trend for stock return correlations within Europe from 1980 till 2003 and they found a significant upward trend for these correlations, but only in Europe. A permanent increase in correlations would be indicated by an increase in betas, but unfortunately, this is not the case. For correlations between industries however, they found that correlations have decreased during the nineties, but that after 2000 this trend has been stopped and even reversed. Bekaert et al. (2006) thus found no evidence for a trend over the years. The third thing they examined was the correlation between stocks of the same style across different countries. Bekaert et al. (2006) found a pattern that stocks with a small value are less correlated than stocks with a high value. They also found that this difference has increased over time. The last thing Bekaert et al. (2006) found is the absence of a trend in changes of firm-level idiosyncratic variances over the time they examined. Eiling, Gerard and De Roon (2004) agree with the preceding academics on the development of the relative importance of country versus industry effects. They found that until the introduction of the Euro country-based strategies outperform industry-based strategies. After the introduction of the Euro this outperformance disappeared and nowadays industry-based and country-based strategies are equally effective. One of the reasons they mention is that after the introduction of the Euro, currency risk between Euro-countries has disappeared. De Santis and Gerard (1998) indicate that currency risk premiums are substantial and 6

8 economically significant. When different exchange rates exist between countries, this leads to different currency risk premiums. As a result there will be more market segmentation and lower cross-country correlations. On the other hand, when currency risk is eliminated, this would lead to higher cross-country correlations. Adjaouté and Danthine (2001), and De Santis, Gerard and Hillion (2003) support this view. They found a significant increase in correlations between returns on Euro-countries equity indices after the Euro was introduced. But, when they adjusted the results for currency effects they find the same increase in correlations. This suggests that the elimination of currency effects after the introduction of the Euro, is not the main reason for the increase of the cross-country correlations. So, the adoption of a single currency for a region has only a small impact on the diversification strategies conducted by investors. Despite the expectation that the introduction will have a limited effect on crosscountry correlations, Eiling et al. (2004) found a significant increase in cross-country correlations in the Euro-zone. In 1998, Goldman and Sachs and Watson and Wyatt conducted a survey, according to which 60% of the fund managers switched their portfolio strategy from country to industry based. There are several other papers addressing this issue. Ferreira and Ferreira (2003) found that country factors still dominate in the Euro-zone, although the importance of industry effects has increased. Rouwenhorst (1999) and Ehling and Ramos (2004) agree with Ferreira and Ferreira (2003) and found evidence for the dominance of country factors in the Euro-zone. Adjaouté and Danthine (2002) conversely, found that industry factors have become more important than country factors. Eiling et al. (2004) conclude that investing in country-based portfolios is no longer needed to get an optimal risk-revenue portfolio. They also conclude that differences in performance of industry- and country-based portfolios are not affected by changes in market volatility. When Eiling et al. (2004) control for the TMT-bubble, their conclusions still hold. Risk Ferreira and Gama s (2005) motivation for their study is that investors portfolios are exposed to a high total risk. This statement is motivated by the fact that investors do not hold the market portfolio of the countries they invest in. When portfolios are hardly diversified the diversifiable risk is high. To reach a particular level of diversification the investor needs a lot of assets. Therefore the benefits of investing in foreign countries become harder to achieve, as well as the compensation for undertaking such a strategy. Ferreira and Gama (2005) also state that when investors have a maximum amount of assets available and/or encounter transaction costs, an increase in diversifiable risk means a decrease in the level of diversification of their portfolios. 7

9 In their paper, Ferreira and Gama (2005) distinguish three types of risk that are important for international tradable equity. They split these two risks in two parts: systematic risk and nonsystematic risk. Systematic risk is the risk of holding the market portfolio. As the market moves, each individual asset or portfolio is more or less affected by these movements. Nonsystematic, unsystematic, idiosyncratic or specific risk refers to the risk of a price change due to the specific circumstances of a specific asset. Non-systematic risk is diversifiable to a certain extent in a global portfolio. Ferreira and Gama (2005) take two sources, which are: geographic location and industry affiliation. Ferreira and Gama (2005) conclude from their research that the world risk, so the systematic risk, has been the least important source of total risk from 1974 to They found no significant trend in the volatility of systematic risk. In contrast, they found significant results for both industry and country volatility. Industry volatility shows a steep increase from 1995 till Ferreira and Gama (2005) indicate that the internet-bubble by itself does not explain the increase in industry risk, but that the TMT industry plays a weighty role in this development. The economical crash of October 1987 influenced both world and country risks, but had a smaller effect on local industry risk. Actually, since the 1987 crash, local industry volatility started to increase until now. The period from 1990 till 1995 can be characterized as an abnormal period in history, since the share of country risk in total risk is atypically high and the total risk itself is averagely lower than in the neighbouring years. In the period the benefits of international portfolio diversification have increased. Besides that, it became harder to achieve global diversification using local industry portfolios, because more assets are needed to gain that result. The increase in the relative importance of industry over country risk in the second part of the 1990 s indicates that diversification over industries became a more important tool for risk reduction. 8

10 3 Structural factors Campa and Fernandes (2006) found that the importance of country factors is higher for countries with a lower gross national product. When a country is more international financially integrated the importance of country factors decreases. When industries in a country have a high degree of product specialization, country factors are also more important than in a country s industry with a low degree of product specialization. Also, when the financial market is more active in one country, this increases the importance of countryspecific factors. Baele and Inghelbrecht (2006) did not focus on the industry characteristics but focused on the variation of market betas. They state that the relative influence of country and industry factors vary over time. A large body of research indicate that these fluctuations are extremely countercyclical. During growing markets the correlation between international equity returns tend to be higher when prices of these securities are falling and lower when prices of these securities are rising. Baele and Inghelbrecht (2006) found that the influence of temporary and structural effects is particularly influenced by fluctuations of market betas. There are several researches that indicate this relation. Among others, Bekaert and Harvey (1997), Chen and Zhang (1997), and Baele (2005) state that further integration of economies and financial markets lead to a structural increase in the countries regional and global market betas and therefore higher cross-market correlations. They also found evidence for fluctuations in market betas even when structural shifts in the economy are absent. Several papers indicate that betas are a function of variables of a country s economic state. Santos and Veronesi (2004) found that not only the level of the betas but also the dispersion of the betas fluctuate over time. Baele and Inghelbrecht (2006) show that these time-varying dispersion in betas is critical for describing cross-industry and cross-country risk and correlations over time. At last, the change of industry characteristics, caused by competition or changes in technology can cause shifts in industry betas. Brooks and Del Negro (2004) show that during the nineties the correlation between U.S. stock returns and other developed countries stock returns has risen from 0.4 to 0.9. They give several possible explanations for this rise. One reason can be that investors during this period tend to be less chauvinistic in their investment decisions. Because of this development country-specific sentiments play a smaller role in portfolio decisions and investors became global portfolio holders. A second reason can be that firms nowadays attract funds more globally and market their products all over the world. Result of this is that firms are exposed more and more to the global market fluctuations and that national equity markets are comoving more and more in the same direction. The third reason Brooks and Del Negro 9

11 (2004) mention is that it is possible that the increase in comovement in the nineties simply is a result from the stock market bubble of Brooks and Del Negro (2004) investigated the relative importance of industry effects outside the TMT-sector (Technology, Media and Telecommunications) because they were suspecting that the increasing importance of industry effects were mainly caused by stocks inside the TMT-sector. They found no consistent rise in the relative importance outside the TMT-sector. Brooks and Del Negro (2004) found that the relative importance of industry effect increases when the economy is in a distress period. They also found that country-specific factors has a larger influence on the variation of stockreturns in industries in which non-traded goods are sold than in traded goods industries. Diermeier and Solnik (2001) state that in practice, portfolio managers use one basic assumption, regarding country and industry effects in constructing their portfolio. This assumption is that country factors are the biggest source of influence for variation in stockprices. And that these country factors are hardly correlated. By this, they mean that the stockprice of, for instance, BMW behaves more like another German company like Henkel, than it behaves like a car-manufacturer from another country like Renault. Diermeier and Solnik (2001) also state that the stock price of a company is influenced by all the country factors the company is active in. To which extent the stock price is influenced can be seen in proportion to the value of each country s component in the company s activities. Another conclusion of Diermeier and Solnik (2001) is that the return of an individual company is strongly related to what degree a company develops international activities. Griffin and Karolyi (1998) also focus on international diversification. In their paper, they give several reasons how investors can gain from international diversification. Some analysts maintain the opinion that the gains from international diversification come from differences in economic conditions underlying foreign capital markets which are caused by differences in monetary and fiscal policies, changes in interest rates, budget deficits and national growth rates. Another group of analysts thinks that the gains of international diversification comes from the different industrial structures across countries. Roll (1992) suggests in his research that there are three forces causing variation in countries index portfolio returns. These forces are first the technical procedures of index construction, second, the industrial composition of the index, and third, the behaviour of both real and nominal exchange rate behaviour. He also argues that the industrial composition is very important for explaining cross-sectional variation in volatility and the correlation structure of country index returns. When an investor invests in the Swiss market portfolio this portfolio represents a disproportionate amount of banking stocks, while investing in the Dutch market portfolio means holding a large share of energy sector stocks. 10

12 Griffin and Karolyi (1998) re-examine Heston and Rouwenhorst s (1994) results by employing the new Dow Jones World Stock Index database. This index contains of daily index prices for 66 industry classifications and more than 25 countries. Griffin and Karolyi (1998) examined data over the period from January 1992 to April Results of this analysis agree upon Heston and Rouwenhorst s (1994) finding that little of the variation in country indexes can be explained by the industrial structure of that country. Heston and Rouwenhorst (1994) state that part of the benefit of international diversification actually comes from industrial diversification. They reason that, because industries are imperfectly correlated, equity markets with a difference in industry structure should be imperfectly correlated as well. These industrial differences also explain the difference in volatility of some markets: when, on average, banking stocks are less volatile than energy stocks, one could argue that the Dutch index could be more volatile than the Swiss index. Another explanation Heston and Rouwenhorst (1994) give for the low international return correlations is that, differences in legislation, local fiscal and monetary policies and regional economic shocks cause large country-specific variation in returns. The last explanation assumes that country-specific factors are substantial enough to dominate any industry effects. When one takes this view, the imperfect correlation between equity markets of Switzerland and The Netherlands is caused by specific, independent shocks and not by the fact that Switzerland has more banks. Griffin and Karolyi (1998) found differences in relative importance of country factors among different industries. They divide industries in two groups: industries that do not produce goods that are traded internationally and industries that do produce goods that are traded internationally. For nontraded-goods industries such as heavy construction and media, country factors have a relatively bigger impact on the variation of index returns. For tradedgoods industries such as computers and automobiles, on the other hand, industry factors have a relatively bigger impact on the variation of index returns. They also found information about covariances in traded-goods industries. Cross-country covariances for firms within a particular industry are higher than cross-country covariances for firms in different industries. Kallberg and Pasquariello (2006) confirm this finding in their research. In their paper they investigate excess comovement at the industry level. They used the concepts of information, momentum, liquidity and product market shocks to explain these comovements and found that for traded-goods industries like Basic Industries and Cyclical Consumer Goods the explanatory power is much smaller than for nontraded-goods industries like Information Technology and Financials. When we link this to the differences between developed countries like the G6-countries (US, Japan, UK, France, Germany and Canada) and emerging markets Griffin and Karloyi (1998) found that the limits for pure country diversification are higher for diversification within only 11

13 the G6-countries than for diversification within G6-countries plus emerging countries. Contrary, for the nontraded-goods industries there is hardly any difference in cross-country covariances between stocks in the same industry and stocks that are in different industries. 3.1 Integration of financial markets One of the most mentioned factors influencing the development of the relative importance of country versus industry effects is the integration of financial markets. Baca et al. (2000) state that the relative importance of country and industry effects in global stock returns is mainly caused by the degree to which capital markets and global economies are integrated. Academic research before 2000 has not supported the idea that global integration is increasing but Baca et al. (2000) found that it actually has increased. Although there are differences among countries. In Japan capital markets exhibit the most variation in returns within the country and the UK and the US markets exhibit the least variation in returns. This means that the capital market in Japan is the most segmented of the countries in this study. Campa and Fernandes (2006) also found that the most important issue driving the relative importance of country- and industry effects is the integration of the specific country in the world financial markets. The more a country is integrated in world financial markets, the more important industry-specific factors will become. When Campa and Fernandes (2006) look at the industry-level, they found that a higher international financial integration has a positive influence at the importance of industry factors. When industrial activity is concentrated in a few countries this has a negative impact on the importance of industry factors. Kearney and Lucey (2004) elaborate some more about the integration of financial markets. They distinguish 3 approaches to define the extent of international financial integration and divide these in two categories: direct and indirect measures. A direct measure is the extent to which the return rates on financial assets with the same maturity and risk characteristics are equal across political jurisdictions. The second approach is an indirect approach and is the extent to which there exists a complete set of financial markets that allows investors to diversify their portfolio to insure themselves against the full range of risks that can occur. The second indirect approach is the extent to which domestic investment is originating from global savings rather than from domestic savings. Kearney and Lucey (2004) see three sets of implications of this integration. First of all, the effectiveness of international portfolio diversification will weaken as returns are more equal across countries. Secondly, the more complete the world s financial markets, the more powerful the economies of the individual 12

14 countries will be. Third, household savings rates will change over time with the further integration of financial markets. Baele and Inghelbrecht (2006) think the effect of integration of the markets on industry risk and industry correlations is less obvious. On the one side, globalization would lead to an increasingly important effect of global factors on the pricing of assets. Conversely, while the world is integrating more and more, investors can get more access to information about the markets and thus can take into account more specific industry factors in the pricing of assets. The increasing importance of global factors would lead to increasing cross-industry correlations, while the increasing information would lead to the opposite effect. 13

15 4 Comovement: Sentimental versus Fundamental view Barberis, Shleifer and Wurgler (2005) made a distinction between two views of return comovement. In the traditional view, the returns of two assets are correlated because frictions in the fundamental values of these assets are correlated. Frictions in the value of the asset are caused by news about for instance an asset s earnings or about the particular market as a whole. Also stocks and bonds with the same time to maturity and rating tend to be correlated. The traditional theory is based upon an economy with no sudden changes and with rational investors. This means that the comovements in the price of different assets solely are reflected by comovements in fundamental values. This means that every change in price will be caused by changes in fundamental values of the market or of the assets. Unlimited arbitrage in this economy causes that the comovement of prices of assets only reflect the comovement of fundamental values. In a more realistic economy however, frictions occur and investors are not that rational as in the economy described above. That is why Barberis et al. (2005) delink the comovement in prices of assets from comovements in fundamental values and distinguish another view, which is called the sentimental-based or friction-based theory. In these sentimental and friction-based theories Barberis et al. (2005) examine three views of comovement. The first view is called the category view, which is examined by Barberis and Schleifer (2003). They reason that many investors do not allocate funds at the individual asset level but group assets into categories for reasons of simplicity. Grouping determinants are for instance the size of the asset, the kind of industry or the type of bond. When these investors base their allocation of funds on sentiments and their sentiments are correlated they create correlated asset price changes for assets that are grouped in the same category, even if the case flows of these assets are uncorrelated. The second kind of comovement is called the habitat view. According to this view many investors only invest in a small subset of all available assets due to their habits. These habits can be caused by many reasons, such as international trading restrictions, transaction costs or lack of information. International trading restrictions can force investors to invest only in some specific countries or regions. For some investors it can be costly to get access to particular assets, so these assets will not be included in their habitat subset. When a change occurs in an investor s sentiment he increases or decreases his exposure to the assets in his habitat subset and with that create a correlated price change for assets that are traded by a specific group of investors. In the habitat view, this likely will be individual investors. 14

16 Third, Barberis et al. (2005) examine the information diffusion view. The main issue in this view is that, as a result of market friction, information is incorporated more quickly into the prices of some assets than into other assets. Some reasons behind this quicker incorporation in asset prices are the cost of trade, accessibility of information and the accessibility of resources. Some assets are more costly to trade than others and therefore the information will be incorporated later into the asset s price. When certain investors can get quicker access to breaking news and have the resources to quickly respond to that news, this information will be quicker incorporated in the price of assets traded. In the information diffusion view, assets that incorporate information into their prices at the same rate have a correlated return. Engsted and Tanggaard (2007) also investigated the influence of information on the comovement of stocks in different countries. They analyzed US and German bond returns and decomposed these returns into three news elements: news about future inflation, news about future real interest rates and news about future excess bond returns. Sutton (2000) found that comovement in long-term bond yields is primarly caused by news about future excess bond returns. Barr and Priestley (2004) found that 70% of the variation in expected bond returns is due to world risk and 30% can be explained by local country-specific risk. Driessen, Melenberg, and Nijman (2003), on the other hand, found that the positive correlation of bond returns is mainly caused by the positive correlation of the levels of the term structures between countries. Engsted and Tanggaard (2007) found that the most important factor, driving the comovement of bond return fluctuations is the news about future inflation. 15

17 5 Method Description Heston and Rouwenhorst (1994) examined the influence of industrial structure on the crosssectional correlation and volatility structure of country index returns. They used data about 12 European Countries between 1978 and In this part the method they used to examine these relationships will be described. Data Heston and Rouwenhorst (1994) used data for the 829 firms that make up the Morgan Stanley Capital International indices of twelve European Countries. These indices have a flexible composition because of bankruptcies, mergers and delistings. When a firm was dropped out of the index but remained listed on the local stock exchange, Heston and Rouwenhorst (1994) kept this firm in their sample as long as information on the firm s returns was available. They made a second adjustment for the data. Only shares that can be held by international investors are included in the sample. Then they divided the sample in seven categories, based on industry classifications from the Financial Times Actuaries/Goldman Sachs: Finance, Insurance and Real Estate Energy Utilities Transportation and Storage Consumer Goods and Services Capital Goods Basic Industries Most firms are assigned to the four categories: finance, basic industries, consumer goods and services and capital goods. In some countries there are no firms in some categories so these were excluded from the sample. Heston and Rouwenhorst (1994) found a variation of returns and volatility across markets and industries. They also found a low correlation among country index returns and that is the reason they attempt to answer two questions: first, is there a relationship between the relative performance of countries and their industrial structure? And second, is industrial diversification the primary factor behind the benefits of international diversification? Method 16

18 To answer the preceding questions they used the method described below. They measure the return on the ith security that belongs to industry j and country k in order to separate country performance from industry performance using the following model: Rit = α + β + γ + ε, (a) t jt kt it Where α t is the base level of the return in period t, β jt is the industry effect, γ kt the country effect, and ε it is the error term and represents the firm-specific disturbance. Model (a) allows segregated influences of country and industry effects, but eliminates interaction effects between these effects. The error terms are assumed to have a mean of zero and finite variance for returns in all industries and countries, and are not correlated across firms. Heston and Rouwenhorst (1994) defined dummies for both industries and countries. I ij is equal to one if security i belongs to industry j and zero if not. C ik is equal to one if security i belongs to country k and equal to zero if not. For each period equation (a), after plugging in the dummies, can be rewritten as R i = α + β γ C + ε i (b) 1 I i1 β 2I i2 β 7 I i7 γ 1Ci1 γ 2Ci i12. They want to estimate (b) by cross-sectional regression, but this is not directly possible because of perfect multicollinearity between the regressors. They encountered a problem, because there is no possibility to identify country and industry effects. Because every firm is in one industry and in one country, they can only identify cross-sectional differences between countries. On the other hand, it is possible to measure country and industry effects relative to a benchmark. Heston and Rouwenhorst chose to take the average firm in the sample as their benchmark. To do this, they imposed the following restrictions for each time period t: 7 j= 1 12 k= 1 n β = 0, and (c1) j j m γ = 0. (c2) k k n j and m k represent the number of assets in industry j and country k. In normal least-squares estimation of a regression the estimated residuals are zero for every country and every industry. When you add all these residuals over all industries and all countries you get the estimated residual for the aggregate European index. Because you add 17

19 all zeros the estimated residual for the aggregate European index is zero. While this is zero, the least-squares estimate for α from equation (a) is equal to the estimate of the return of the equally-weighted European market. This means that the geographically-diversified portfolio is equal to the European equally-weighted index and therefore has no country-specific effects. The same reasoning holds for the industrially-diversified portfolio which is free of industryspecific effects. Therefore Heston and Rouwenhorst introduced a decomposition method ew R k, to split the actual equally-weighted index of a specific country, k into a part that is the same for all countries, α, which is the average of the industry effects of the securities in the index, and a country-specific part, γ : k R ew k 7 1 = α + β j + β j I ij + γ k. (d) m k i j= 1 For the industry index, the same procedure can be done. Each equally-weighted industry index return, ew R j, can be split into a part that is the same for all industries, α, which is the weighted average of the country effects of the securities in the index, and an industry-specific part, β : j R ew j 12 1 = α + β j + γ k Cik. (e) n j i k = 1 Now, you can compose new restrictions that imply that the value-weighted European index has neither industry nor country effects: 7 j= 1 w β = 0 (f1) j j 12 k= 1 v γ = 0. (f2) k k 18

20 In equations (f1) and (f2) = k k w j j country k and v = 1. w j and v k are the value weights of specific industry j and specific Now, the weighted least-squares estimate of the regression intercept becomes equal to the European value-weighted index. After composing the model described above, Heston and Rouwenhorst (1994) did a cross-sectional regression for every month and obtained a time series of industrially-diversified country returns and geographically-diversified industry returns. They used these returns to gain insight into the source of variation in the industry and country index returns. 19

21 6 Conclusion Almost all literature agrees on the historical development of the relative importance of country versus industry effects. For a large period of time international diversification was superior over diversification over industries to reduce portfolio risk. During this period capital markets in specific countries were fairly autonomous capital markets with not that much influence from foreign markets. Diversification benefits could be achieved by holding a portfolio with assets from different countries. During the nineties capital markets constantly became more integrated and the relative importance of international diversification was diminishing. Diversification over industries became more and more important and now more or less reached the same importance level as diversification over countries. The increase of relative importance of industry-diversification during the period of the collapse of the TMT-bubble was according to some researchers just a temporary phenomenon, while others title this as a structural shift to industry-diversification as the optimal diversification method. Future developments will make clear if the importance of industryfactors will continue to rise or that it will just be a temporary trend. Structural factors that have a significant impact on the development of the importance of country and industry factors are: the value of the stock (Bekaert et al., 2006), the adoption of a single currency (Adjaouté and Danthine, 2001 and De Santis et al., 2003), the level of the gross national product, the activities of one country s financial market, the degree of product specialization and (Campa and Fernandes, 2006), the state of the economy (Baele and Inghelbrecht, 2006 and Brooks and Del Negro, 2004), the change of industry characteristics (Baele and Inghelbrecht, 2006), chauvinism of investors and more international fund attraction and product marketing (Brooks and Del Negro, 2004), the production of traded- or nontraded goods (Brooks and Del Negro, 2004, Griffin and Karolyi, 1998 and Kallberg and Pasquariello, 2006), and the exchange rates and the industrial composition (Roll, 1992). Stocks with a higher value are more correlated than stocks with a lower value within a given industry across countries. This means that higher stock prices will lead to an increase of the importance of diversification over industries. Adjaouté and Danthine (2001) and De Santis et al. (2003) found a significant increase of between country correlation but after controlling for currency effects they found the same increase. This means that the adoption of a single currency has only a small impact on cross-country correlations. When a country has a high gross national product, high activity of the financial market and a high degree of product specialization the importance of country factors increases. When the economy is growing, the correlation between international equity returns tend to be higher when prices of these 20

22 securities are falling and lower when prices of these securities are rising. In a distress period, the relative importance of industry factors increases. When investors are more chauvenistic, this leads to less correlation between different countries correlations. Because companies attract funds and market their products more internationally this also leads to more correlation between different countries. Both these developments ask for more diversification over industries. When industries are producing more traded goods, country-specific factors have less influence on the variation in stock returns. The main factor directing the relative importance of country versus industry effects is the international integration of financial markets. Baca et al. (2000) found that financial markets actually have integrated during the recent years unless there are differences between countries. Campa and Fernandes (2006) state that the more a country is integrated in world financial markets the more important industry factors will be. Baele and Inghelbrecht (2006) think the development of the relative importance is less obvious. The signal two opposite movements: first, increasing integration leads to an increase in global factors. Contrary, increasing integration leads to a more transparant market in which investors can take more industry factors into account while pricing assets. 21

23 Reference list - Adjaouté, K., J.P. Danthine (2001), EMU and Portfolio Diversification Opportunities, CEPR discussion paper, No Adjaouté, K., J.P. Danthine (2002), European Financial Integration and Equity Returns: A Theory-based Assessment, Conference paper, Second ECB Central Banking Conference - Baca, Sean P., L. Garbe, Brian, and A. Weiss, Richard, 2000, The rise of sector effects in major equity markets, Financial Analyst Journal 56, Baele, Lieven, 2005, Volatility spillover effects in European equity markets, Journal of Financial and Quantitative Analysis 40, Baele, Lieven and Inghelbrecht, Koen, 2006, Structural versus Temporary Drivers of Country and Industry Risk, Working Paper, Universiteit Gent. - Barberis, Nicholas, Shleifer, Andrei, Style investing. Journal of Financial Economics 68, Barberis, Nicholas, Shleifer, Andrei and Wurgler, Jeffrey, 2004, Comovement, Journal of Financial Economics 75, Barr, D. G. and Priestley, R., 2004, Expected returns, risk and the integration of international bond markets. Journal of International Money and Finance 23, Bekaert, Geert, and Campbell R. Harvey, 1997, Emerging equity market volatility, Journal of Financial Economics 43, Bekaert, Geert, Hodrick, Robert and Xiaoyan Zhang, 2006, International Stock Return Comovements, Working Paper. - Brooks, Robin, and Marco Del Negro, 2002, International Diversification Strategies, Working Paper, Federal Reserve Bank of Atlanta. - Brooks, Robin, and Marco Del Negro, 2004, The rise in comovement across national stock markets: Market integration or IT bubble?, Journal of Empirical Finance 11, Campa, Jose Manuel, and Nuno Fernandez, 2006, Sources of gains from international portfolio diversification, Journal of Emperical Finance 13, Carrieri, Francesca, Vihang Errunza, and Sergei Sarkissian, 2004b, The dynamics of geographic versus sectoral diversification: Is there a link to the real economy?, Working Paper McGill Univeristy, Cavaglia, Stefano, David Cho, and Brian Singer, 2001, Risks of sector rotation strategies, Journal of Portfolio Management 27,

24 - Chen, Nai-Fu, and Feng Zhang, 1997, Correlations, trades and stock returns of the Pacific-Basin markets, Pacific-Basin Finance Journal 7, De Santis, G., B. Gerard (1998), How Big is the Premium for Currency Risk?, Journal of Financial Economics 49, De Santis, G., B. Gerard, P. Hillion (2003), The Relevance of Currency Risk in the EMU, Journal of Economics and Business 55, Diermeier, Jeff, and Bruno Solnik, 2001, Global pricing of equity, Financial Analyst Journal 57, Driessen, J., Melenberg, B., and Nijman, T., 2003, Common factors in international bond returns. Journal of International Money and Finance 22, Ehling, P., S.B. Ramos (2004), Geographic versus Industry Diversification: Constraints Matter, Working paper, No 192 NCCR FINRISK. - Eiling, Esther, Bruno Gerard, and Frans de Roon, 2004, Asset allocation in the eurozone: Industry or country based?, Working Paper Engsted, Tom and Tanggaard, Carsten, 2007, The comovement of US and German bond markets, International Review of Financial Analysis 16, Ferreira, M.A., M.A. Ferreira, The Importance of Industry and Country Effects in the EMU Equity Markets, Working paper, ISCTE School of Business. - Ferreira, Miguel A., and Paulo M. Gama, 2005, Have world, country and industry risks changed over time? An investigation of the developed stock markets volatility, Journal of Financial and Quantitative Analysis 40, Griffin, John M., and René M. Stulz, 2001, International competition and exchange rate shocks: A cross-country industry analysis of stock returns, Review of Financial Studies 14, Heston, Steven L., and K. Geert Rouwenhorst, 1994, Does industrial structure explain the benefits of international diversification?, Journal of Financial Economics 36, Kallberg, Jarl and Pasquariello, Paolo, 2006, Time Series and Cross-Sectional Excess Comovement in Stock Indexes, Working Paper, University Of Michigan Business School. - Kearney, Colm and Lucey, Brian M., 2004, International equity market integration: Theory, evidence and implications, International Review of Financial Analysis 13, Roll, R., Industrial structure and the comparative behavior of international stock market indices. Journal of Finance 47, Rouwenhorst, Geert, 1999, European equity markets and the EMU, Financial Analysts Journal 55,

25 - Santos, Tano, and Pietro Veronesi, 2004, Conditional betas, NBER Working Paper 10413, Sutton, G. D., 2000, Is there excess comovement of bond yields between countries?, Journal of International Money and Finance 19,