Version: May, Current version: June 8, 2016

Transcription

1 The Causes and Consequence of Fund Level Home Bias Takato Hiraki (a), Ming Liu (b) and Xue Wang (c) Version: May, 2016 Current version: June 8, 2016 Abstract We examine the causes and consequence of fund level home bias using a sample of U.S. global equity mutual funds. We show that home bias at the fund level is much smaller than at the country level. Raw returns do not vary much across funds with the different level of home bias measured by their portfolio weights in U.S. domestic stocks. However, risk-adjusted returns adjusted by the global six (Fama-French five plus momentum) factors show that funds invested more in the U.S. home-market outperform those more invested in the international markets. Fund managers with foreign educational background invest less in U.S. stocks, but record poor riskadjusted performance. More home-biased female and less home-based MBA degree holding fund managers also tend to deliver poor risk adjusted returns. Overall, our finding is consistent with both information and behavioral hypotheses, depending on fund s and manager s home bias characteristics. Finally, we show that differently characterized global funds commonly exhibit flight home behavior during the financial crisis. JEL Classification: G11; G15; G23 Contact Information: (a) Tokyo University of Science, hiraki@rs.tus.ac.jp (b) International University of Japan, mliu@iuj.ac.jp (c) Renmin University of China, xuewang@ruc.edu.cn 1

2 1. Introduction The puzzling phenomenon of home bias in financial markets has attracted many researchers. It is puzzling that investors overinvest in domestic financial assets relative to their weights in the global market portfolio by forgoing the benefits of international diversification. Including French and Poterba (1991), Cooper and Kaplanis (1994), and Tesar and Werner (1995), early papers have demonstrated strong home bias in international investment. Over the past two decades, global equity markets have become more integrated but home bias has remained much unchanged (Karolyi and Stulz, 2003). Researchers have proposed various explanations to home bias: barriers on international investment, hedging of foreign exchange risk, differing labor market risk, withholding tax, information asymmetry and behavioral bias. However, none of the factors alone seems suffice to justify the magnitude of globally observed home bias. Cooper, Sercu and Vanpee (2012) suggest that a combination of the above factors would help explain the puzzle, with information asymmetry and economic openness as the most prominent. One problem with the country-level studies on home bias is that the dispersion of home bias among investors remains undiscovered. In this paper, we provide new insights into the home bias puzzle by utilizing fund level holdings of U.S.-bases global equity funds investing in the U.S. and non-u.s. equities. We explore the relation between funds home bias and their performances as well as manager characteristics. Specifically, we examine the determinants of fund-level home bias using a sample of U.S. global equity mutual funds over Compared to the extant cross-country studies, this study has one advantage in that we eliminate the potential effects of macro and institutional factors on fund s decisions in global investment, such as capital restrictions, withholding tax, country level uncertainties etc. This is because these 1 Hau and Rey (2008) document a large dispersion on the fund level home bias with a sample of both domestic and international mutual funds. 2

3 factors would have the same impacts on all global funds operating in the U.S. Therefore, we have a cleaner setting to empirically test the two hypotheses: information and behavioral biasbased explanations of home bias. To our knowledge, this is the first paper that conducts a comprehensive study on fund level home bias, performance and fund manager characteristics. Several theory papers support information-based explanation for home bias. Gehrig (1993) proposes information asymmetry as one possible explanation. The paper shows that, based on a noisy rational expectations model with two countries and two risky assets, the higher precision on domestic than on foreign signals results in investors bias towards domestic assets in equilibrium. Brennan and Cao (1997) derive the dynamic information asymmetry implications of investment flow rather than the static implications in Gehrig (1993). Their main result shows that the purchase of foreign assets by domestic investors is positively correlated with returns of foreign assets, and that home bias holds as long as the domestic market co-varies with the world market. Van Nieuwerburgh and Veldkamp (2009) develop a two-country general equilibrium model with heterogeneous investors in each country. Similar to the previous papers, the model begins with investors who have more (less) information endowment on domestic (foreign) assets than the average investor of the two countries. Their model allows investors to choose which information to learn in maximizing their mean variance utility. The equilibrium outcome is that home investors, with an initial information advantage on home assets over foreign (and average) investors, choose to learn home risk factors which results in home bias in their portfolios because home assets appear less risky to them. Empirical studies mostly support the information-based explanation of home bias. Kang and Stulz (1997) find that foreign investors in Japan tend to overinvest in stocks of large and 3

4 export-oriented companies, consistent with the information-based explanation. Ahearne, Griever and Warnock (2004) find that the more companies a foreign country lists in the U.S. stock market, the more U.S. investors invest in that country, again supporting the information-based explanation of home bias. Andrade and Chhaochharia (2010) find that U.S. Foreign Direct Investment (FDI) in a destination country in the 1990s is positively related to U.S. portfolio investment in that country in , consistent with the learning story of Van Nieuwerburgh and Veldkamp (2009). Home bias may also be caused by behavioral reasons. For example, investors may perceive familiar assets to have higher expected returns and less return dispersions (Huberman, 2001; Hiraki et al., 2003; Pool et. al, 2012). Graham, Harvey and Huang, (2009) show that the more competent the investors think themselves, the more they invest in international stocks. Morse and Shive (2011) show that the more patriotic a country is believed, the more domestic assets the investors of that country hold. To study the causes and consequence of home bias at the fund level is of great importance and interest in international finance. Firstly, the observed country level home bias is an aggregation of investment decisions at the individual investor s level. A better understanding of international investment at the granular level helps us solve the home bias puzzle documented at the country level. Secondly, global fund managers are presumably more sophisticated and have better learning capacity with research resources than the average investors modeled in the previous (country-level) home bias papers. Moreover, they may have different objectives than individual investors due to the principal-agent relations (He and Xiong, 2013). It would be interesting to test the implications of information asymmetries and/or behavioral biases from a viewpoint of professional money managers. Since fund managers may have better learning 4

5 ability than individual investors, they may easily transmit the increased precision on home risk factors to increased precision on global risk factors. For example, they may apply what they learn in the domestic markets about specific industries to the global markets. One way to differentiate the information and the behavioral hypothesis on home bias is to examine the performance. Better information might imply a positive relation between home bias inclination and performance while behavioral bias would at best imply a neutral relation. We sort funds into quintiles based on their portfolio weights in U.S. domestic stocks and measure subsequent six-month returns for each quintile. We find that funds with the highest investment in U.S. stocks (the most home biased) have indistinguishable net returns from those with the lowest investment in U.S. stocks (the least home biased). If anything, the middle quintiles (funds with intermediate home bias inclination) slightly underperform the other quintiles. Next we risk-adjust fund portfolio returns using the U.S. 4-factor (U.S. Fama-French three factors plus U.S. momentum factor), global 4-factor (global Fama-French three factors plus global momentum factor), U.S. 6-factor (U.S. Fama-French five factors plus U.S. momentum factor), global 6-factor (global Fama-French five factors plus global momentum factor) models. All factor returns data are downloaded from Professor Ken French s website. In the case of global 6-factor adjusted returns, the most home biased founds outperform the least home biased by percent per month, which translates into 2.24 percent per annum. The factor loadings estimated with the global six factors show that the most home biased funds have uniquely different loadings on size, profitability and investment factors from other quintiles, especially, the lest home-biased. This characteristic is weak with the comparable U.S. factors. The result, based on the most prominent risk-adjusting model, supports the information rather than the 5

6 behavioral hypothesis explaining a basic relationship between home-bias and performance. Our result highlights the importance on how to benchmark global mutual fund returns. We hand-collected fund manager attributes to study the determinants of fund level home bias. We find two interesting points: first, fund managers with foreign educational background and MBA degree tend to invest more in international stcks; and second, female fund managers and managers with CFA tend to invest more stocks in the U.S. market. Examining the determinants of fund performance, we first find that higher level of home bias tends to associate with higher risk-adjusted returns, and second, that fund manager s foreign educational background tends to associate with lower returns. Overall, while our findings support the view that home bias can be explained by the information hypothesis, the result associated with some manager s characteristics, especially, foreign educational background is consistent with the behavioral hypothesis. Our final empirical investigation on the flight home effect of economic crisis shows that global funds pull investments back to the U.S. market prior to and during the most recent financial crisis. This effect holds in nearly all cases of fund s and manager s characteristics. Fund manager s foreign educational background and the level of home bias before the crisis do not explain much this flight home flow movement during the pre-crisis and early crisis period. One notable exception applies to funds with large family size which tend to be less exposed to the flight home effect. It could be that funds with large resources (proxied by fund family size) are less likely to flight home. 6

7 The paper relates to several strands of literature. 2 First, we add new insights to the existing work on home bias at the individual fund level with the data of more precise asset holdings by U.S. global funds. Hau and Rey (2008) show that there is wide dispersion in fund level home bias across countries and funds. Coval and Moskowitz (1999) find that U.S. mutual fund managers have strong preference for local firms, a phenomenon they refer to home bias at home. Other papers study whether households underinvest internationally, using Swedish data (Calvet, Campbell and Sodini, 2007; Kalsson and Norden, 2007; Norden 2010), survey data from the U.S. investors (Graham, Harvey and Huang, 2009), proprietary data set from a U.S. investment advisor (Bekaert et al., 2015). Our study relates to this strand by investigating differing degrees of home bias among U.S.-based global funds, their performance and the determinants of home bias at the fund level using fund s holdings data. The second strand to which this study might contribute is the relation between mutual fund manager (team) characteristics and investment behavior as well as performance. Chevalier and Ellison (1999) find that managers attending high-sat undergraduate institutions outperform the control group. Cici et al. (2015) show that mutual fund managers prior experience in industries other than financial sector improves their ability to pick stocks from these industries. Patel and Sarkissian (2014) find that Morningstar provides more accurate data on fund managers and that a team with three members performs better than teams with less or more members. Massa and Schumacher (2015) show that it is optimal for fund families to outsource fund managers when investing foreign countries due to the information advantage. None of these studies, however, relates the fund and manager attributes to fund level home bias. 2 See, for example, review articles Karolyi and Stulz (2003), Coeurdacier and Rey (2012) and Cooper, Sercu and Vanpee (2012). 7

8 Lastly, our study is related to the literature on portfolio concentration and its performance implications. We investigate a possible reason for performance differentials between more homebased and less home-based funds. There is rich literature in this area (Kacperczyk et al., 2005; Brands et al., 2005; Fedenia et al., 2011; Ivkovic et al., 2008; Goetzmann and Kumar, 2008; Kacperczyk et al., 2014; Hiraki et al., 2015). While the evidence is mixed, majority of the findings support the notion that investors with more private information hold concentrated portfolios. Again, none of these studies, however, relates the fund and manager attributes to fund concentration in the context of fund level home bias. The paper is organized as follows. In Section 2, we introduce sample and data sources. Section 3 presents empirical results. Section 4 shows evidence on the flight home phenomenon. Section 5 concludes the study. 2. Data Our data are from various sources. The mutual fund holdings data are from Morningstar and fund characteristics are from CRSP Mutual Fund Database. Both databases are survivorshipbias free. We merge these two datasets and include only global equity funds in the sample. We select global equity funds to study home bias because they may invest both foreign and domestic stocks. We follow fund category provided by Morningstar and fund investment objective codes by CRSP. We manually check the fund prospectus to make sure whether the investment objective of our sample fund is to invest in global equities. In doing so, we are able to remove the institutional constraints that may affect fund s decision on international diversification. As a final check, we closely examine whether funds have experienced a very large change in the portfolio weight in U.S. stocks. For example, suppose that a fund may have a 0% investment 8

9 weight in U.S. stocks at one point in sample and the number increase to 45% six months later. It is likely that the fund has changed from "international" to "global equity" fund. In this case, we remove the fund portfolio records, i.e., "international" funds and include "global" funds since they have changed the U.S. weight suddenly from 0% to 45% investment in U.S. stocks. The holdings data set from Morningstar includes individual holdings country and industry information. About 5% of the stock holdings data of the global funds do not have country or industry information. These records correspond to about 9500 unique holding security names. We manually match these names with stock names retrieved from Datastream in order to populate the country and industry data. We match the industry classification of Morningstar with that of Datastream. Our final sample includes 320 global funds from December 1999 to December We select the last fund portfolio date by the end of June and December in each year during sample period, which gives us 4,161 fund portfolios (82% of the fund report dates fall in June or December). From year-end market capitalization of stocks from 51 markets 3 in Datastream, we calculate the market capitalization value of the world portfolio and the weight of U.S. market. We use these calculated market weights as a benchmark for global diversification. Finally, we obtain fund manager tenure information from Morningstar Direct and supplement this data set with hand-collected data from Internet sources such as Linkedin, fund home page, SEC filing, Bloomberg, and Zoominfo. The main fund manager characters are: start and end date at the fund, gender, undergraduate institution, MBA school, other graduate studies schools, and CFA. From the fund portfolio report date and fund manager start / end date at the fund, we know the managers working in each of the semi-annual before and at the fund portfolio report date. If there is more than one manager, we take the average of fund managers 3 The 51 markets we use account for nearly 99% of the total global market. 9

10 characteristic variables. For example, AQR Global Equity Fund has five managers in the second half of All of them are male and three of them have MBA degree. So the values for managers female and MBA variables are 0 and 0.6, respectively. We are able to collect fund manager information for 315 (out of 320) funds and 4,074 (out of 4,161) fund portfolios. Table 1 reports the summary statistics of fund and fund manager data. Panel A shows that on average, fund holds stocks from 17.8 countries in the portfolio. The portfolio weight in U.S. stocks is 43.7%, which is in sharp contrast to the aggregate data from household survey, which can be around or above 90%. As shown in Figure 1, compared to the U.S. stock weight in the world portfolio, the global funds overweight U.S. stocks merely by 4%. In other words, the home bias puzzle hardly exists among the global fund managers. The mean and median monthly net returns are 0.595% and 1.062, respectively. Panel B, Table 1 displays the fund managers variables. We are able to collect information for a total of 1,312 managers for 315 funds. The average number of managers per fund is About one third of the team has foreign educational background. There are a small number of female fund managers (with a total of 151 female managers out of 1,312 managers), but nearly half of the manager team have CFA charter or MBA degree. In spite of the fact that the relative home bias is very limited within global funds as an investment style, the variation in the fund-level home bias indeed exists and possibly provides us with interesting cross-sectional results. Panel A of Figure 1 documents the median (absolute) home bias of global fund portfolios in terms of the percentage weight in U.S. domestic stocks from 1999 to It has been moderate and stable, close but not more than 50 percent in the U.S. stock weight throughout the sample period. On average, the median fund s relative home bias is 4.07 percent. It seems that U.S. global funds exhibit much less home bias. This is because 10

11 the sample represents U.S. global equity funds. In contrast, Andrade and Chhaochharias (2010) report in their Table 1 that during , U.S. residents hold in their portfolio a mere 12.6 percent in foreign stocks, or 87.4 percent in U.S. stocks. In Panel A, we also examine a sample of 25 global funds which exist throughout , an entire sample period. The pattern is similar to the case of full sample. In both full sample and subsample of funds, we observe a small but identifiable increase in home bias after the financial crisis, which is consistent with the flight home effect (Bae and Zhang, 2015). Panel B of Figure 1 shows the distribution of fund level home investment. The largest proportion of global funds investments in the U.S. market ranges between 40 and 50 percent, followed by the 30-to-40 percent range. This distribution shows a moderate dispersion of (absolute) home bias among the funds specialized in global investments. It is interesting to investigate why home bias differs among these funds, which motivates our cross-sectional investigation in the subsequent sections. 3. Empirical results on home concentration and performance 3.1 Raw returns and risk-adjusted returns Table 2 displays the average monthly returns of fund quintiles sorted by their U.S. stock weights. We sort them at the end of each half year. Fund net returns are obtained from CRSP, and we back out gross returns by adding the expense ratio to net returns. We also use U.S./global four-factor (Fama-French U.S./global three factors plus US/global momentum factor) and sixfactor (Fama-French U.S./global five factors plus U.S./global momentum factor) models to adjust the returns for risk. All factors are obtained from Professor Ken French s website. The net and gross returns show no statistically significant difference between the fund groups with high 11

12 and low weights in U.S. stocks. The average fund portfolio net returns for high and low home biased are 0.499% and 0.503%, respectively. The return differences after risk-adjustment are also indistinguishable from zero except for the global six-factor adjusted returns, which yields significant 0.187% per month and translates into economically significant 2.24% per year performance. The result that global equity mutual funds with higher investment in U.S. stocks outperform those with low investment in U.S. stocks provides supportive evidence for the information hypothesis of home bias. Moreover, the result highlights the importance of the choice of benchmark when evaluating mutual fund returns. Chan, Dimmock and Lakonishok (2009) show that holding based benchmark and attributes/factor based benchmark may produce different inference on domestic equity fund performance. This issue can potentially become more pronounced for investments in the global equity market. We show that different factor models can produce indeed very different results for performance valuation. This aspect is investigated through our analysis of factor loadings as follows. In Table 3, we report the factor loadings of fund quintile returns on U.S. and global six factors. The adjusted R 2 is higher across quintiles in panel B, all greater than 0.926, than in panel A. Risk exposure (especially to size) and significance patterns between the low and the high home bias quintile are similar across the six U.S. factors as indicated the column Difference(1-5) in panel A. In contrast, the result in Panel B shows that funds with low investment in U.S. stocks (i.e., high investment in international stocks) have a higher and more significant loading on international size factor, especially, suggesting these funds may allocate more capital on small stocks in the international market than other funds. These funds also have significant loadings on the two newly developed factors in the Fama-French five factor model: profitability and investment. For example, the funds with low portfolio weight in U.S. stocks (or, high portfolio 12

13 weight on international stocks) have a significant loading on RMW (robust minus weak) of 0.166, while the funds with high portfolio weight in U.S. stocks (i.e., home bias) have a negative loading of to the same factor. It seems that there is a wide dispersion on types of stocks held by global mutual funds. Based on the above discussion, it is possible to argue that risk-adjusted global fund performance and performance variation across quintiles are more effectively explained by the global six factor model than the U.S. six and other less representative U.S. and global factor models. The highest overall explanatory power and an ability to capture differing factor contribution across quintile portfolios are achieved by the global six factor model. The information hypothesis of home bias is supported by this global six factor model: even global fund managers have more information advantage in domestic stocks than foreign stocks. Global equity mutual funds with higher investment in U.S. stocks outperform those with lower investment in U.S. stocks. At this moment, however, a support for the information hypothesis does not necessarily mean that the behavioral hypothesis is refuted. Fund level home bias is still caused by manager s attributes. We discuss this issue below. 3.2 The determinants of home bias and fund performance regression analysis Previous literature has proposed information and behavioral bias hypothesis for home bias puzzle. Although variations exist, a standard information-based explanation argues that investors have higher precision of information on home assets and capitalize this advantage in equilibrium. One implication of this argument is that investors may earn higher returns on home 13

14 asset. 4 Behavioral explanations, on the other hand, assume that investors without information advantage overweigh home assets due to behavioral reasons, such as familiarity. In this case, investors may (incorrectly) perceive they have better information on home assets. In a global setting, it is an empirical matter whether global fund managers have information advantage (or perceived advantage) on foreign assets. We set up a dummy variable for each fund manager to indicate if she has studied undergraduate or graduate programs outside the U.S. We take average of this variable of all managers if the fund is managed by a team. Similarly, we also set variables for a female manager, CFA holder, manager with the MBA degree. In Table 4, we report the regression result of the determinants on fund-level portfolio weights in U.S. stocks. The independent variables include fund manager characteristics (averaged across the management team if applied) that we hypothesize and other fund attributes that might affect home bias of funds. The result shows that foreign educational background significantly reduces investment in the U.S. market. For example, if a management team increases the proportion of managers with foreign educational background from 0% to 100%, the fund will reduce the investment weight in U.S. stocks by 2%. Moreover, funds managed by female managers and managers with CFA tend to invest more in U.S. stocks, while fund managers with the MBA degree tend to invest more in the international market. Finally, funds from large fund family seem to invest more in international markets. The coefficients associated with these variables are all significant at the 1% level. At this moment, each significant result is consistent with either information or behavioral hypothesis, not both at the same time. 4 There is a caveat when testing the cross-section difference on the level home bias using the information-based home bias model, because the investors invest less in home assets, thus more on international assets, may also be an outcome of their information advantage on foreign assets. 14

15 Next, we propose the panel regression of fund risk-adjusted performance on fund manager characteristics and other fund variables. One purpose of the regression analysis is to differentiate information and behavioral explanations of home bias. In each half year, we measure the fund and fund manager characteristics used as independent variables. The dependent variable risk-adjusted returns in the six months after the fund portfolio formation is derived through the Fama-French global five factors and the global momentum factor, called the global 6 factors. We estimate the coefficients (factor loadings) of these six factors for each fund using previous 36 months returns. The risk-adjusted returns are calculated by subtracting the sum of the products of the six factors and corresponding factor loadings from realized returns on each fund. We then run the panel regression and adjust the standard errors on both fund cross section and time dimensions, using the method developed by Petersen (2009). Table 5 shows the results of fund performance determinants in the context of home bias. The higher level of investment in the U.S. market (Portwt_US) predicts higher risk-adjusted returns, which is significant at the 10% level in model (1) without manager's characteristics. As expected, the performance effect of the home bias inclination variable becomes insignificant in model (4) with all fund and manager character variables. 5 In the meantime, the foreign educational background predicts a negative effect on risk-adjusted returns whenever included as in models (2) through (4). This negative and significant relation is not consistent with the informationbased explanation, but more consistent with the behavior explanation. When selecting international stocks, this characteristic does not lead such a fund to better performance but leads to poorer performance. Interestingly, fund management teams with female managers seem also underperforming while, not statistically significant, CFA seems adding value to fund 5 There might be an endogeneity issue between home bias and manager characteristic variables on the right hand of the regression equation. 15

16 performance. Since foreign educational background decreases both the level of home basis and fund performance and female managers invest more in familiar U.S. stocks to decrease riskadjusted performance, the behavioral hypothesis is more consistent with manager s characteristics affecting home bias at the fund level. 4. Flight home during the pre- and early financial crisis period The flight home effect during the period has been documented in the study in international banking loans (e.g., Giannetti and Laeven, 2012a, b). Flight home before and during the domestic crisis is also pervasive in global equity portfolio investment (see for example, Bae and Zhang, 2015). Flight home may affect the level of home bias. In this section, we study whether there is the flight home effect among global equity funds. We focus our study on funds holdings in the early stage of the financial crisis. At the end of 2006 and 2007, we calculate the difference between fund portfolio weight in U.S. stocks and the U.S. stock weight in the world portfolio. If the calculated value is positive, it indicates an overweight in U.S. stocks, or (relative) home bias. We then compare the difference of this value at the end of 2006 and We show the result in Table 6, there are 95 funds in both 2006 and From the end of 2006 to 2007, the level of home bias increases from 7.5% to 10.8%, indicating a flight home effect of global equity funds by the end of We repeat the tests using the average of two years home bias measures: to There is an increase of the flight home effect and it mainly comes from the before crisis time period 2005 to Do funds with high level or low level of home bias before the crisis exhibit more flight home? We classify funds by whether they have home bias before the crisis and examine the 16

17 flight home for both groups. The final part of Table 6 shows that flight home is pervasive regardless of the level of home bias the funds have before the crisis. In Table 7 we show the results of cross-sectional regression on the determinants of flight home flow movement. The constant term is positive and significant with a high magnitude relative to other insignificant and marginally significant independent variables. Funds from large families less exhibit the flight home effect, which is statistically significant at the 1% level. Funds with large resources (proxied by fund family size) are less likely to flight home. It seems that fund managers with CFA tend to flight home more than others, which is significant at the 10% level. 5. Conclusion We study the causes and consequence of fund level home bias using a sample of U.S. global equity funds. We find that funds with different degree of home bias do not exhibit different raw returns. However, when fund returns are adjusted for relevant risk factors included in the global six-factor model, funds with higher level of investment in U.S. stocks outperform those with lower weights in U.S. stocks. This indicates that high level of home bias may be caused by information advantage on domestic assets. It also highlights the importance of benchmark when evaluating global equity fund performance on a risk-adjusted basis. Our tests on fund managers behaviors show that managers with foreign educational background invest more (less) in international (U.S.) stocks, consistent either with information or behavioral explanation. An examination on the relation between fund and manager attributes and fund performance shows that fund managers with foreign educational background deliver inferior risk-adjusted performance based on the most convincing global factor model. This 17

18 suggests that foreign background does not necessarily translate into superior information on international stocks. With this respect, the behavioral hypothesis is more consistent than the information hypothesis. We also show that teams with female fund managers tend to invest more in U.S. stocks and underperform and similarly more domestic-oriented CFA holders tend not to be outperformed. We also find that these CFA fund managers tend to flight home more prior to and during the latest financial crisis and funds from large families less exhibit the flight home effect, one special form of fund slide toward home bias. Overall, our finding is consistent with both information and behavioral hypotheses. Depending on manager s characteristics, an improved risk adjusted performance of more homebiased global equity funds by information is partly offset by the negative performance effect of manager s biased behavior or decision on global vs. domestic asset allocation. Our study provides some evidence to deny positive roles expected to global funds investing globally. This is a partial explanation of the home bias puzzle. References: Ahearne, A. G., W. L. Griever, and F. E.Warnock (2004), Information costs and home bias: An analysis of U.S. holdings of foreign equities. Journal of International Economics 62(2), Andrade, S. C. and V. Chhaochharia (2010), Information immobility and foreign portfolio investment. Review of Financial Studies 23(6), Bae, K.-H. and X. Zhang (2015), The cost of stock market integration in emerging markets. Asia- Pacific Journal of Financial Studies 44, Bekaert G., K. Hoyem, W.Y. Hu, and E. Ravina (2015), Who is internationally diversified? Evidence from (k) Plan. Working Paper. Brands, S., S. Brown, D. Gallagher (2005), Portfolio concentration and investment manager performance. International Review of Finance 5(3 4),

19 Brennan, M. J. and H. H. Cao (1997), International portfolio investment flows. Journal of Finance 52(5), Calvet, L., J. Campbell, and P. Sodini (2007), Down or Out: Assessing the Welfare Cost of Household Investment Mistakes. Journal of Political Economy, 115(5), Chevalier J. and G. Ellison (1999), Are some mutual fund managers better than others? Cross-sectional patterns in behavior and performance. Journal of Finance 54(3), Cici G., M. Gehde-Trapp, M.A. Goricke, A. Kempf (2015), What They Did in their Previous Lives? The Investment Value of Mutual Fund Managers Experience outside the Financial Sector. Working Paper. Coeurdacier, N. and H. Rey (2012), Home Bias in Open Economy Financial Macroeconomics. Journal of Economic Literature 51(1) Coval, J. and T. Moskowitz (1999), Home bias at home: Local equity preference in domestic portfolios. Journal of Finance 54(6), Cooper, I. and E. Kaplanis (1994), Home bias in equity portfolios, inflation hedging, and international capital market equilibrium. Review of Financial Studies 7(1), Cooper, I., P. Sercu and R. Vanpee (2012), The Equity Home Bias Puzzle: A Survey. In: Foundations and Trends in Finance, vol. 7, No. 4. Elton, E.J. and M.J. Gruber (2013), Mutual funds. Financial Markets and Asset Pricing. In Handbook of Economics and Finance, Part B, vol. 2. Elsevier, pp Fedenia, M., S. Shafferb, and H. Skiba (2011), Information immobility, industry concentration, and institutional investors performance. Journal of Banking and Finance 37 (6), Ferson, W.E. (2013), Investment performance: A review and synthesis. Financial Markets and Asset Pricing: Handbook of Economics and Finance, Part B, vol. 2. Elsevier, pp French, K. R. and J. M. Poterba (1991), Investor diversification and international equity markets. American Economic Review 81(2), Gehrig, T. (1993), An information based explanation of the domestic bias in international equity investment. Scandinavian Journal of Economics 95(1), Giannetti, M. and L. Laeven (2012), The flight home effect: evidence from the syndicated loan market during financial crises. Journal of Financial Economics 104 (1): Giannetti, M. and L. Laeven (2012) Flight Home, Flight Abroad, and International Cridit Cycles, American Economic Review, Papers and Proceedings 2012, 102 (3): Graham, J. R., C. R. Harvey, and H. Huang (2009), Investor competence, trading frequency, and home bias. Management Science 55(7),

20 Hau, H, and H. Rey (2008), Home bias at the fund level, American Economic Review 98, He, Z., and W. Xiong (2013), Delegated asset management, investment mandates, and capital immobility, Journal of Financial Economics, 107, Hiraki, T., A. Ito, and F. Kuroki (2003), Investor familarity and home bias. Asia-Pacific Financial Markets 10, Hiraki, T., M. Liu, S., X. Wang (2015), Country and industry concentration and the performance of international mutual funds. Journal of Banking and Finance 59, Huberman, G. (2001), Familiarity breeds investment. Review of Financial Studies 14(3), Ivkovic, Z., C. Sialm, and S. Weisbenner (2008), Portfolio concentration and the performance of individual investors. Journal of Financial and Quantitative Analysis 43 (3), Kacperczyk, M., C. Sialm, L. Zheng (2005), On the industry concentration of actively managed equity mutual funds. Journal of Finance 60 (4), Kacperczyk, M., S. Van Nieuwerburgh, L. Veldkamp (2014), Time-varying fund manager skill. Journal of Finance 69 (4), Kang, J.-K. and R. M. Stulz (1997), Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan. Journal of Financial Economics 46(1), Karlsson, A. and L. Nord en (2007), Home sweet home: Home bias and international diversification among individual investors. Journal of Banking and Finance 31(2), Karolyi, G. A. and R. M. Stulz (2003), Are financial assets priced locally or globally? In: G. M. Constantinides, M. Harris, and R. M. Stulz (eds.): Handbook of the Economics of Finance. Elsevier, pp Massa, M. and D. Schumacher (2015), Information barriers in global markets: Evidence from international subcontracting relationships. Working Paper. Morse, A. and S. Shive (2011), Patriotism in your portfolio. Journal of Financial Markets 14(2), Norden, L. (2010), Individual home bias, portfolio churning and performance, European Journal of Finance 16(4), Patel, S. and S. Sarkissian (2014), To group or mot to group? Evidence from CRSP, Morningstar Principia, and Morningstar Direct Mutual Fund Databases. Working Paper. Pool, V.K., Stoffman, N., Yonker, S.E. (2012), No place like home: familiarity in mutual fund manager portfolio choice. Review of Financial Studies 25(8), Tesar, L. L. and I. M. Werner (1995), Home bias and high turnover. Journal of International Money and Finance 14(4),

21 Van Nieuwerburgh, S. and L. Veldkamp (2009), Information immobility and the home bias puzzle. Journal of Finance 64(3),

22 60% Figure 1. Panel A. Portfolio Weight on US Stocks % 40% 30% 20% 10% 0% US in World Portfolio Median Fund US Holdings Median Fund US Holdings (25 Funds Exist Through ) 22

23 Frequency Figure 1. Panel B. Portfolio Weight on US Stocks by Global Funds % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Portfolio weight on US stocks 23

24 Table 1 This table shows the summary statistics of the funds. Num_fund shows the total number of global funds in our sample. Num_stock shows the number of stocks held by each fund-portfolio report date. Portweight_US is the portfolio weight of US stocks. Num_country is the number of countries held by the funds. TNA is total net asset of the funds in 2014 value. MRet is the monthly (next of fee) return. MExp is the expense ratio and Mflow is the monthly fund flows (in %). Panel B displays the fund manager variables. We first find managers working in the fund for each half year, and define the dummy variables such as CFA, MBA degree etc. Then we calculate the average of all fund managers if it is managed by a team. Num_Mgr is the average number of managers for each fund-portfolio date. Pcnt_Foreignedu is the percentage of managers (or, the average of the dummy variable of each manager) that have education background (undergraduate or graduate) outside US. Pcnt_Female, Pcnt_CFA and Pcnt_MBA are the percentage of managers who are female, with a CFA charter, a MBA degree, respectively. Mean Median 10th Percentile 90th Percentile Panel A: Fund variables Num_Fund 320 Num_Portfolios 4161 Num_Stock Num_Country Portweight_US Mret (%) TNA (Mil, in 2014 $) MExp (%) Mflow (%) Turnover (%) FamTNA (Mil, in 2014 $) Panel B: Fund manager Variables Num_Fund 315 Num_Portfolios 4074 Total_number_mgr 1312 Num_Mgr Pcnt_Foreignedu Pcnt_Female Pcnt_CFA Pcnt_MBA

25 Table 2 This table displays the fund portfolio performance when funds are sorted into groups based on portfolio weight on US stocks. In each half year, we sort funds into quintiles based on portfolio weight on US stocks. Quintile 1 (5) represent the quintile with low (high) investment in US stocks. We then measure the fund performance of the next 6 months. Fund net returns are obtained from CRSP. Average net and gross returns of the fund quintile portfolios are reported, where gross returns are expense ratio added back to net returns. We also report the risk-adjusted returns of the fund portfolios using Fama-French US and global three- and five-factor models plus US and global momentum factor, resulting in the adjustment with US/global four- and six-factor models. We report t-value in parenthesis and ***, ** and * represents 1%, 5% and 10% significance level, respectively. Net return Gross return US 4-factor Global 4- factor US 6-factor Global 6- factor 1 - low (1.48) (1.80) (0.41) (0.33) (0.43) (-0.74) (1.43) (1.63) (0.01) (0.11) (0.24) (-0.17) (1.11) (1.28) (-0.83) (-0.91) (-0.36) (-0.76) (1.12) (1.34) (-0.74) (-0.78) (0.14) (0.16) 5 - high (1.40) (1.67) (0.29) (1.01) (0.61) (1.64) Difference (1-5) ** (0.05) (0.11) (0.26) (-0.71) (-0.05) (-2.27) Difference (1-3) * (1.06) (1.50) (1.77) (1.14) (1.14) (0.21) 25

26 Table 3 This table reports the factor loadings of fund portfolio returns on US and Global Fama-French five-factor and momentum factor. The dependent variable is the time-series returns of fund portfolios where the portfolios are grouped by the holdings of US stocks. The independent variables are the US and global Fama-French five factors plus momentum factor downloaded from Professor Ken French website. The factors are SMB (small minus big), HML (high minus low B/M), RMW (robust minus weak operating profit), CMA (conservative minus aggressive Investment) and WML (winner minus loser). We report t-value in parenthesis and ***, ** and * represents 1%, 5% and 10% significance level, respectively. 1 - low high Difference(1-5) Difference(1-3) Panel A: US factors Mkt_RF 0.773*** 0.789*** 0.817*** 0.816*** 0.798*** ** (24.13) (23.69) (25.93) (26.02) (32.46) (-1.10) (-2.02) SMB ** ** ** (-0.33) (2.33) (1.31) (2.24) (1.21) (-1.78) (-2.36) HML 0.136** 0.137** 0.101* *** (2.52) (2.45) (1.90) (1.35) (2.81) (0.51) (0.95) RMW * *** ** (-0.11) (-0.95) (-1.83) (-3.17) (-1.04) (0.97) (2.46) CMA (-0.21) (-0.11) (0.23) (-0.49) (-0.75) (0.53) (-0.63) WML *** 0.091*** ** (1.44) (1.18) (3.06) (4.16) (0.49) (1.52) (-2.28) Adj-R

27 Panel B: Global factors Mkt_RF 0.836*** 0.853*** 0.862*** 0.845*** 0.771*** 0.065*** (46.76) (40.13) (36.42) (36.41) (34.16) (2.94) (-1.13) SMB 0.089*** 0.214*** 0.090** 0.203*** (2.62) (5.31) (2.00) (4.60) (0.52) (1.59) (-0.02) HML (0.23) (1.10) (-0.12) (0.22) (1.45) (-1.30) (0.29) RMW 0.166*** *** ** 0.286*** 0.227*** (3.69) (1.07) (-1.02) (-3.81) (-2.11) (5.14) (3.84) CMA 0.105* *** (1.90) (0.28) (0.77) (-0.83) (-1.43) (3.00) (0.67) WML *** 0.129*** *** (1.10) (1.39) (4.20) (6.04) (0.32) (0.57) (-3.40) Adj-R

28 Table 4 This table reports the regression result of the determinants of fund level home bias. The dependent variable is funds' portfolio weight on US stocks. We take the last portfolio report date for each calendar year half for each fund. The independent variables include fund manager characteristics and fund level variables, which are taken in the six-month window before the portfolio report date. Pcnt_foreignedu is the key variable and represents the proportion of fund managers with education background (undergraduate or post-graduate) outside US; Num_mgr represents the number of fund managers; pcnt_cfa, pcnt_mba and pcnt_female represents the percentage of fund managers who are female manager, with CFA charter and MBA degree respectively. Fund level variables are the monthly average values of the fund characters in the six-month window. The variables include log total net asset (lg_tna), monthly fund return (Mret), expense ratio (MExp) and log family TNA (lg_famtna). The panel regressions adjust fund fixed effect. (1) (2) (3) (4) Pcnt_Foreignedu *** *** *** *** (-2.99) (-2.94) (-2.70) (-2.75) Num_Mgr (-0.56) (-0.15) (0.27) Pcnt_Female *** *** *** (2.99) (2.77) (2.59) Pcnt_CFA ** ** * (2.41) (2.38) (1.70) Pcnt_MBA *** *** *** (-3.59) (-2.89) (-3.79) Lg_TNA (0.02) (0.02) Mret *** *** (-5.17) (-4.48) MExp (0.51) (0.36) Lg_FamTNA * (-1.65) Fixed effect Yes Yes Yes Yes Num. of Obs R

29 Table 5 This table reports the panel regression results on the determinants of fund performance. In each year halve, we obtain fund variables including portfolio weight on US holdings, log of total net asset, expense ratio, fund flow (percentage) and turnover ratio as well as manager characteristics including number of manager and proportion of managers with foreign education background, female managers, CFA, and MBA. We obtain the fund performance in the following six months of each year halve. The performance is calculated as fund excess net return subtracting the product of six global factors (Fama-French global five-factor plus the global momentum factor) and the loading of fund on these factors. The loadings are estimated using previous 36 months returns. We adjust standard errors on both fund level and time level clustering. ***, **, and * represent 1%, 5% and 10% significance level. (1) (2) (3) (4) Intercept (-1.07) (0.63) (-0.95) (-0.88) Portwt_US * (1.73) (1.49) (1.56) Lg_TNA (0.63) (0.69) MExp (-0.16) (-0.12) Mflow (1.44) (1.46) Turnover (0.47) (0.63) Pcnt_Forgnedu ** * * (-2.13) (-1.79) (-1.88) Num_Mgr (-0.28) (-0.03) (-0.14) Pcnt_Female ** ** ** (-2.14) (-2.01) (-2.05) Pcnt_CFA (1.46) (1.57) (1.45) Pcnt_MBA (-0.26) (-0.39) (-0.46) Num. of Obs R

30 Table 6 This table shows the flight home of US global funds during financial crisis. We compare the fund level home bias before financial crisis ( ), and during financial crisis ( ). Fund level home bias is defined as the difference between fund's portfolio weight on US stocks and US market weigh in world market at the end of the year. The change of this fund level home bias from before financial crisis to during financial crisis is a measure of flight home. US market weigh in world market is calculated from Datastream for 51 global markets. For each fund, average value of this home bias measure is calculated for years 2005 and 2006, and 2007 and There are 95 funds that exist in both of these two periods. Flight home from 2006 to 2007 # of funds Mean HB_ HB_ difference 0.033*** p-value <.0001 Flight home from to HB_ HB_ difference 0.045*** p-value <.0001 Flight home by funds with positive / negative home bias before crisis (2006) HB_Positive HB_Negative difference p-value