Mutual fund home bias and market uncertainty

Transcription

1 Mutual fund home bias and market uncertainty Nicole Choi a and Hilla Skiba b This draft: May 2014 Abstract: We study the relationship between market uncertainty and mutual fund home bias across 49 countries. We assert that market uncertainty exists over time and across countries. We predict that (1) market uncertainty is positively related to home bias, (2) home bias is positively related to fund performance, and (3) the positive relationship is stronger when market uncertainty is high. Results provide support for empirical predictions. Mutual funds from countries with more uncertainty have higher home bias in their portfolios and home bias is positively related to performance of mutual funds, especially when the home country is characterized by high uncertainty. Keywords: Home bias; Market uncertainty; Flight home; Mutual funds JEL codes: G11; G15; G23; Z10 a Nicole Choi, University of Wyoming, nchoi@uwyo.edu b Hilla Skiba, Univesrity of Wyoming, hskiba@uwyo.edu

2 Introduction International under-diversification is an important topic in finance literature. The home bias phenomenon has been documented empirically in investors portfolios since seminal work by French and Poterba (1991). The authors show that investors allocate disproportionately small amounts of equity outside their home markets, relative to the market capitalization weights, so that the investors home markets, on average, carry more weight in their portfolios. In other words, investors are under-diversified, even though traditional portfolio theory suggests large gains from international diversification (Grubel 1968; Levy and Sarnat 1970). Common explanations for international under-diversification include market segmentation, familiarity bias, and information barriers (Hamberg, Mavruk, and Sjögren 2013; Huberman 2001; Eichler 2012; Coval and Moskowitz 1999; Chan, Covrig, and Ng 2005; Grinblatt and Keloharju 2001, among others). The main contribution of this paper is to test empirically if the degree of home bias is related to market uncertainty faced by investors. The relationship between uncertainty and home bias is motivated by theory in Van Nieuwerburgh and Veldkamp (2009). When investors face uncertainty in asset allocation, information about home market assets is easier to access. Information about home market assets reduces related uncertainty, which in turn increases the home market assets risk-reward ratio. The resulting portfolios can therefore be home biased but mean-variance efficient. In the first part of our study, we conjecture that investors face varying levels of uncertainty in the global asset markets. We assert that market uncertainty can vary over time and across countries. We predict that during time periods of heightened market uncertainty, home bias is magnified because learning about foreign securities becomes more difficult in general. Similarly, we predict that learning about securities in a country that has high barriers to 2

3 information, and is more uncertain as a result, is more difficult. As a consequence, foreign investors are less willing to invest and they underweight the uncertain country, which in turn increases the domestic overweight of the securities. In the second part of the paper, we test if home bias in portfolios is positively related to performance. Existing literature on the topic is inconclusive, and several papers have shown evidence for positive, neutral, and negative effects of home bias on performance (Giannetti and Laeven 2014; Coval and Moskowitz 2001; Choi et al. 2014, among others). We contribute to the literature on consequences of home bias on performance by testing if home bias is more beneficial to the investor when the market uncertainty is high. This hypothesis is motivated by the conclusion of Van Nieuwerburgh and Veldkamp (2009) that home biased portfolios are mean-variance efficient. We posit that if specialization in home market securities enhances performance, then this should be especially true when the foreign investors face high uncertainty in the investors home country due to either time-varying heightened market uncertainty or due to home country-specific information barriers. The empirical prediction is then that the interaction between market uncertainty and home bias is positively related to performance. Our analysis focuses on international under-diversification in a large sample of mutual funds from 2002 to We use a comprehensive dataset of mutual funds holdings and performance to test our hypotheses. The mutual funds are from 49 markets and we have data on their detailed portfolio holdings on a quarterly basis. The holdings data allow us also to compute performance measures to mutual funds portfolios and to their home market holdings, to test if home country concentration combined with market uncertainty enhances performance. The time period of the study spans nine years from the third quarter of 2002 to the last quarter of The time span captures the global financial crisis from 2007 through 2009 as a natural 3

4 experiment window for extreme global uncertainty that is time-varying. We also employ other time-varying uncertainty measures effects on home bias and performance. These measures include the US VIX index and value-weighted regional VIX-like indexes, that are constructed from several regional implied volatility indexes. Many other recent papers have shown the importance of financial crisis and volatility on global equity markets and investor behavior (for example, Forbes and Warnock 2012; Milesi-Ferretti and Tille 2011; Guiso et al. 2011; Giannetti and Laeven 2012). We contribute to this important literature by documenting the effect of market turmoil on professional investors portfolios and performance. The results support most of our hypotheses. First, we show that mutual fund investors home bias increases during heightened market uncertainty. We also show that if a mutual fund s home country is characterized by high information barriers (i.e., is more uncertain), then the home bias of the mutual fund is higher compared with mutual funds from less uncertain home countries. Second, we show that home bias has a neutral to a positive effect on mutual funds overall portfolio performance and on performance of mutual funds home market securities. Third, we show that most variables in our study that proxy for uncertainty have a negative effect on mutual funds performance. Both the time-varying and country-specific market uncertainty deteriorate performance. Fourth, we find evidence that the interaction of market uncertainty and home bias is positively related to performance. In other words, home bias has a more positive effect on performance during periods of heightened uncertainty and when investors home country is more uncertain. This is especially true for the country-specific uncertainty proxies, although we show that during the financial crisis period, mutual funds home bias is positively related to performance. The result on home bias and performance complements the result of 4

5 Giannetti and Laeven (2014) who study US mutual funds local bias and performance during financial crisis. This paper makes several contributions to the literature regarding the international underdiversification puzzle. First, we utilize a rich, international sample of mutual funds holdings and performance to study causes and consequences of under-diversification. Second, we document a statistically and economically significant relationship between market uncertainty and mutual funds home bias. Market uncertainty is measured both over time and across countries. Especially the time-varying component of home bias has not been widely studied in the prior literature. Third, we find that market uncertainty has a negative effect on mutual funds performance, but investors can overcome some of the negative effect by focusing their holdings in home market securities. Overall, evidence presented supports the theoretical foundation of rational under-diversification that is a result of costly information. Portfolios that are tilted toward home market securities are not necessarily mean-variance inefficient. The rest of the paper is organized as follows: Section II reviews related literature and the testable hypotheses; Section III explains the data and methodology; Section IV reviews the results from empirical analysis; and Section V concludes. II. Related Literature and Testable Hypotheses A. Home Bias in Portfolios Several finance papers document that investors have a preference for home market securities. Home bias has been shown to matter both at the country level and in local markets. In other words, investors allocate larger than expected shares to their home country securities as well as to geographically proximate securities inside their home country. 5

6 In an international setting, home bias has been documented at the aggregate country level by French and Poterba (1991), who show that the average home bias is positive in all countries in their sample. Chan, Covrig, and Ng (2005) document home bias in a large sample of global mutual funds and show that home bias is positive in all countries and driven by capital controls, stock market development, and familiarity bias. Coval and Moskowitz (1999) document local bias in US mutual funds domestic holdings. Similarly, Huberman (2001) shows that US investors exhibit a strong familiarity bias in their domestic holdings, which results in regionally under-diversified portfolios. Nofsinger and Varma (2012) document individual investors preference for local utility stocks over utility companies located outside their state of residence. The authors find the preference for local utility stocks is driven by familiarity bias and even sophisticated and affluent investors are subject to the bias. Relatively few papers have studied time-varying characteristics of home bias and international under-diversification of investors portfolios, often because of data limitations. However, some important explanations for home bias can be discerned only based on time-series variation. For example, advances in technology and increased availability of international investment vehicles have made information easier to access. Because of this we would expect under-diversification have reduced since early empirical papers. Giannetti and Laeven (2014) are among the first to document that there is a significant relationship over time between home bias and stock market conditions. The authors show that the local bias in US mutual funds domestic holdings, first documented by Coval and Moskowitz (1999), varies over time, so that during times of increased market uncertainty, the local bias in the US investor portfolios increases significantly. Giannetti and Laeven (2014) also document that the managers of locally concentrated portfolios earn superior returns to local stocks 6

7 compared to distant stocks. The authors attribute this finding to an information advantage, arguing that during periods of heightened market uncertainty, information barriers increase, and a local presence provides managers with a larger advantage compared to more certain times. B. Market Uncertainty and Investor Decision Making The effect of uncertainty on economic decisions is well-documented in finance literature. In his seminal paper, Ellsberg (1961) shows that investors are ambiguity averse and choose to bet on a certain lottery over an uncertain one. Many recent papers have studied the effect of time-varying market uncertainty, specifically the effect of the financial crisis of , on asset allocation. Most papers find that during market uncertainty investors have a tendency to rebalance their portfolios toward certainty. In addition to Giannetti and Laeven s (2014) mutual fund paper discussed in the previous section, Giannetti and Laeven (2012) document a flight home effect in the market of syndicated loans, where lenders favor domestic borrowers during financial crisis. Vermeulen (2013) shows that investors rebalance their portfolios toward less correlated markets during the financial crisis. Milesi-Ferretti and Tille (2011) show that during financial crisis, global capital and trade flows collapse and that the collapse in capital flows is much more dramatic and shorter-lived in emerging markets. Forbes and Warnock (2012) also analyze international capital flows and show that global risk is associated with extreme capital flow episodes, and that domestic macroeconomic characteristics are less important in explaining periods of surges and stops. Beber, Brandt, and Kavajecz (2008) document that investors hold high quality and liquid bonds during crisis periods, and as a result there exist both flight to liquidity and flight to quality during the crisis period in the bond markets. Country-specific uncertainty has also been shown to matter in asset allocation decisions. Anderson et al. (2011) and Beugelsdijk and Frijns (2010) show that if investors home markets 7

8 are culturally characterized by uncertainty-avoiding investors, then those investors allocate less capital in foreign markets. Leuz, Lins, and Warnock (2010) find that foreigners invest less in countries that are characterized by many of the country-specific uncertainty proxies we use in this paper. Specifically, the authors find that foreigners underweight countries where outsider protection and disclosure are poor. Eichler (2012) shows that US investors allocate more capital toward foreign markets with more comprehensive disclosure. Thapa and Poshakwale (2012) show that foreigners prefer markets that are more liquid, efficient and have lower trading costs. At the security-level, in a single market study, Kang and Stulz (1997) show that foreigners investing in Japan show a preference for transparent firms. C. Home Bias and Performance In a continuing debate on why international under-diversification takes place, some researchers argue that proximity to assets increases investors information advantage regarding those assets and that investors rationally overweight home market securities and proximate foreign markets. Other researchers argue that under-diversification is irrational and driven by familiarity bias. The evidence on whether familiarity bias or information advantage truly drives portfolio allocation decisions is inconclusive at best. For example, Coval and Moskowitz (2001) show that locally concentrated portfolios outperform diversified portfolios in the US. Giannetti and Laeven (2014) show that market uncertainty increases the positive effect that geographic concentration has on US mutual funds performance. Ivkovic and Weisbenner (2005) find that individual investors also overweight local stocks as well as generate positive abnormal performance from the local holdings. All of these studies, however, focus on one market and the effect of local bias on performance. 8

9 On an international scale, Choi et al. (2014) show the opposite. Home market does not appear to provide an advantage to investors on a large scale in a global setting, and home bias and country overweights in portfolios do not enhance performance in home or foreign securities on a risk-adjusted basis. Similarly, Fedenia et al. (2013) show that foreign investors country overweight does not improve performance in that country s securities. It is thus still debatable whether home biased asset allocation is driven by familiarity bias or rational behavior motivated by costly information barriers, when an investor chooses to specialize in a narrower set of highly correlated assets as Van Nieuwerburgh and Veldkamp s (2009) theoretical model suggests. D. Hypotheses The theoretical motivation for our study is provided by Van Nieuwerburgh and Veldkamp (2009), who show that home bias can be an optimal choice in the portfolio. Their theory is in turn motivated by Merton (1987), who posits that information imperfections in asset markets lead investors to focus their holdings. According to Van Nieuwerburgh and Veldkamp (2009), barriers to information result in investments in a small set of highly correlated assets. Investors face uncertainty and barriers to information in asset allocation, especially outside their home markets. Since information about home market assets is easier to access, home market assets are less uncertain to the investor. In other words, the same asset can have a different risk-reward ratio to the home market and to the foreign investors. As a result, investors may overweight home market assets, but still have a mean-variance efficient portfolio. Based on the Van Nieuwerburgh and Veldkamp s (2009) result, we expect to observe different levels of home bias depending on the degree of market uncertainty faced by the investors. If market uncertainty drives home bias, then we expect to observe high levels of home bias if the investor is faced with high uncertainty. Formally, we test the following hypothesis: 9

10 H1: Market uncertainty is positively related to home bias As discussed in the previous sections, we believe that market uncertainty can be both time-varying and country-specific. The time-varying uncertainty is assumed to be high during turbulent, volatile markets. As a consequence, we expect to observe high levels of home bias in investors portfolios during heightened market uncertainty. The logic and analysis in our paper with respect to time-varying home bias is similar to that presented in Giannetti and Laeven (2014), where the authors study US-based mutual funds time-varying local concentration. We test if home bias (as opposed to local bias) of the funds holdings increases during heightened uncertainty. Market uncertainty can also be country-specific. When foreign investors consider their international allocation decisions, certain countries have higher barriers to information and are less transparent. In other words, some countries are more uncertain to foreign investors than others. In a case of an uncertain country environment, we expect there to be lower levels of investment by foreigners and higher ownership by local investors. Thus, we expect to observe a difference in home bias across countries, and country-specific uncertainty proxies that are positively related to home bias. Throughout the study, we measure home bias as the difference between the investors portfolio weight of home market holdings and the actual home market capitalization weight. The measure will be explained in more detail in the methods section. In the second part of the paper, we test if home bias enhances investors performance, and if home bias enhances performance more when market uncertainty is high. According to Van Nieuwerburgh and Veldkamp (2009), home biased portfolios may be optimal. If investors are home biased, they have more incentive to learn about home market assets and the gap in 10

11 information between domestic and foreign investors increases. Therefore, performance should not only be enhanced by home bias, but also be enhanced more by home bias when the market uncertainty is high. Formally, we test the following hypotheses: H2: Home bias enhances investors performance H3: Home bias enhances investors performance when market uncertainty is high We test H2 in two different ways. First, we test if home bias improves performance, while holding any country variation constant in the analysis. Second, we test if home bias improves performance, while controlling for time. To test H3, we interact the market uncertainty proxies with the home bias measure, and expect to observe a positive relationship between high uncertainty and home bias on performance. Again, we repeat the analysis with time-varying and country-specific uncertainty measures. III. Data and Methodology A. Data We combine several datasets to study the allocation and performance of mutual funds over time. The mutual fund holdings data are from FactSet Company, available through Wharton Research Data Services. These holdings data are among the most extensive global mutual fund holdings data that we are aware of. In this study, we limit ourselves to mutual funds that hold both foreign and domestic securities. We have quarterly holdings data for 7,629 mutual funds at the portfolio level. The holdings data used in the study are from the first quarter of 2002 to the fourth quarter of This time span allows us to investigate global market conditions from relatively stable performance in the early part of the sample, to the financial crisis and heightened market uncertainty period during , to the recovery period of The mutual funds in the sample are located in 49 different home markets, in both the developed and emerging world. 11

12 We match the mutual funds holdings to security-level data to construct measures of home bias and performance for the portfolios. The security performance data are also from FactSet. Both the holdings data and the security data report the country of the funds and stocks headquarters. We define holdings as domestic if the fund and the stock share the same country of domicile. To compute portfolio home bias, we compare mutual funds actual allocation to home countries to the home countries expected allocation. The expected allocation is computed based on WorldScope s data on publicly traded firms total shares outstanding, the float shares outstanding (non-privately held shares), and the market price of the securities in USD. This allows us to compute expected investment weights for securities and countries, based on both the total market capitalization as well as the float capitalization. We describe the computation of the expected weights and home bias in more detail in Section B. The purpose of this paper is to investigate whether market uncertainty over time and across countries is related to home bias and performance of the funds. The time-varying uncertainty is first measured by an indicator variable of the global financial crisis for the period from the fourth quarter of 2007 to the second quarter of We also collect several indexes that measure global volatility over time. The detailed construction of volatility measures is described in Section B.2. All volatility measures are from Bloomberg. To test if country-specific uncertainty affects home bias of portfolios, we include several country-specific uncertainty proxies in the regressions. These proxies include: the Anti-selfdealing index from Djankov et al. (2008), CIFAR from Bushman, Piotroski, and Smith (2004), Internet users per 100 people from the World Bank, the index of Economic freedom, and some of its relevant components (Business freedom, Investment freedom, and Freedom from corruption) 12

13 from World heritage Foundation, and Trade openness, Market development, and GDP per capita from the World Bank. Table 1 shows the main country-specific variables used in the study to proxy for mutual funds home markets uncertainty. Table 1 also reports the number of mutual funds in each market and the average home bias in each market. Reported numbers for time-varying variables are time-series averaged. B. Methodology B.1. Home Bias Home bias of each mutual fund is computed based on the actual weight of the mutual fund s home market holdings relative to the home market s weight as a share of the total world portfolio. The first term of equation (1) is the actual home market weight of the mutual fund s portfolio, and it is computed based on the capitalization of all securities in mutual fund i s portfolio that has headquarters in home market I in quarter t, scaled by the total value of the mutual fund s portfolio in quarter t in all J markets. The second term of equation (1) is the expected allocation to the home market. It is computed based on the market value of all securities in market I in quarter t scaled by the total market value of the countries in our sample in quarter t. More formally, the home bias measure, HB, for each fund i in home market I in quarter t is HB TMV TMV i I,t I, t i I, t J J TMV TMV i I, t I, t I 1 I 1. (1) We compute the expected weights of each market based on both the total number of firms shares outstanding as well as the float number of shares outstanding, which excludes all closely 13

14 held shares that cannot be accessed, so that those firms with fewer investable shares have relatively lower expected investment values. All analyses are run with both measures, but only reported for the float-based weights. Table 1 shows summary statistics for the sample of 7,629 mutual funds. The second column of Table 1 shows the time series average home bias, HB, for each country from equation (1). There is a large variation in the home bias across countries. The home bias measure, which is positive for all countries in the sample, ranges from low values for Namibia, Ireland, Belgium, and the Czech Republic (0.02, 0.03, 0.03, and 0.06, respectively) 1 to high values for Brazil, the Philippines, Turkey, and Pakistan (0.98, 1.00, 1.00, and 1.00, respectively). 2 B.2. Time-varying Uncertainty Measures We measure time-varying uncertainty in two ways. The first measure is an indicator variable for the financial crisis period. This indicator variable is equal to one from the third quarter of 2007 to the second quarter of 2009, when most of the market turmoil takes place. The second measure is motivated by Laeven and Giannetti (2014), Forbes and Warnock (2012), and Guiso et al. (2011), among others, who measure time-varying market uncertainty for the US market and global market with the VIX index. Since our paper focuses on global uncertainty, we use several regional VIX-like indexes in addition to the US. VIX to capture global volatility. The volatility indexes include: ASX of Australia, Hang Seng s HSI Volatility Index, India VIX, Implied Volatility Index of Taiwan, AEX for the Netherlands, BEL 20 Volatility Index for Belgium, CAC 40 Volatility Index for France, FTSE 100 Volatility Index for the United Kingdom, VDAX Volatility Index for Germany, and the CBOE s VIX for the United 1 Note that some of the countries with very low levels of home bias are markets where many institutions are headquartered but not necessarily present. We exclude these markets in robustness checks. 2 All investors in our sample have at least some foreign holdings and the home markets have some value. In some cases these foreign holdings are very small and the home bias is rounded to 100%. Also note that some countries restrict foreign investment by institutions. We also exclude these markets in robustness checks. 14

15 States. We also compute one comprehensive Global Volatility Index based on the country volatility indexes. The Global Volatility Index is computed by first standardizing each regional index and then computing equal- and value-weighted averages of the standardized indexes for each time period. Figure 1 shows the volatility measures during our sample period and also highlights the financial crisis. It is clear from the figure that the volatility measures are highly correlated and especially high in the early quarters of the sample as well as during the financial crisis. B.3. Country-specific Uncertainty Index We collect a large number of country-specific uncertainty measures. The first captures accounting transparency. This variable, CIFAR, is introduced by Bushman, Piotroski, and Smith (2004) and is meant to capture transparency of firms to external market participants. CIFAR is an index that is developed for each country in the sample, based on the transparency of firms financial statements. It considers 90 items from firms annual reports, and represents the average number of these items for each country, so that a high score indicates high transparency in the market. The second measure of country-specific uncertainty is the Anti-self-dealing index from Djankov et al. (2008). The index variable measures shareholder protection in different countries. It is constructed based on a large set of underlying components that proxy for countries legal environments. This index captures private enforcement mechanisms, such as disclosure and litigation of self-dealing transactions. The index takes on high values for countries with high levels of shareholder protection and is typically higher in common law countries compared with civil law countries. 15

16 The third measure of country-specific uncertainty is internet users per 100 people. This variable is from the World Bank. We hypothesize that markets with high levels of internet users will be more transparent and information about firms can be accessed more easily by foreign investors. The fourth measure of country-specific uncertainty is related to countries economic freedom. The index of Economic freedom is constructed by Heritage Foundation. 3 The index value is higher for countries characterized by more economic freedom and opportunity. In some analyses we also use components of the index that most directly relate to stock markets, for example the Business freedom index, Investment freedom index, and Freedom from corruption index. We also include variables describing macroeconomic characteristics of the countries that are related to uncertainty. These macroeconomic variables include the GDP per capita, total trade scaled by GDP (export and imports), and stock market development, measured as the market capitalization scaled by GDP. These variables are also from the World Bank. We assume that all of these country characteristics are positively related to transparency and negatively related to uncertainty. From these uncertainty proxies we generate two uncertainty indexes. First is the Uncertainty index, which is the summation of uncertainty variables standardized values, after which we standardize the index itself. The second index for uncertainty includes the first and the second principal components (PC 1 and PC 2) generated from the uncertainty variables. 3 Index of Economic Freedom from the Heritage Foundation: 16

17 B.4. Performance Measures To test hypotheses 2 and 3, relating to whether home market concentration provides investors with an information advantage when uncertainty is high, we test if the home bias measure is positively related to overall portfolio performance of the mutual fund. To construct a measure of performance from holdings data, we compute three-month returns to the mutual fund s holdings after they are reported. To measure a fund s performance, we first compute returns to each of the securities in the portfolio, and then value-weight the returns to portfolio level based on the total value of the mutual fund s portfolio. More formally, the value weighted excess return to the mutual fund i s portfolio at quarter t is, (2) R w R R i, t i, j, t j, t Jt j J where R j,t is a quarterly return of a security j that is headquartered in a country J, w i,j,t is fund i s weight on security j at quarter t, and R Jt, is equally-weighted return index of all available securities in FactSet for country J. In some tests, we compute excess return to funds home securities in excess of home market s benchmark index by summing weighted returns of securities domiciled in funds home countries minus funds home country s equally-weighted market index. In most specifications, we also adjust the excess return to the risk of the securities used in the return calculation. The risk adjustment is such that we compute a beta for each security in the portfolio based on three-year returns leading up to the reporting period. Formally, the portfolio level risk-adjusted excess return is R adj,, i t it, (3) j J R w i, j, t j, t where R i,t is from equation (2) and jt, is beta of security j. 17

18 IV. Results A. Time-varying Uncertainty and Home Bias In this section, we test hypothesis 1 regarding whether home bias increases as market uncertainty increases. First, we test if time-varying uncertainty increases home bias. In this analysis, the dependent variable is home bias from equation (1) and the independent variables include several measures of time-varying uncertainty. These time-varying uncertainty measures are the Financial crisis indicator and several volatility measures that include: the US VIX, Euro VIX, and value- and equally-weighted Global VIX. We expect all of these volatility measures to take on positive signs in the regressions. Table 2 shows results from OLS regressions where the dependent variable is mutual fund s portfolio home bias from equation (1) and the independent variables include the timevarying uncertainty measures. We also control for the mutual fund s market value, MF MV, (logarithm of USD market value of equity) and include country fixed effects. The results in Table 2 support hypothesis 1 with respect to the financial crisis indicator. The Financial crisis indicator is positive and significant, which indicates during the financial crisis, the mutual fund s home bias increases. However, the volatility measures are not positive as expected, and some of them are actually negative and significant, indicating a decrease in home bias when the volatility is high. This result contradicts the result from Giannetti and Laeven (2014), who find that local concentration of US mutual funds US holdings increased with high VIX during the 2000s. Overall, the results of Table 2 provide weak evidence at best for hypothesis 1 that mutual fund s home bias increases during a period of heightened market uncertainty. 18

19 B. Country-specific Uncertainty and Home Bias Next, we test hypothesis 1 in a country-specific uncertainty analysis. Again, the expectation is that as country-specific uncertainty increases, home bias increases as well. In this analysis, the dependent variable is again home bias from equation (1). In Table 3, Panel A, the first set of independent variables includes the Anti-self-dealing index, which captures the country s legal strength, CIFAR which captures the accounting transparency, and Internet which is a number of internet users per 100 people. All these variables decrease as uncertainty increases, thus they are expected to take on negative signs. The second set of independent variables includes the index of Economic freedom, and some of its relevant components. These components include indexes for Business freedom, Investment freedom, and Freedom from corruption. All the indexes increase with more freedom and are expected to be larger values for less uncertain markets. The third set of independent variables includes macroeconomic variables that proxy for uncertainty. These variables include: Trade openness ((Imports+exports)/GDP), Market development (Market cap/gdp), and GDP per capita. All these are expected to have higher values for less uncertain countries and to take on negative signs in the regressions. In Table 3, Panel B, we proxy for country-specific uncertainty by the uncertainty indexes described in Section III B.3. Uncertainty index is the standardized index of the summation of all uncertainty variables standardized values. PC 1 and PC 2 are the first and the second principle components generated from the uncertainty variables. These uncertainty indexes also increase as the uncertainty decreases and are expected to take on negative signs, consistent with the empirical prediction of hypothesis 1. Results presented in Table 3 provide support for hypothesis 1. In Panel A, all but one of the uncertainty proxies have negative and significant signs, so that mutual funds from countries with more (less) uncertain investment environments have higher (lower) levels of home bias. 19

20 Also, the GDP per capita measure is negative and significant, as expected. The economic significance of the uncertainty variables is also large. A one standard deviation increase in the proxies reduces the observed home bias from the low value in Anti-self-dealing index of 1.5% to the high value in GDP per capita (when included without other uncertainty proxies) of 16%. In Table 3, Panel B, we also find strong support for hypothesis 1, when we capture country-specific uncertainty with Uncertainty index, PC 1 and PC 2. All of the indexes are negative and statistically and economically significant in all specifications. This is true whether we control for the country s GDP per capita or not. The economic significance of the indexes is large. A one standard deviation increases in the Uncertainty index, PC 1, and PC 2 decrease the observed home bias by 16%, 11%, and 8%, respectively. Overall, the results of Table 3 provide strong support for hypothesis 1 that market uncertainty, when measured based on country-specific uncertainty variables, is positively related to the magnitude of mutual funds home bias. C. Home Bias, Uncertainty, and Performance Next, we test hypotheses 2 and 3 regarding whether home bias enhances performance, and whether home bias enhances performance more when the market uncertainty is high. Again, we measure market uncertainty based on both time-varying and country-specific uncertainty proxies. The empirical predictions are that home bias is positively related to performance and that performance is enhanced at a greater degree when the market uncertainty is higher. In other words, specialization in home market securities is more beneficial when interpreting local information becomes more difficult for foreign investors. Table 4 shows results from analysis that tests the effect of time-varying uncertainty and home bias on performance. We test the effect of time-varying uncertainty and home bias on 20

21 performance of the mutual fund s aggregate portfolio in excess of the world benchmark index in Panel A, and in the home market securities of the mutual fund in excess of the home market index in Panel B. The dependent variables are risk-adjusted return of a mutual fund (equation 3) for specifications (1) to (4) and market-adjusted return (equation 2) for specification (5). The independent variables include the Home bias measure from equation (1) as well as the timevarying uncertainty measures from Table 2: the Financial crisis indicator, and several volatility indexes. In addition, we include the interaction terms between Home bias and the uncertainty variable used in each specification. Consistent with hypothesis 2 and 3, predicted signs are positive on the home bias variable and positive on the interaction term between the home bias and each uncertainty measure. In other words, we expect to observe a positive effect of home bias on performance that will be more positive when market uncertainty is high. The results presented in Table 4 provide weak support for hypotheses 2 and 3. First, Home bias is positive and marginally significant when included without the other controls in specification (1) of Panel B. In Panel A, Home bias is also positive but not statistically significant. The only interaction term that is positive and statistically significant is in specification (2) of Panel A, when time-varying uncertainty is measured by the Financial crisis indicator. The result indicates weak support for the assertion that home bias enhances mutual funds portfolio performance in excess of the global index during the financial crisis period. However, the same indicator is no longer significant in Panel B, when the dependent variable is the performance in home market securities in excess of the home market index. Overall, we do not find strong support for either hypothesis 2 or 3 when we analyze the effect of home bias and market uncertainty on performance over time. The result conflicts with findings of both Coval and Moskowitz s (2001) and Giannetti and Laeven s (2014) US-specific findings. First, we do 21

22 not find strong support that home bias enhances performance. Second, we do not find that mutual funds benefit from home market concentration during heightened market uncertainty. We also find that the financial crisis and market volatility have a negative and significant effect on mutual funds performance. Size of the mutual fund has either a neutral or positive effect on performance. Next, we test hypotheses 2 and 3 similar to the previous analysis, but we proxy for market uncertainty with country-specific measures. Tables 5 and 6 show results from analyses of the effect of country-specific uncertainty and home bias on risk-adjusted performance in their home market. In Table 5, the independent variables include the Home bias and several country-specific uncertainty measures. The uncertainty measures in Table 5 are the variables from Table 3 as well as interaction terms between Home bias and the uncertainty variable of each specification. The results of Table 5 provide strong support for both hypotheses 2 and 3. First, in support of hypothesis 2, Home bias is now a positive and statistically significant determinant of performance of mutual funds in their home market securities in excess of the home market s benchmark index. In both specifications (1) and (2) the coefficient of Home bias is positive and statistically significant. In economic terms, a one standard deviation increase in the mutual fund s home bias corresponds to about a 1% annual increase in the dependent variable, which is the return to the home market securities in excess of the benchmark and scaled by the beta of the home market holdings. Second, in support of hypothesis 3, we find that the interaction terms between Home bias and most of the uncertainty measures are negative and significant. The interaction terms are negative and significant when uncertainty is measured by the Anti-self-dealing index, CIFAR, 22

23 Investment freedom, and Internet. Overall, home bias has a positive effect on performance, but less so when the country has lower level of uncertainty, consistent with hypothesis 3. In Table 6, we show results from regressions that repeat the analysis of Table 5, but we measure country-specific uncertainty with the Uncertainty index variable, PC 1, and PC 2. First, we find that the uncertainty variables are not statistically significant determinants of performance on their own in either specification (1) or (3). But consistent with results of Table 5, the interaction term between Home bias and both Uncertainty index and PC 1 are negative and significant. Again, home bias appears to enhance performance of mutual funds in their home market securities in excess of the home market benchmark, but less so if the home market is characterized by lower uncertainty, consistent with hypothesis 3. Overall, the results from this section are similar to Sections A and B. The time-varying uncertainty measures do not support our hypotheses, but the country-specific uncertainty measures are mostly in line with the paper s empirical predictions. V. Conclusions This paper tests if market uncertainty is related to portfolio home bias. We assert that market uncertainty exists over time and across countries. Based on the theoretical finding by Van Nieuwerburgh and Veldkamp (2009), we predict that (1) market uncertainty is positively related to home bias; (2) that home bias is positively related to mutual funds performance; and (3) that home bias is positively related to mutual fund s performance especially when market uncertainty is high. We find support for our hypotheses mainly when we study market uncertainty that is country-specific. The main result of the paper is that when a country has high uncertainty, the home bias of its mutual fund investors increases. We also find that home bias of holdings 23

24 enhances the performance of mutual funds from countries that are characterized by high uncertainty by more compared to the performance of mutual funds from less uncertain home countries. These results are in line with the theoretical foundation from Van Nieuwerburgh and Veldkamp (2009), who assert that information barriers lead investors to learn about their home market assets and that specialization in a narrow set of assets results in better information, lower risk, and mean-variance efficient portfolios. We show empirically that because uncertainty across countries varies, investors from countries with high uncertainty benefit from specialization in home market assets more because the barriers to learn about the assets are higher for foreign investors in that market. Although we do not find support for our hypotheses with respect to the time-varying components of uncertainty, home bias, or performance, the results provide some interesting findings that can be explored in future research. Few papers so far have studied time-varying home bias. We show that instead of a decline in home bias, which is what we might expect with technological improvements in trading and increased transparency, we are more likely to observe a cyclical pattern of home bias in mutual funds portfolios. Our time-varying performance results also show that global mutual funds performance is worse during heightened market uncertainty. 24

25 Literature Anderson, Christopher W., Mark Fedenia, Mark Hirschey, and Hilla Skiba, Cultural influences on home bias and international diversification by institutional investors. Journal of Banking and Finance 35, Beber, Alessandro, Michael W. Brandt, and Kenneth A. Kavajecz, Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market. Review of Financial Studies 22, Beugelsdijk, Sjoerd, and Bart Frijns, A cultural explanation of the foreign bias in international asset allocation. Journal of Banking and Finance 34(9), Bushman, Robert M., Joseph D. Piotroski, and Abbie J. Smith, What determines corporate transparency? Journal of Accounting Research 42, Chan, Kalok, Vicentiu Covrig, and Lilian Ng, What determines the domestic bias and foreign bias? Evidence from mutual fund equity allocations worldwide. Journal of Finance 60, Choi, Nicole, Mark Fedenia, Hilla Skiba, and Tatyana Sokolyk, Information acquisition, international under-diversification and portfolio performance of institutional investors. Working paper, University of Wyoming. Coval, Joshua D., and Tobias J. Moskowitz, Home bias at home: Local equity preference in domestic portfolios. Journal of Finance 54(6), Coval, Joshua D., and Tobias J. Moskowitz, The geography of investment: Informed trading and asset prices. Journal of Political Economy 109, Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, The law and economics of self-dealing. Journal of Financial Economics 88(3), Eichler, Stefan, Equity home bias and corporate disclosure. Journal of International 25

26 Money and Finance 31, Ellsberg, Daniel, Risk, ambiguity, and the savage axioms. Quarterly Journal of Economics 75(4), Fedenia, Mark, Sherrill Shaffer, and Hilla Skiba, Information immobility, industry concentration, and institutional investors performance. Journal of Banking and Finance 37(6), Forbes, Kristin, and Francis Warnock, Capital flow waves: Surges, stops, flight, and retrenchment. Journal of International Economics 88(2), French, Kenneth, and James Poterba, Investor diversification and international equity markets. American Economic Review 81, Giannetti, Mariassunta, and Luc Laeven, The flight home effect: Evidence from the syndicated loan markets during financial crisis. Journal of Financial Economics 104(1), Giannetti, Mariassunta, and Luc Laeven, Local bias, stock prices, and market conditions: Evidence from U.S. mutual funds. ECGI Finance Working paper No. 335/2012. Grinblatt, Mark, and Matti Keloharju, How distance, language and culture influence stockholdings and trades. Journal of Finance 56(3), Grubel, Herbert, Internationally diversified portfolios: Welfare gains and capital flows. American Economic Review 58(5), Guiso, Luigi, Paola Sapienza, and Luigi Zingales, Time varying risk aversion. Working paper, University of Chicago. 26

27 Hamberg, Mattias, Taylan Mavruk, and Stefan Sjögren, Investment allocation decisions, home bias and the mandatory IFRS adoption. Journal of International Money and Finance 36, Huberman, Gur, Familiarity breeds investment. Review of Financial Studies 14(3), Ivkovic, Zoran, and Scott Weisbenner, Local does as local is: Information content of the geography of individual investors common stock investments. Journal of Finance 60(1), Kang, Jun-Koo and Rene M. Stulz, Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan. Journal of Financial Economics 46(1), Leuz, Christian, Karl V. Lins, and Francis E. Warnock, Do foreigners invest less in poorly governed firms? Review of Financial Studies 23(3), Levy, Haim, and Marshall Sarnat, International diversification of investment portfolios. American Economic Review 60(4), Merton, Robert C., A simple model of capital market equilibrium with incomplete information. Journal of Finance 42(3), Milesi-Ferretti, Gian-Maria and Cédric Tille, The great retrenchment: International capital flows during the global financial crisis. Economic Policy 26, Nofsinger, John, and Abhishek Varma, Individuals and their local utility stocks: Preference for the familiar. The Financial Review 47, Thapa, Chandra, and Sunil S. Poshakwale, Country-specific equity market characteristics and foreign equity portfolio allocation. Journal of International Money and Finance 31,

28 Van Nieuwerburgh, Stijn, and Laura Veldkamp, Information immobility and the home bias puzzle. Journal of Finance 64(3), Vermeulen, Robert, International diversification during the financial crisis: A blessing for equity investors? Journal of International Money and Finance 35,