The role of financial intermediaries in the international sharing of risk

Transcription

1 TILBURG UNIVERSITY The role of financial intermediaries in the international sharing of risk BSc Thesis Economics P.J.M. de Kort ANR: Supervisor: Prof. dr. W. Wagner Number of words: 6925

2 1 Introduction During the last decades, most barriers to international asset trade have been removed. Capital controls have been reduced or completely eliminated, information and communication technologies have improved significantly and much regulation has been harmonized globally. These developments have been studied extensively in various fields of economics. The research conducted in this thesis is from a financial economics perspective and is directed towards the international sharing of risk via equity portfolio diversification. Both theoretically and empirically confirmed the potential benefits of international equity portfolio diversification are considerable. For example, Dimson, Marsh and Staunton (2002:120) estimate a risk reduction of 10 to 20 percent for an investor holding a fully diversified world portfolio. Assuming investors are risk averse, holding a more internationally diversified portfolio would clearly increase utility. Next to that, as explained above, the costs do not appear to be large anymore. Still, data on foreign asset holdings suggest that investors exhibit home bias rather than sharing risks fully internationally. Interesting questions that arise from this remarkable observation include why is there no more diversification? And, what determines the actual level of international diversification? The research conducted in this thesis is most related towards the latter question. The main focus of this thesis is on explaining the effect of financial intermediaries on international equity portfolio diversification. In doing this, the thesis is aimed at filling the gap between the literature on international banking and the literature on equity home bias by using the economic theory of financial intermediation. According to economic theory, financial intermediaries are established, among others, to overcome information asymmetries, reduce transaction costs and share and diversify risks. Domestically, financial intermediaries share risks by diversifying asset holdings across sectors, regions and generations. Next to that, they also spread risks internationally. The large existing literature that analyzes the diversification of bank portfolios shows that banks do not fully diversify their portfolios to maximally share risks internationally (see for example, Buch, Driscoll and Ostergaard, 2005 and Schoenmaker and Wagner, 2011). Also on equity home bias a vast literature has emerged after the influential papers of French and Poterba (1991) and Tesar and Werner (1995). Although both strands of literature give multiple different reasons for the lack of international portfolio diversification, the role of financial intermediaries in international diversification is not explicitly analyzed. Barron and Ni (2008) 1

3 address a potential role indirectly in their study on the role of asymmetric information in explaining home bias. However, their work is mainly theoretical and their predictions are only tested briefly on a sample of U.S. pension funds. Thus, the research question that this thesis answers is the following: What is the effect of financial intermediaries on international diversification of equity portfolio holdings? In order to answer this question completely, it is broken down into two sub questions: What is the overall effect of financial intermediaries on diversification? And, what is the individual effect of several different types of financial intermediaries on international equity portfolio diversification? Throughout this thesis, the terms financial intermediaries and financial institutions are used interchangeably to denote banks, mutual funds, insurance companies and pension funds. In assessing the effect of financial intermediaries, only their size and relative importance is taken into account. t studied are other characteristics of the financial system such as competition, ownership and efficiency. As is already indicated several times, only equity portfolio holdings are analyzed. Of course, also other securities and derivatives offer significant diversification opportunities but to limit the length of the thesis, these are not studied. The main method that is used to answer the research question and sub questions stated above is regression analysis. First, the overall effect of financial intermediaries is examined. Afterwards, different types of financial institutions are entered separately into the regressions in order to determine their individual effect. To measure equity portfolio investment, data from the IMF s Coordinated Investment Portfolio Survey (CIPS) will be used for a sample of 67 countries covering the period Other data sources are the IMF s International Financial Statistics and the ECB for data on the size of banks, the OECD Institutional Investors statistics for data on the size of other financial intermediaries and the World Bank indicators. The rest of this thesis is structured as follows. In the next chapter, the applicable theories are discussed and the existing literature on the topic is reviewed. In chapter 3 the hypotheses are developed, followed in chapter 4 by a description of the data used. Chapter 5 presents the results and findings and in the last chapter, chapter 6, conclusions are drawn. 2

4 2 Theory and literature review This chapter is structured as follows. First, an overview of applicable theories is presented in section 2.1. In sections 2.2 and 2.3 these theories are explained in more detail and related to some of the existing literature on international diversification of financial intermediaries portfolios and equity home bias respectively. Section 2.4 concludes. 2.1 Applicable theories The most important theory for this thesis is the theory of financial intermediation. The theory of financial intermediation can best be described as a set of theories that explain the existence and functions of financial institutions in the economy. Since there is a very large and broad literature on this theory, only the elements that are most relevant for to the research in this thesis are discussed in this chapter. The theory of financial intermediation is the most important theory used in this thesis since it is at the basis of the research question. The theory explains the reasons why financial institutions could have an effect on international equity portfolio diversification. It also gives a prediction about their expected effect. The other theories that are relevant for this thesis are the capital asset pricing model (henceforth CAPM) and its international extension, the international asset pricing model or international CAPM. The CAPM is relevant since it clarifies the rationale behind and the benefits of (international) equity portfolio diversification. Additionally, the theory on portfolio selection determines the measure of diversification. It provides the theory behind the full-diversification benchmark against which the level of diversification is compared. 2.2 Theory of financial intermediation The theory on financial intermediation asserts that financial institutions are established mainly to overcome informational asymmetries or at least reduce the associated cost. In general, it is costly for to acquire information about potential investments (Levine, 1997: 694). Financial institutions can, because of economies of scale and scope obtain information at lower costs per unit invested than individuals (Levine, 1997: 695). Furthermore, over time they develop more expertise in evaluating investment projects (Mishkin, 2004: 178). For both reasons, financial institutions can reduce the ex-ante adverse selection problem inherent in financial 3

5 transactions. Also ex-post financial intermediaries can have a role in the reduction of moral hazard problems. The main reason for this is again the large scale of financial intermediaries which allows them to monitor more closely an investment (Levine, 1997: 697). Some economists argue that financial intermediaries only have sufficient incentives to gather information and monitor a firm if the investment in it is non-tradable (see, for example Mishkin, 2004). Otherwise, the information is incorporated in the price of the security quickly (assuming liquid and efficient financial markets) and the financial institution that gathers the information only pays the costs of it while not being able to appropriate the benefits. The incentive for free riding in financial markets is indeed a widely recognized problem reducing the production of information. However, financial intermediaries have a lower incentive to free ride than individual investors since they have much larger stakes in companies. Therefore, they should be expected to have a larger incentive to acquire information. For international equity investment Barron and Ni (2008) have formalized this argument in a model and show that there is indeed a direct positive association between portfolio size and information acquisition. Next to using the information gathered about potential investments for resource allocation purposes, financial intermediaries can also sell information to other investors (Mishkin, 2004: 176). They can do this for example to lend credibility to information provided by a firm or to make it easier for individual investors to digest information. Besides reducing transaction costs due to informational asymmetries, financial intermediaries can also reduce other (more or less fixed) transaction costs associated with investing in financial markets (Mishkin, 2004: 173). They can do this, since they often trade in large blocks and thus make use of economies of scale. Furthermore, financial intermediaries, in particular banks, can offer investors the tools to invest with, for example by acting as a broker or custodian or by developing investment products. According to the theory of financial intermediation, the last function of financial intermediaries that is relevant for this thesis is diversification or risk sharing (Levine, 1997). The large size of financial institutions allows them to allocate capital to many different investments. Provided that the returns on these investments are less than perfectly correlated, some of the risk of the investments is diversified away. Domestically, financial institutions share risks by diversifying asset holdings across sectors, regions and generations. Next to that, they can also share risks internationally. On this issue, most existing literature shows that 4

6 financial intermediaries themselves do not fully diversify their portfolios to maximally share risks internationally. For example, Buch, Driscoll and Ostergaard (2005) find that banks are over-invested in the domestic economy relative to the optimal benchmark which is computed under different assumptions of exchange rate risk hedging. Shoenmaker and Wagner (2011) find mixed international diversification results for European banks in individual countries and an overall over-exposure of European banks to the US. In the last study on banking portfolio diversification that will be considered here, García-Herrero and Vasquez (2007: 16) report that on average the banks in their sample hold 82.4 per cent of their assets at home. From a portfolio theory approach, an average of 60.1 per cent of assets held at home would be optimal. Next to research analyzing international diversification in banking portfolios, some research also has been done on international diversification of mutual fund equity holdings. An interesting study on this topic has been conducted by Ni (2009). He finds that, in general, mutual fund managers exhibit home bias but also that managers of larger mutual funds hold more diversified portfolios. 2.3 CAPM and international asset pricing model The standard, domestically oriented CAPM asserts that the market portfolio coincides with the efficient portfolio (Sharpe, 1965). Following the mean-variance approach introduced by Markowitz (1959), the efficient portfolio is a portfolio of securities that minimizes volatility (measured by standard deviation) for a given rate of return or maximizes the rate of return for a given volatility (Markowitz, 1959: 22). According the CAPM, the market portfolio must be such an efficient portfolio since otherwise there would be no stock market equilibrium (Sharpe, 1964). However, for the CAPM to hold some simplifying assumptions are needed. The most important ones are that financial markets are competitive, borrowing and lending at the risk-free interest rate is possible for all investors and expectations are homogeneous (Sharpe, 1964: 433). Although in reality clearly not all these assumptions hold, the CAPM is a useful starting point widely used by economists. The original CAPM explained above fully focuses on domestic diversification. International extensions of this model have been made by Levy and Sarnat (1970) and Solnik (1973). The efficient international portfolio can be defined, analogous to the definition of Markowitz (1959), as the portfolio that maximizes rate of return for a given variance or minimizes variance for a given rate of return (Levy and Sarnat, 1970). As demonstrated by Solnik (1973: 39) the world marked-weighted portfolio coincides with this efficient international portfolio. 5

7 This result holds under the same assumptions as for the general CAPM and additionally under the assumptions that there are no constraints to international capital flows and investors only consume in their home countries (Solnik, 1973). The theoretical discussion above thus predicts that investors hold securities in proportion to the world market portfolio. Nevertheless, past research has shown that, in practice, investors exhibit a significant degree of home bias. They tend to overinvest in domestic stocks and underinvest in foreign ones. In one of the earliest papers on this issue, French and Poterba (1991) explain home bias mainly by referring to investor behavior in forming their expectations of the return and risk of a stock. In their 1995 study of Canadian and US based investors, Tesar and Werner find a strong home bias and show that to the extent there is international investment, this is guided by geographic proximity. However, based on an analysis of turnover data, they also find that transaction costs are not an important impediment to international diversification. Warnock (2002) confirms this result using direct data on transaction costs. The last paper discussed here, is one by Lewis (1999). She calculates that for the sample of countries used the minimum variance portfolio would have a share of 39.5 per cent in foreign equity. The observed share of foreign equity is only 8 per cent. 2.4 Conclusions of the chapter To sum up, the main theories that are used in this thesis are the theory of financial intermediation, the CAPM and its international extension, the international asset pricing model. The first theory explains the existence of financial intermediaries by referring to their roles in reducing informational asymmetries and other trading costs and providing risk sharing opportunities. This theory therefore, forms the basis for the hypotheses that will be developed in the next chapter. The second and third theories stress the importance of (international) diversification in determining the optimal equity portfolio. These theories thus establish the benchmark for evaluating the level of international equity portfolio diversification that will be used in chapter 5. 6

8 3 Hypotheses The structure of this chapter is as follows. In section 3.1 the main hypothesis regarding the overall effect of financial intermediaries is developed. This is followed by the hypotheses on the effect of different types of intermediaries in section Overall effect of financial intermediaries Following the same order as in section 2.2 of this thesis, the function of lowering information acquisition costs leads to the first effect of financial intermediaries on international equity portfolio diversification. In general, the costs of acquiring information are higher for international investments than for domestic ones. Potential reasons for this are lower availability of soft information, less familiarity with foreign firms, differences in accounting and corporate governance standards which make the information available less comparable, language differences and time lags in information availability. Investors are thus better informed about domestic firms than about foreign firms. As shown by Brennan and Cao (1997) this can lead to home bias or, in other words, a lack of international equity portfolio diversification. However, financial intermediaries could reduce the abovementioned costs of acquiring information about international investment possibilities. As outlined in chapter 2, due to economies of scale and scope, financial intermediaries can significantly reduce information gathering costs. Furthermore, also for scale reasons they have a higher incentive than individual investors to acquire information about a potential investment. This thus implies that financial intermediaries should hold more internationally diversified equity portfolios. A second possible channel through which the effect of financial intermediaries on international equity portfolio diversification might operate is the reduction of (fixed) transaction costs. Although part of the home bias literature reviewed in section 2.3 showed that transaction costs were not inhibitive to international investment, these studies do not distinguish between transaction costs to different investors. Therefore, assuming that transaction costs are lower for financial intermediaries this implies a positive effect of financial intermediaries on international equity portfolio diversification. The last potential effect considered is the diversification function of financial intermediaries. Since the minimal units of investment are more or less fixed, a priori, diversification is easier to achieve in a large portfolio than in a small one. However, financial intermediaries can also 7

9 diversify domestically across regions, sectors and in some cases also generations. These alternative diversification opportunities could reduce the incentive to share risk internationally. Nevertheless, all in all, the discussion above leads to the following hypothesis: Hypothesis 1: The overall effect of financial intermediaries on international equity portfolio diversification is positive. 3.2 Individual effect of different types of financial intermediaries The first type of financial intermediaries considered is banks. For banks the general effects explained above all apply. In particular the effect of increasing information availability is important for banks and runs mainly through the sale of information to investors via analyst reports and investment advices. Furthermore, since banks operate more and more on a global scale, economies of scope in gathering information about international equity investments become larger which reduces the costs of acquiring information. However, this globalization of banking can also have a negative effect on equity portfolio diversification as it might act as a substitute. Next to reducing informational costs, also in the reduction of other transaction costs banks can be of particular importance as they can provide investors easier access to trade in international equity markets and provide asset servicing tasks. These arguments thus lead to the following hypothesis: Hypothesis 2: The effect of banks on international equity portfolio diversification is positive. The next financial intermediary considered is mutual funds. Also for this group in principle the general effects described in section 3.1 apply. Only the reduction in transaction costs may be less applicable for mutual funds since investors also pay a fee for the management of the funds. Nevertheless, especially the effects of providing investment tools and diversification in the funds portfolios are important. As to the former, for example investment in emerging market equity can be more easily done via a mutual fund than on an individual stock basis. As to the latter, diversification possibilities are one of the main reasons of existence of mutual funds. However, the effect depends heavily on the focus of the mutual fund considered. Domestically focused funds of course have no effect on international equity portfolio diversification while mutual funds focusing on a group of countries do. Nevertheless, on a 8

10 country-wide scale, both types of funds can be expected to be available. Hence, the reasoning above yields the following hypothesis: Hypothesis 3: The effect of mutual funds on international equity portfolio diversification is positive. Whereas the expected effects of the first two types of financial intermediaries was relatively unambiguously positive, for insurance companies this is not so clear. On the one hand, there are of course the positive effects of diversification and information gathering. On the other hand, insurance companies might lower the incentive for international diversification. As argued by Pagano (1993: 617), the existence of insurance markets might decrease precautionary savings by reducing endowment risk. Analogous to this reasoning, since insurance reduces the fluctuation in individuals wealth, it may increase the willingness to accept risk in investing. This increased willingness to take risk could be in the form of reduced diversification. Furthermore, in particular life insurance companies have an additional diversification possibility as they can spread risk over different generations. This will be explained further in the discussion of pension funds. From this the following hypothesis can be derived: Hypothesis 4: The effect of insurance companies on international equity portfolio diversification is negative. The last type of financial intermediaries considered in this thesis is pension funds. For pension funds reductions in information gathering costs and diversification in the funds portfolios are the most important factors that have a positive effect on international diversification. However, the latter effect will be very small. The main reason for this is that pension funds are often restricted by regulators in the portfolio share that can be invested in a particular asset class and/or region. They may even be encouraged to invest domestically and mainly in longterm (government) bonds which again limits the scope for equity diversification in funds portfolios. Furthermore, similar to insurance companies, pension funds might reduce the incentive to share risk internationally. The reason is that pension funds have an additional domestic diversification possibility since they can spread the effect of a stock market crash over different generations by both reducing the pension payouts to current retirees and raising the contributions paid by current workers. Of course, the extent to which this is possible depends on the legal and regulatory framework. Nevertheless, the existence of an additional 9

11 domestic diversification opportunity could reduce the incentive to share risks internationally. Therefore, the hypothesis becomes as follows: Hypothesis 5: The effect of pension funds on international equity portfolio diversification is negative. 10

12 4 Data This chapter proceeds as follows. Section 4.1 describes the databases used for the analysis. Next, section 4.2 describes the dependent and explanatory variables that will be used in the regressions of the next chapter. 4.1 Databases The main database used is the IMF s Coordinated Investment Portfolio Survey (CIPS). This survey is conducted on an annual basis since 2001 and provides data on year-end holdings of international portfolio investment at the country level. The data used in this thesis are taken from table 12.1, which reports the aggregate international equity investment of a country. According to the metadata questionnaires available, data are often collected by the national central bank and financial intermediaries are in most countries mandated to report their holdings. As a result, data coverage of the international equity holdings by financial intermediaries is either full or at least above 90% in most of the countries. For countries, participation in the survey is on a voluntarily basis but the number of countries participating has increased up to 73 in However, due to data limitations concerning the explanatory variables, the sample is restricted to 65 countries. Due to data limitations of the other databases used, this sample will be restricted further in some of the regressions. For Eurozone countries, data on total assets of banks are taken from the ECB. Data on the total assets of banks in other countries are from the IMF s International Financial Statistics (IFS) database. Also data on GDP levels and exports and imports of goods are taken from this database. For data on the size of other financial intermediaries the OECD Institutional Investors statistics database is used. This dataset covers 34 countries for the period For most countries data are available on asset holdings of mutual funds, insurance corporations and pension funds. However, some countries do not report data on all three categories of financial intermediaries mentioned above. Finally, data on stock market capitalization were extracted from the World Bank indicators database. The same holds for data on the remaining explanatory variables which will be explained further in the next section. 11

13 4.2 Variables The dependent variable in all the regressions is international equity portfolio diversification (IEPD) measured as: (1) where, is the foreign asset position in portfolio equity, = is total domestic wealth in portfolio equity, denotes the domestic stock market capitalization and is the world stock market capitalization. This measure equals 1 if international diversification is perfect (according to the CAPM) and 0 if all portfolio equity is held domestically. This measure follows the measure used by Bracke and Schmitz (2008), but is adapted to reflect diversification instead of home bias. Since this thesis analyses the effect of financial intermediaries on international diversification, the main explanatory variables are FI_total, Bank, MF, Insurance and PF. FI_ total measures total assets of all categories of financial intermediaries and is expressed as a fraction of GDP. Bank, MF, Insurance and PF measure the assets of banks, mutual funds, insurance companies and pension funds respectively and are also expressed as a fraction of GDP. As discussed extensively in chapters 2 and 3, the main channel through which financial intermediaries are expected to affect international diversification is economies of scale and scope. This is best measured by the size of financial institutions in an economy. In the regressions, controls are included for several macroeconomic variables and measures of development of the domestic stock market. First of all, GDP per capita is, for reasons of comparability, measured in current US dollars and used as a measure of overall economic development. Further, also GDP per capita growth is included. This is a proxy for domestic return and thus likely correlated with international diversification. Moreover, as is known from the large literature on finance and growth, GDP growth is also (positively) related to the size of the financial sector. Omission would thus result in an upward bias of the results. As a proxy for macroeconomic stability, inflation (i.e. the annual percentage change in GDP deflator) is used. The last macroeconomic control variable is international trade openness. This is an index of exports plus imports as a fraction of GDP. International trade openness is included since it may be a substitute for equity portfolio diversification as it also allows to share risk internationally. As measures of domestic development of the stock market, the 12

14 number of domestic listed firms and stock market turnover as a percentage of market capitalization are included. The former measures domestic portfolio diversification opportunities whereas the latter is a measure of liquidity of the domestic stock market. In some of the regression specifications also dummy variables for EU and EMU membership are included since member countries face fewer barriers to international investment within the union and thus might be more integrated. Table 1 presents summary statistics of the variables. 13

15 5 Results and findings This chapter is structured as follows. Section 5.1 presents the aggregate effects of financial intermediaries on international equity portfolio diversification. In section 5.2, this is followed by the effects of different types of financial intermediaries. Section 5.3 performs several robustness checks and section 5.4 analyzes whether the results are affected by differences in financial structure. Section 5.5 concludes. 5.1 The aggregate effect of financial intermediaries A first approximation of the aggregate effect of financial intermediaries is given by the regression results of table 2. In regressions 1 to 5 which are the base regressions, FI_total enters positively and statistically significantly. However, the effect is economically quite small. A one basis point increase in the ratio of assets to GDP brings international equity portfolio diversification only about basis points closer to the benchmark. Nevertheless, this result is relatively robust for different regression specifications and the variation in the size of financial intermediaries explains a large part of the variation in international diversification. A point that has to be kept in mind, though, is that FI_total is a weighted average of different financial intermediaries in which banks carry by far the largest weight. To illustrate this point, the correlation between FI-total and Bank is very high at 0.97 which implies that mainly the effect of banks is estimated. Furthermore, due to data limitations the measure could not be constructed for all countries. Therefore, for a full picture of the overall effect of financial intermediaries, it is necessary to also look at the other regressions in which the different intermediaries enter separately. This will be discussed further below. 5.2 The individual effect of different financial intermediaries To determine the individual effect of the four intermediaries, first all types are entered separately into the regressions. The results of these regressions are reported in tables 3 to 6. However, if only one type of intermediary is included at the time, the effects found for a particular intermediary could reflect to some extent the overall effect of financial intermediation. To test whether this is the case, all intermediaries are entered together in the regressions of table 7. A cautionary note on the interpretation of the results from this table is needed, though, since correlations among different types of intermediaries are in some cases very high. This might thus lead to multicollinearity issues. As will be discussed further in the section about mutual funds, in particular, the correlation between Bank and MF is very high at 14

16 approximately 0.9. Also the correlation between Bank and Insurance is at 0.8 relatively high. The correlations between PF and the other financial intermediation variables are the lowest, pension funds and mutual funds are even slightly negatively correlated. Only the correlation with Insurance is at approximately 0.7 quite high. Following the order of financial intermediaries used in chapter 3, banks are the first type of financial intermediaries to be discussed separately. Table 3 presents the results of the estimations of the effect of banks on international equity portfolio diversification. Similar to the findings in the previous section, Bank enters positively and statistically significant in all base regressions. Furthermore, compared to the other financial intermediaries the effect is large. However, this might be partly due to the fact that the sample size in these regressions is larger than in the regressions of the other financial intermediaries. As stated before, the estimates for the effect of banks may thus reflect both the effect of financial intermediation and the individual effect of banks. Looking at table 7, this is indeed confirmed. Although the estimate of Bank remains positive, it is not significant anymore especially after controlling for domestic stock market development and time effects. The picture for mutual funds is similar to that of banks. MF enters positively and statistically significant in all regressions of table 4 and the effect is also larger than the overall effect of financial intermediaries albeit smaller than the effect of banks. However, the strongly positive effect completely disappears when controlling for other financial intermediaries. A possible reason for this is that there is a large overlap of assets of mutual funds and assets of other financial intermediaries since other intermediaries can invest part of their assets into mutual funds. Looking at correlations between the categories of intermediaries confirms this. As stated before, especially, the correlation between mutual funds and banks is quite close to unity. Therefore, controlling for banks, removes the effect of mutual funds. For the third category of intermediaries, insurance companies, the results are mixed. Generally their effect is not very different from 0 although significance varies. Furthermore, it seems that most of the positive effect estimated in the first 4 regressions of table 5 is driven by time effects. After controlling for these, the estimate for insurance drops severely and becomes very insignificant. The same occurs in the regressions of table 7. Whereas Insurance enters positively in the first 3 regression, the estimated effect turns even negative in regression 4. 15

17 The last category of financial intermediaries, pension funds seems to have a negative effect on international equity portfolio diversification. In most regressions PF enters negatively although again significance varies. Nevertheless, pension funds seem to explain little of the variance in international diversification as shown by the low in regression 1 of table 6. Furthermore, also in this case time effects seem to have an important role. In contrast to the findings for insurance companies, controlling for time effects makes the estimates less negative. The findings of table 6 are confirmed by those in table 7. In all of the regressions in this table the effect of pension funds is negative and relatively large although again taking out time effects makes the estimated effect insignificant. 5.3 Robustness checks Being a country known for its large financial sector, most of the extreme (very large) observations on size of financial intermediaries were from Luxembourg. This is especially the case for the FI_total, Bank and MF variables. Furthermore, with values of IEPD around 0.9, international diversification in Luxembourg is quite high as well. Therefore, to check whether the observations for Luxembourg have biased the results, regression 6 in tables 2 to 7 includes a dummy variable for Luxembourg. As it turns out, the observations for Luxembourg have indeed significantly changed the results although the extent varies for different types of intermediaries. As a result of controlling for Luxembourg, the beta of the aggregate financial intermediation measure, FI_total, turns from positive and highly significant to negative although statistically insignificant. This should however not come as a surprise given that on the FI_total measure the observations are very extreme and the sample size of the regressions is relatively small. The sample size of the regressions for banks in table 3 is much larger and hence the effect of controlling for Luxembourg is smaller although still considerable. However, somewhat surprisingly given that Luxembourg has a very large mutual fund sector, the estimated effect of mutual funds on international diversification increases when taking out Luxembourg. On the contrary, the beta of Insurance turns negative and significant whereas it was insignificantly positive without controlling for Luxembourg. The result for pension funds is least affected as the beta remains negative and insignificant. As a second robustness check, the FI_total, Bank, MF, Insurance and PF variables are included on a logarithmic scale in order to reduce the importance of extreme observations (results not reported). Contrary to what would be expected based on the summary statistics; this did not change the results for the FI_total measure. Sign and significance remain largely 16

18 the same (significance level changes only moderately). The same holds for the larger bank sample. Including Bank on a log-scale makes the findings even somewhat stronger. The results for the mutual fund and insurance regressions do not change much when including the variables of interest on a logarithmic scale. Again, this is somewhat surprising given that especially MF exhibits a large variation. Finally, the results for PF are not altered at all by the different functional specification. 5.4 Financial structure The analysis conducted in the previous sections does not distinguish between countries that have different financial structures. However, this distinction might be important as in more market-based financial systems the effect of financial intermediaries could potentially be smaller than in more bank-based systems. A possible explanation for this is that in the former systems, financial intermediaries hold less equity and hence the effect of the diversification function is smaller. To analyze whether this is confirmed by the data, first the results on the domestic stock market development variables will be examined. Afterwards, the effect of financial structure will be determined more directly by including an interaction term of the measures of financial intermediation and stock market capitalization as a fraction of GDP. Looking at the results for the domestic stock market development variables, especially Listed seems to be important. This variable enters negatively and statistically significant at the 1% level in all regressions in which it is included. Hence, this result suggests that a more developed domestic stock market in terms of the number of listed firms reduces international diversification. The same holds for turnover, although the effects are much weaker. Furthermore, comparing regressions 2 and 3 of tables 2 to 7 shows that inclusion of the stock market development variables significantly reduces the effects of financial intermediaries on international diversification. For FI_total, MF and Insurance, the estimated effect approximately halves. The reduction in the estimate of PF is even stronger, only one-third remains after the inclusion of Listed and Turnover. Somewhat surprisingly, given that the distinction is between bank-based and market-based systems, the effect on banks of including the stock market variables is smallest. So, these results suggest that exclusion of the stock market variables has led to a significant upward bias in the estimates of regressions 1 and 2. They also give a first intuition that the effect of financial intermediaries on international diversification may indeed be smaller in market-based systems than in bank-based systems. 17

19 Although the analysis above already suggests that the effect of financial intermediaries could be smaller in countries with more developed stock markets, it is insightful to also test for this directly by including an interaction term. Therefore, an interaction term of the measures of financial intermediation and stock market capitalization as a fraction of GDP is included in the regressions of table 8. Including this interaction term changes the results for FI_total and Bank only marginally. The interaction term indeed enters with the expected negative sign, confirming that a in a more market-based financial system the effect of financial intermediaries on international equity portfolio diversification is smaller. Looking at the other categories of financial institutions shows that the effects of including the interaction term are larger. The estimated effect of mutual funds is reduced by one-sixth. The effects of the interaction term are even larger for insurers and pension funds. The estimated effects of these intermediaries respectively halve and disappear completely. Similar to the analysis of Listed and Turnover, this is somewhat counterintuitive given that the distinction between marketbased and bank-based financial systems would be expected to matter most for banks themselves. 5.5 Conclusions of the chapter In sum, based on the findings of this chapter, the overall effect of financial intermediaries on international equity portfolio diversification is positive but small. However, the majority of this positive effect is driven by the largest category of intermediaries, banks. Also mutual funds have a positive effect on international diversification but this disappears completely when taking into account the other types of financial intermediaries. Insurance companies have a negligible effect. On the contrary, pension funds have an outright negative effect on international equity portfolio diversification. Taken together these findings thus partly confirm the theory discussed in chapters 2 and 3. Financial intermediaries can indeed increase international diversification by increasing information availability and reducing (fixed) trading costs. However, to the extent that they improve the opportunities for domestic risk sharing, they might reduce the need for international diversification. This result is strengthened by looking at the stock market development variables. A more developed domestic stock market reduces the need to diversify internationally. Next to that, a larger stock market as measured by market capitalization reduces the effect of financial intermediaries on international equity portfolio diversification. The extent to which this occurs varies for different types of financial intermediaries. For banks and mutual funds the 18

20 effect is rather small while much if not all of the estimated effect of insurance companies and pension funds disappears after including an interaction term with stock market capitalization. 19

21 6 Conclusions The lack of international risk sharing has by now become a well-documented phenomenon in both the international banking literature and even more directly in the home bias literature. Both strands of literature find that the potential benefits of international diversification are large but due to a variety of reasons remain unexploited. However, following the theory of financial intermediation, financial institutions could have a role in overcoming some of the main sources of the lack of international equity portfolio diversification. The empirical findings of this thesis indeed confirm that financial intermediaries have a positive effect on international equity portfolio diversification. Increasing the size of financial institutions in an economy generally leads to an increase in international diversification of equity holdings although the increase is small. However, this outcome depends strongly on the type of intermediary considered and thus on the structure of the financial sector in a particular country. As expected banks have a positive effect on international diversification. Furthermore, their effect is the largest of the different types of intermediaries considered. Also the effect of mutual funds is positive but their effect is much smaller than that of banks and even negligible if other financial intermediaries are taken into account. The role of insurance companies is generally very small as well although not in all cases negative as was expected. The effect of the last type of financial institutions, pension funds, was found to be negative as predicted. t only the structure of the financial sector matters. Also whether a system is predominantly bank-based or market-based has some influence on the results. In more market-based systems the effect of financial intermediaries is generally smaller. However, contrary to what would be expected, the estimate for the effect of banks is least affected. Taken together the abovementioned findings thus show that there is a tradeoff between sharing risk domestically and internationally. Financial institutions that improve the opportunities for domestic risk sharing, for example, across generations, reduce the international sharing of risk. On the contrary, financial intermediaries that make it less costly to invest abroad, for example by reducing information asymmetries or other trading costs, improve international diversification. However, in almost all of the countries examined, by far the largest part of financial intermediation is done by banks. And, since this type of intermediary was found to have a positive effect on international diversification, it can be stated that based on the findings in this thesis, the effect of financial intermediaries on the international diversification of equity holdings is positive. 20

22 Although the findings of this thesis have already given some insights on the role of financial intermediaries in the international diversification of equity holdings, there are still many more interesting issues to examine on this topic. For example, as already stated in the introduction it might be interesting to look at international diversification of portfolios of bonds or other financial instruments rather than equity. Furthermore, another possibility to refine the analysis conducted here is to use a more refined measure of international diversification. This would allow controlling for more issues such as the extent of information asymmetry between investors in different countries or the benefits of diversification which depend on the correlations of business cycles and stock market returns. Also incorporating cross-listings and the diversification opportunity offered by holding shares of multinational firms may be an important addition to this thesis since these may both be substitutes for international equity diversification. A final interesting area for future research could be to analyze the structure of financial intermediaries. In particular for banks it might be insightful to distinguish in the analysis between the effects of investment banks and commercial banks or foreign owned and domestically owned banks. 21

23 References Barron, J.M. & Ni, J. (2008), Endogenous Asymmetric Information and International Equity Home Bias: the Effects of Portfolio Size and Information Costs, Journal of International Money and Finance, 27, Brennan, M.J. & Cao, H. H. (1997) International Portfolio Investment Flows, The Journal of Finance, 52, (5), Buch, C.M. & Driscoll J.C. & Ostergaard, C. (2005), Cross-Border Diversification in Bank Asset Portfolios, ECB working paper series, no Dimson, E. & Marsh, P.R. & Staunton, M. (2002), Triumph of the Optimists: 101 years of global investment returns. Princeton, NJ: Princeton University Press. French, K.R. & Poterba, J.M. (1991), Investor Diversification and International Equity Markets, The American Economic Review, 81, (2), García-Herrero, A. & Vazquez, F. (2007), International Diversification Gains and Home Bias in Banking, IMF Working Paper, WP/07/281. Levine, R. (1997), Financial Development and Economic Growth: Views and agenda, Journal of Economic Literature, 35, (2), (2004), Finance and Growth: Theory and Evidence, NBER Working Paper Series, no Levy, H & Sarnat, M. (1970), International Diversification of Investment Portfolios, The American Economic Review, 60, (4), Lewis, K.K. (1999), Trying to Explain Home Bias in Equities and Consumption, Journal of Economic Literature, 37, (2), Markowitz, H.M. (1959), Portfolio Selection: Efficient Diversification of Investments, 2 nd edition (1976), Clinton: The Colonial Press Inc. Mishkin, F.S. (2004), The Economics of Money, Banking and Financial Markets, 9 th edition (2010), Boston: Pearson Education Inc. 22

24 Ni, J. (2009), The Effects of Portfolio Size on International Equity Home Bias Puzzle, International Review of Economics and Finance, 18, Pagano, M. (1993) Financial Markets and Growth: An Overview, European Economic Review, 37, Schoenmaker, D. & Wagner, W. (2011), Cross-Border Banking in Europe and Financial Stability, Tinbergen Institute discussion paper. Sharpe, W.F. (1964), Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk, The Journal of Finance, 19, (3), Solnik, B.H. (1973), European Capital Markets: Towards a General Theory of International Investment, Lexington: D.C. Heath and Company. Tesar, L.L. & Werner, I.M. (1995), Home Bias and High Turnover, Journal of International Money and Finance, 14, (4), Warnock, F.E. (2002), Home Bias and High Turnover Reconsidered, Journal of International Money and Finance, 21, Wooldridge, J.M. (2006) Introductory Econometrics: A Modern Approach, 4 th edition, South- Western Cengage Learning. 23

25 Appendix Table 1 Variable N Mean Standard deviation Minimum Maximum Bank EU EMU FI_total GDP per capita (1) GDP per capita growth IEPD Inflation (1) Insurance Listed (1) MF PF Trade Turnover (1) (1) Natural logarithm is used to reduce the importance of extreme observations. 24