Foreign focused mutual funds and exchange traded funds: Do they improve portfolio management?

Transcription

1 Foreign focused mutual funds and exchange traded funds: Do they improve portfolio management? D. Eli Sherrill a, Sara E. Shirley b, Jeffrey R. Stark c a College of Business Illinois State University Campus Box 5480 Normal, IL desherr@ilstu.edu b Jones College of Business Middle Tennessee State University 1301 East Main St. Murfreesboro, TN sara.shirley@mtsu.edu c Jones College of Business Middle Tennessee State University 1301 East Main St. Murfreesboro, TN jeff.stark@mtsu.edu Abstract Approximately one third of all actively managed foreign equity mutual funds hold exchange traded funds (ETFs). We show that, dependent on the type of ETF held, mutual funds with large ETF positions experience noticeable improvements in the management of their portfolios. In particular, when large positions in ETFs are taken, mutual fund performance marginally increases, and cash holdings, portfolio risk, and turnover all decrease. We also observe that foreign equity mutual funds tend to hold ETFs within their investment objective during periods of large flows. 1

2 1. Introduction The exchange traded fund (ETF) market continues to grow, increasing from $608 billion in 2007 to $2,524 billion by year end 2016 (ICI Factbook, 2017). As the popularity of ETFs rises, their use among actively managed mutual funds becomes a more relevant topic of exploration. Sherrill et al. (2017) are the first authors to examine ETF use by actively managed domestic equity mutual funds, finding widespread use of ETFs, and evidence that ETF use is associated with lower performance, larger cash positions, and poor market timing ability. While their study finds no meaningful improvement associated with holding an ETF, we propose that foreign focused mutual funds have potentially greater gains from involvement in the ETF markets. These potential benefits are magnified among actively managed mutual funds that hold foreign equities because of the increased costs associated with barriers to international investment (Stulz, 1981). If an actively managed foreign equity mutual fund chooses to hold an ETF, it must be the case that the mutual fund did so for a reason. The first and most obvious reason is to improve fund performance. Additionally, ETF positions can help reduce the amount of cash required to handle large fund flows, thus reducing the effects of cash drag (Wermers, 2000) and flow induced trading (Edelen, 1999). ETFs have the potential to reduce portfolio risk and turnover as well. If a mutual fund can invest in one or two ETFs rather than possibly dozens of individual stocks, the mutual fund has broadened its exposure (reduced risk) and done so through one or two trades rather than dozens (lower turnover). Despite relatively little academic research on foreign equity mutual funds, their size is considerable; as of year-end 2016 their net assets totaled over $2 trillion dollars spread out over 2

3 1,500 unique mutual funds (ICI, 2017). 1 In response to the amount of money invested in foreign equity mutual funds and the possible benefits associated with ETF use, we examine the impact of holding ETF positions among U.S. based mutual funds investing internationally. Given the possible benefits of ETF use discussed above, we first provide an overview of what types of ETFs foreign investing mutual funds hold. We find that mutual funds hold ETFs with higher performance, and consistent with the potential benefits relating to improving portfolio efficiencies rather than utilizing ETFs as long term investments, we find that mutual funds are drawn to larger, less expensive, and more liquid ETFs. In looking at the types of mutual funds that hold ETFs, one striking difference arises between our sample and the sample of Sherrill et al. (2017). Whereas the Sherrill et al. study finds that the smallest domestic equity mutual funds hold ETFs, ETF-using foreign equity focused mutual funds are closer in size to their non-user counterparts ($821 million versus $1,064 million). Finally, we observe that ETF usage is widespread across all investment objectives (Emerging Markets, Global, International, and Other). We next explore the association between mutual fund characteristics and ETF use across our broad sample. Relative to non-user mutual funds, we find that mutual funds with large ETF positions underperform and have greater turnover, whereas funds with small ETF positions have greater risk and greater turnover. These associations inform us on the general attributions of the mutual funds using ETFs, though they do not allow us to determine how ETF positions impact the overall portfolio. To determine how mutual funds use ETFs, we look at the differences in portfolio characteristics between months that an ETF is held and months that an ETF is not held. We 1 As a note of comparison, US domestic equity mutual funds (those typically examined in academic research) total just over $6 trillion dollars spread across nearly 3,300 funds (ICI, 2017). 3

4 further this by dividing our ETF user mutual funds into subsamples based on the type of ETF they hold the majority of their portfolio in. These classifications include same objective ETF, equity domestic ETF, and other ETF objective. Our subgroup results provide evidence that U.S. based foreign equity mutual funds utilize ETF positions to improve the portfolio. Large ETF positions within the same investment objective of the mutual fund are associated with increased returns, decreased cash holdings, decreased risk, large inflows, and decreased turnover during the months an ETF is held. These results are consistent with ETF use improving the management of the portfolio by improving performance, reducing the cash drag on performance, reducing overall portfolio risk, helping to alleviate the effects of flow induced trading, and allowing the mutual fund to trade less frequently. While prior literature looking at ETF holdings by actively managed domestic equity mutual funds does not find evidence in favor of their use, herein we provide results supporting their use among U.S. based foreign equity mutual funds. 2. Hypothesis Design In this section we develop our hypotheses relating mutual fund characteristics to ETF use. Characteristics of interest include mutual fund performance, the amount of cash held, riskiness of portfolio returns, fund flows, and portfolio turnover Performance The first question that must be addressed is related to the performance or ability of foreign equity mutual funds. If mutual funds hold ETF positions they must be part of the fund s strategy, a strategy that presumably is designed to improve fund performance. We believe this is a natural assumption considering the career concerns of not outperforming (Chevalier and 4

5 Ellison, 1999), the flow to performance relationship (Sirri and Tufano, 1998), and compensation incentives (Kempf et al., 2009). Following these findings, we propose a hypothesis related to ETF use and mutual fund performance. H1: Foreign equity mutual funds use ETFs to generate outperformance Cash held Wermers (2000) documents a drag on performance from holding cash. The underlying argument is that cash does not generate a return. Therefore, the larger the portion of your portfolio allocated to cash, the larger the portion that cannot earn a competitive return. ETF use can help alleviate this. Because many ETFs are liquid and easily tradeable, U.S. based foreign equity mutual funds can choose to hold ETFs in place of cash, thus earning a return while maintaining the liquid benefits of their original cash positions. H2: Foreign equity mutual funds use ETFs to replace cash holdings Portfolio risk Dating back to Sharpe (1964), the benefits of efficient diversification have been extolled in financial literature. By utilizing ETF holdings, a U.S. based foreign equity mutual fund is capable of obtaining diversified exposures to a large number of equity positions through the purchase of a single security (an ETF) rather than having to purchase many individual positions. H3: Foreign equity mutual funds use ETFs to reduce the risk of their portfolio returns. 5

6 2.4. Flow Edelen (1999) and Rakowski (2010) show that mutual fund flows lead to flow induced trading and a drag on performance. The source of this reduction in performance comes from the need of a mutual fund to either quickly sell assets at a suboptimal time to generate cash for outflows, or to invest inflows quickly in suboptimal positions. ETFs can help alleviate these concerns by allowing a fund manager to have access to a liquid asset to either sell in the case of outflows or to place money into in the case of inflows. In either scenario, the mutual fund is able to utilize ETFs to mitigate the impact of flow driven trading. H4: Foreign equity mutual funds use ETFs during periods of large (positive or negative) fund flows Turnover Wermers (2000) documents a significant negative impact on net mutual fund performance that is attributable to transaction costs. The greater a mutual fund s portfolio turnover, the greater the transaction cost drag. By utilizing ETFs to gain access to a wide range of underlying equity positions through the purchase of just a single ETF, a mutual fund can greatly reduce their portfolio turnover relative to filling out a portfolio with the purchase of many individual securities. H5: Foreign equity mutual funds use ETFs to reduce portfolio turnover 6

7 3. Data and Sample Creation In this section we discuss our data sources and the steps we follow to create our sample. We then present descriptive statistics looking at ETF holdings and what types of foreign equity mutual funds most commonly hold ETFs Data and Sample Creation We obtain information on mutual fund returns, holdings, characteristics, and family characteristics from the Center for Research in Security Prices Survivor-Bias-Free US Mutual Fund Database (CRSP MF). We aggregate mutual fund share classes belonging to the same mutual fund into one TNA-weighted portfolio observation. CRSP MF provides ETF expense ratios and fund characteristics and the CRSP US Stock database (CRSP US) provides ETF prices, returns, shares outstanding, and volume traded. Our sample begins in 2004, the first full year that CRSP MF reports holdings with consistency, and ends in We retain returns at the daily level to allow for monthly calculations of risk. All other mutual fund and ETF variables are at monthly frequencies for analysis, unless otherwise noted. To construct our sample of U.S. based foreign equity mutual funds that hold passively managed ETFs, we drop fund-of-funds and utilize Lipper objective codes to retain only those funds classified as foreign equity. 2 We remove fund observations prior to their first offer date as reported in CRSP MF to avoid incubation bias (Evans, 2010). CRSP MF holdings information allows us to identify all passively managed ETF holdings. Following Chen et al. (2013), we utilize monthly holdings data to identify ETF positions and assume a maximum holding period 2 We identify foreign equity funds as those with a Lipper objective code of IF, GL, IS, GS, GIF, GTK, IRE, GRE, GNR, GFS, GH, XJ PC, LT, JA, INR, EU, EM, CH, and CN 7

8 of six months. When monthly holdings are not available, we use quarterly holdings subject to the same six-month restriction. After six consecutive months with no updated holdings data, we set the fund s holdings to missing. The resulting sample is 523 unique foreign equity mutual funds with passive ETF positions and 1,102 unique foreign equity mutual funds that never hold an ETF. The number of fund month observations is 28,170 for ETF holding funds and 68,378 for non-etf-user mutual funds ETF Sample Descriptive Statistics We first look to establish the importance of examining ETF holdings among foreign equity mutual funds by demonstrating their widespread use. In Table 1 we presents descriptive statistics of ETF use for our entire sample and by mutual fund investment objective. Of the 1,625 mutual funds in our full sample, 523 (32.18%) use an ETF. Furthermore, ETF use is widespread across all investment objectives with 40.89% of Emerging Market, 27.78% of Global, 36.73% of International, and 21.86% of other objective mutual funds holding ETFs. We also show that ETF use is not confined to one particular type of ETF. Emerging Market and International mutual funds primarily hold ETFs within their investment objective, while Global and other mutual funds primarily hold ETFs classified as domestic equity. [Insert Table 1 near here] Having provided initial evidence on the widespread use of ETFs among U.S. based foreign equity mutual funds, we turn to the type of ETFs held and the type of mutual fund holding ETFs. Table 2 presents cross-sectional ETF characteristics associated with ownership in 8

9 Panel A and mutual fund characteristics in Panel B. We observe that mutual funds tend to hold ETFs that are significantly larger ($1,781 million vs. $90 million), older (5.05 years vs years), have higher monthly returns (0.31% objective excess vs % objective excess), and that are more liquid (339 thousand shares traded daily vs. 18 thousand). In looking at which types of mutual funds hold ETFs, we observe that objective excess returns are lower by 4.5 basis points per month, cash held is lower by 0.38%, monthly inflows are lower by 0.32%, and annual turnover is higher by 23.93%. In stark contrast to the findings of Sherrill et al. (2017) who find that ETF use is primarily concentrated among small domestic equity mutual funds, we observe no significant difference between the sizes of U.S. based foreign equity mutual fund users and non-users. [Insert Table 2 near here] We confirm the observed univariate relationships between ETF use and mutual fund characteristics in a multivariate setting with the use of a logistic regression (Koski and Pontiff, 1999; and Sherrill et al., 2017): n MF Holds ETF i,t = α i + β j X i,t + ε i,t, (1) j=1 where MF Holds ETFi,t takes on a value of one if mutual fund i holds an ETF in year t. We include as mutual fund control variables Expense Ratio, Size, Return, Risk, Load, Age, Flow, Turnover, and Cash. We measure Expense Ratio as the average percent expense ratio charged by the mutual fund; Size measured as the natural log of mutual fund total net assets (TNA); Return 9

10 measured as the monthly objective excess return; Risk as the standard deviation of daily returns over the month; Load measured as an indicator variable that takes on the value of one if the mutual fund charges a load fee and zero otherwise; Age measured as the mutual fund s age in fractional years; Flow measured as the absolute value of monthly flows to a mutual fund; Turnover measured as the monthly mutual fund portfolio turnover; and Cash measured as the portion of mutual fund portfolio attributed to cash holdings. We look at yearly and crosssectional ETF holdings, where variables are measured once over the ETFs lifetime in the crosssectional tests rather than as described in Eq. (1). We present the cross-sectional relationship between mutual fund characteristics and ETF ownership in column 1 of Table 3. For analysis in columns 2 and 3, we divide our sample at the median of ETF portfolio allocation size during each calendar year and focus on low-users in column 2 and high-users in column 3. For columns 2 and 3, observations are at yearly frequency and only mutual funds that have ever held and ETF are retained in the sample. Our multivariate results show that ETF users in general have lower performance. Among our ETF-use classifications, we show that low-users have lower expense ratios and higher risk, whereas highusers have higher expense ratios, are smaller in size, are younger, experience lower turnover, and hold greater amounts of cash. Though these relationships help to inform us as to the overall mutual fund characteristics associated with ETF ownership, they do not allow us to determine how ETF use impacts mutual fund portfolios. We save this analysis for section 4. [Insert Table 3 near here] 10

11 As a final examination of our sample, we explore the association between mutual fund performance and ETF ownership in Table 4, and the associations between mutual fund cash holdings, risk, flow, and turnover and ETF ownership in Table 5. Both tables utilize a multivariate regression: n Characteristic i,t = α i + β 1 ETF Held i,t 1 + β 2 X i,t 1 + ε i,t, (2) j=2 where t is a given year and i a given mutual fund. The dependent variable in all columns of Table 4 is Objective Excessi,t, the average of a mutual fund s objective excess returns over a year. Columns 1 and 2 of Table 4 include all mutual funds while Columns 3 and 4 include only observations from mutual funds that hold an ETF at some point during our sample period. Table 5 has a dependent variable of Cashi,t in column 1, Riski,t in column 2, Flowi,t in column 3, and Turnoveri,t in column 4. Table 5 includes all mutual funds and is comparable in sample to column 2 of Table 4. The coefficients of interest for Table 5 are Low User and High User which take a value of one if the mutual fund is in the low or high group of mutual funds that use ETFs that year, respectively, and zero otherwise. We include as mutual fund control variables Size, Return, Risk, Load, Age, Flow, Turnover and Cash as defined in Eq. (1). Results presented in Table 4 confirm the univariate results from Table 2. Foreign equity mutual funds that hold ETFs underperform mutual funds that never hold an ETF. However, Table 4 shows that the coefficient for low-user mutual funds is insignificant while high users is negative and significant in all variations of the model. Thus the underperformance of ETF user mutual funds is driven by those mutual funds that are dedicating a large portion of their portfolio 11

12 to ETFs. This result is similar to the results of Sherrill et al. (2017) which looked at the performance of domestic equity mutual funds and ETF use. [Insert Table 4 near here] We present the results from our tests of other mutual fund characteristics in relation to ETF usage in Table 5. ETF usage does not relate significantly to different levels of cash, risk, or flows between low, high, and non ETF using mutual funds. The results in Column 4 suggest that both low and high ETF user mutual funds have significantly higher turnover than non ETF user mutual funds. This finding confirms the univariate results from Table 2 for both low and high ETF user mutual funds. [Insert Table 5 near here] 4. Results Within this section we formally examine our proposed hypotheses. We first look at how ETF ownership impacts mutual fund performance, then the impact on cash holdings, how risk changes with ETF use, how fund flows are associated with ETF holdings, and finally how ETF use affects portfolio turnover. 4.1 Months held versus months not held Thus far we have looked at what mutual fund characteristics relate to ETF use in general. We have not yet looked at how mutual fund characteristics change between months that a fund 12

13 uses an ETF and months that it does not. Table 6 presents the univariate statistics from our sample of ETF user U.S. based foreign equity mutual funds divided into low and high users. We present statistics during months the funds hold an ETF and months they do not. Among high users, we observe a decrease in risk by about 1.4% per year, a decrease in cash by about 0.3%, and a decrease in annual turnover by 19.6%. These significant differences indicate that ETF use meaningfully impacts portfolio characteristics, particularly among those mutual funds that take large ETF positions. [Insert Table 6 near here] 4.2. Performance Although we do not observe a significant difference in mutual fund performance during months a fund holds an ETF among our univariate results, performance is arguably the most important fund characteristic available. As such, we first examine how foreign equity mutual funds utilize ETF holdings as related to mutual fund performance as suggested in H1. This hypothesis (H1) posits that mutual funds hold ETFs to improve performance. We examine this relationship in more detail through a multivariate regression of low users in columns 1 and 2 and high users in columns 3 and 4: n Objective Excess i,t = α i + β 1 ETF Held i,t + β 2 X i,t 1 + ε i,t, (3) j=2 where Objective Excessi,t is measured as the month t return earned by mutual fund i relative to an equal weighted benchmark of returns calculated from other mutual funds over the same month t 13

14 belonging to the same objective classification as mutual fund i. In columns 1 and 3, our variable of interest, ETF Heldi,t, takes on a value of one if mutual fund i holds an ETF during month t and zero otherwise. In columns 2 and 4, we replace our ETF Held indicator variable with three indicator variables identifying each ETF user mutual fund as either a Same Obj ETF user, an Equity Domestic Obj ETF user, or an Other ETF Obj user. We construct these three classifications by identifying which type of ETF constitutes the largest portion of a mutual fund s ETF portfolio. Each ranking is mutually exclusive. As control variables, we measure Expense Ratio as the average percent expense ratio charged by the mutual fund; Size measured as the natural log of ETF total net assets (TNA); Return measured as the monthly objective excess return; Risk as the standard deviation of daily returns over the month; Load measured as an indicator variable that takes on the value of one if the mutual fund charges a load fee and zero otherwise; Age measured as the mutual fund s fraction age in years; Flow measured as the monthly flows to a mutual fund; Turnover measured as the monthly mutual fund portfolio turnover; and Cash measured as the portion of mutual fund portfolio attributed to cash holdings. We further cluster standard errors across both mutual fund and time dimensions and report p- values in parentheses. [Insert Table 7 near here] Consistent with our univariate results, we observe no significant change in fund performance across low or high user mutual funds when looking at our general ETF Held variable (columns 1 and 3). However, when looking at our high user group broken down into type of ETF held, we find a marginally significant improvement to mutual fund performance 14

15 associated with the use of ETFs within the same investment objective as the mutual fund amounting to 9.6 basis points per month (1.15% per year). This indicates that when U.S. based foreign equity mutual funds take large ETF positions within their investment objective, they experience an improvement in performance, thus providing marginal support for our first hypothesis; that ETF use is associated with an improvement in mutual fund performance Cash holdings We now move to our second hypothesis: that ETF use can help reduce the amount of cash held by mutual funds and thus remove some of the cash drag discussed by Wermers (2000). To look at this impact, we repeat the analysis of Eq. (3) while changing our dependent variable from a measure of performance to the amount of cash held by a mutual fund. Furthermore, we drop cash as an independent variable from the analysis. We report our results from this specification in Table 8 and maintain the same column setup regarding low users (columns 1 and 2) and high users (columns 3 and 4) as well as with general use (columns 1 and 3) and specific use (columns 2 and 4). [Insert Table 8 near here] Similar to the results presented in Table 7, when looking at cash we find evidence that our ETF use affect is concentrated among high user mutual funds holding ETFs within their investment objective. For these funds, we observe a decrease in cash holdings of 1.34%. This finding is consistent with the notion of our second hypothesis looking at ETFs as an alternative liquid security that earns a return while maintaining many of the same beneficial characteristics 15

16 traditionally associated with cash holdings. Furthermore, because the observed effect is found among ETFs within the same investment objective as the mutual fund family, it further supports the idea that these mutual funds utilize ETFs to manage cash positions. If the effect were found outside of the same investment objective, the mutual funds would not be using ETFs to gain access to returns consistent with their investment strategy and could be using ETFs to generate performance through active management rather than manage their cash drag Portfolio risk We test hypothesis 3: that ETF use is associated with a reduction in portfolio return risk, by replacing the dependent variable in Eq. (3) with portfolio risk measured as the standard deviation of daily returns over the course of a month. We subsequently remove risk from our included independent variables. We report our risk related results in Table 9, with low users in columns 1 and 2 and high users in columns 3 and 4. General ETF use is explored in columns 1 and 3 and specific ETF use is in columns 2 and 4. [Insert Table 9 near here] Results from Table 9 show that high user mutual funds experience a general reduction in portfolio risk from the use of ETFs as seen in column 3. This provides support of our third hypothesis, but we do not yet know if the reduction is widespread across all ETF types or concentrated among a particular type of ETF. When looking at column 4, we observe negative and significant coefficients on Same Obj ETF and Other ETF Obj. This indicates that the reduction in portfolio risk comes from investment within the same objective as the mutual fund 16

17 and within ETFs outside of equity domestic or the fund s specific investment objective. Given that Sharpe (1964) motivates the benefits of efficient diversification through gaining exposure to additional securities, it is not surprising that the reduction in portfolio risk is spread among various types of ETFs rather than concentrated among one classification Mutual fund flows Edelen (1999) and Rakowski (2010), among others, document the negative effects of fund flows on mutual fund performance. The negative effect is a result of forced trading that accompanies large inflows or outflows. A mutual fund manager has a limited number of opportunities to make value-creating investments. As a fund experiences large outflows, the manager is forced to liquidate certain positions to meet investors demand for cash. Similarly, when a mutual fund experiences large inflows, the fund manager must invest the new money. Oftentimes, this demand for money or additional funds needing investment result in suboptimal investment decisions. By utilizing ETFs when funds experience large fund flows, a manager has a way to allocate new assets in a way that will remain liquid while earning a return until a more optimal investment opportunity presents itself. We explore the possibility that ETF use is associated with fund flows by modifying Eq. (3) to have the absolute value of monthly fund flows as the dependent variable. We report our flow related results in Table 10, with low users in columns 1 and 2 and high users in columns 3 and 4. General ETF use is explored in columns 1 and 3 and specific ETF use is in columns 2 and 4. [Insert Table 10 near here] 17

18 Table 10 shows that mutual funds utilize different types of ETFs in different ways. If a mutual fund is trying to utilize an ETF as a means to help manage large fund flows rather than as a staple in an active strategy, we expect them to use ETFs within the same investment objective as the mutual fund itself. We observe this relationship in column 4. The coefficient in Same Obj ETF is positive and marginally significant, indicating that mutual funds tend to hold ETFs within their investment object when fund flows are larger. Furthermore, we observe negative and significant coefficients on the Equity Domestic Obj ETF and the Other ETF Obj variables. This indicates that ETF use outside of the fund s objective is associated with periods of lower magnitude of fund flows, possibly as part of an active investment strategy rather than as a means to help manage periods of large inflows. These results support our fourth hypothesis relating ETF use to periods of large fund flows Portfolio turnover In our final analysis, we explore the relationship between ETF use and portfolio turnover. Wermers (2000) documents the negative effects of transaction costs on mutual fund performance. As a result of this negative effect, mutual funds utilizing ETFs may be doing so as a way to reduce portfolio turnover. Rather than having to directly purchase dozens of securities and making numerous transactions, a mutual fund can purchase one ETF and gain access to a large number of underlying securities. In doing so, the mutual fund has the potential to decrease their turnover and thus reduce the negative impact of transaction costs of portfolio performance. We explore this possibility by modifying Eq. (3) to have portfolio turnover as the dependent variable. Similar to our other analyses within section 4, we report our flow related results in 18

19 Table 11, with low users in columns 1 and 2 and high users in columns 3 and 4. General ETF use is explored in columns 1 and 3 and specific ETF use is in columns 2 and 4. [Insert Table 11 near here] Results here are mixed. Though the expectation is that ETF use across any type of ETF can reduce portfolio turnover, the observed effect is concentrated among our Other ETF Obj variable. This indicates that mutual funds do not utilize ETFs within their objective or ETFs providing access to US stocks as a means to reduce transactions and turnover. Although the only significant coefficient supports our final hypothesis, the insignificant coefficients on the remaining ETF type variables prevents us from definitively stating that the results of Table 11 provide support for hypothesis Conclusion The results presented throughout this study attempt to address the question: how do U.S. based foreign equity mutual funds utilize ETF holdings? The motivation behind this question is twofold: first, ETFs can provide an efficient and affordable method to improve the efficiency of how a foreign-focused mutual fund manages its portfolio, and second, ETF use is widespread among U.S. based foreign equity mutual funds, with approximately one third of mutual funds using ETFs. To understand how ETF use impacts the management of a portfolio, we explore five related hypotheses: ETF use improves fund performance, ETF use replaces cash holdings, ETF use reduces portfolio risk, ETF use helps manage large fund flows, and ETF use reduces 19

20 portfolio turnover. In all cases, ETFs have the potential to improve the overall management of the mutual fund. Unlike the negative association between ETF use and fund performance found among actively managed domestic equity mutual funds (Sherrill et al., 2017), we document numerous benefits associated with ETF use among foreign equity focused funds. Our results indicate that ETF use is associated with improved fund performance, smaller cash holdings, lower risk, and periods of large fund flows. Taken as a whole, we find that U.S. based foreign equity mutual funds utilize ETFs to assist in the management of their portfolio, to improve fund performance, and to help reduce drags on performance related to holding excess cash and flow induced trading. Furthermore, our results provide insight into an often overlooked area of mutual fund management; that of foreign equity mutual funds. We additionally provide the first support for the beneficial use of ETFs by mutual funds. By examining how ETFs help improve portfolio management among foreign equity mutual funds, we help both practitioners and academics better understand how a particular type of mutual fund operates. 20

21 References Berk, J. B., and R. C. Green, Mutual fund flows and performance in rational markets. Journal of Political Economy 112, Busse, J. A., and Q. Tong, Mutual fund industry selection and persistence. Review of Asset Pricing Studies 2, Carhart, M. M., On persistence in mutual fund performance. Journal of Finance 52, Chevalier, J. A., and G. D. Ellison, Career concerns of mutual fund managers. Quarterly Journal of Economics 114, Cici, G., and L. Palacios, On the use of options by mutual funds: Do they know what they are doing? Journal of Banking and Finance 50, Chen, H., H. Desai, and S. Krishnamurthy, A first look at mutual funds that use short sales. Journal of Financial and Quantitative Analysis 48, Chen, H., N. Jegadeesh, and R. Wermers, The value of active mutual fund management: an examination of the stockholdings and trades of fund managers. Journal of Financial and Quantitative Analysis 35, Cremers, M. K. J., and A. Petajisto, How active is your fund manager? A new measure that predicts performance. Review of Financial Studies 22, Cumber, R. E., and J. D. Glen, Evaluating the performance of international mutual funds. Journal of Finance 45, Doshi, H., R. Elkamhi, and M. Simutin, Managerial activeness and mutual fund performance. Review of Asset Pricing Studies 5, Edelen, R. M., Investor flows and the assessed performance of open-end mutual funds. Journal of Financial Economics 53, Evans, R. B., Mutual fund incubation. Journal of Finance 65, Fama, E. F., and K. R. French, Size, value, and momentum in international stock returns. Journal of Financial Economics 105, French, K., Presidential address: the cost of active investing. Journal of Finance 63, Frino, A., A. Lepone, and B. Wong, Derivative use, fund flows and investment manager performance. Journal of Banking and Finance 33,

22 Henriksson, R., and R. Merton, On market timing and investment performance: Statistical procedures for evaluating forecasting skills. Journal of Business 54, ICI, Investment Company Institute Fact Book Ippolito, R. A., Consumer reaction to measures of poor quality: Evidence from the mutual fund industry. The Journal of Law and Economics 35, Jensen, M. C., The performance of mutual funds in the period Journal of Finance, Kacperczyk, M., S. V. Nieuwerburgh, and L. Veldkamp, Time-varying fund manager skill. Journal of Finance 69, Kacperczyk, M., C. Sialm, and L. Zheng, On the industry concentration of actively managed equity mutual fund. Journal of Finance 60, Kao, G. W., L. T. W. Cheng, and K. C. Chan, International mutual fund selectivity and market timing during up and down market conditions. The Financial Review 33, Kempf, A., S. Ruenzi, and T. Thiele, Employment risk, compensation incentives, and managerial risk taking: Evidence from the mutual fund industry. Journal of Financial Economics 92, Koski, J. L., and J. Pontiff, How are derivatives used? Evidence from the mutual fund industry. Journal of Finance 54, Natter, M., M. Rohleder, D. Schulte, and M. Wilkens, The benefits of option use by mutual funds. Journal of Financial Intermediation 26, Rakowski, D., Fund flow volatility and performance. Journal of Financial and Quantitative Analysis 45, Rohleder, M., D. Schulte, and M. Wilkens, Management of flow risk in mutual funds. Review of Quantitative Finance and Accounting 48, Sharpe, W. F., Capital asset prices: A theory of market equilibrium under conditions of risk. Journal of Finance 19, Sherrill, D. E., S. E. Shirley, and J. R. Stark, Actively managed mutual funds holding passive investments: What do ETF positions tell us about mutual fund ability? Journal of Banking and Finance 76, Sirri, E. R., and P. Tufano, Costly search and mutual fund flows. Journal of Finance 53,

23 Stalter, K., When active fund management makes sense for investors. U.S. News and World Report. Stulz, R. M., On the effects of barriers to international investment. Journal of Finance 36, Treynor, J., and K. Mazuy, Can mutual funds outguess the market? Harvard Business Review 44, Wermers, R., Mutual fund performance: An empirical decomposition into stock-picking talent, style, transactions costs, and expenses. Journal of Finance 55,

24 Table 1 Type of ETF used by mutual fund This table presents the distribution of ETF use my mutual funds. Column 1 provides a total number of mutual funds in this sample and the number based on Lipper Objective. Column 2 details the number of mutual funds that have ever held an ETF. Column 3 provides the number of mutual funds that have ever held an ETF that had the same objective as the mutual fund while column 4 provides the number of mutual funds that held ETFs with Equity Domestic objectives. Fund Count Using ETFs Using Same Objective Using ED ETFs Objective [1] [2] [3] [4] All Equity Foreign 1, Emerging Markets Global International Other Objectives

25 Table 2 Descriptive statistics Panel A contains the descriptive statistics of ETFs held and not held by mutual funds. Panel B contains the descriptive statistics of our sample of mutual funds that have held an ETF and those that have not. Mean and standard deviation values are reported. ETF Size and Mutual Fund Size are the total net assets (TNA) of the fund, ETF Age and Mutual Fund Age are the fund s age, Expense Ratio is the fund s expense ratio. Returns are the average monthly returns of a fund. Flow is the monthly percentage into fund. Price is the average daily price of an ETF. Volume is the average number of shares traded per day. Objective Excess Returns are the difference between the mutual fund s returns and the returns of all other mutual funds in its objective. Risk is the standard deviation of daily returns over a month. Turnover is the mutual fund s turnover ratio. Load Indicator takes a value of one if the mutual fund has a front or rear load and a value of zero otherwise. Cash Held is the percent of the mutual fund s portfolio TNA held in cash. Abs Flow is the absolute value of monthly mutual fund flows. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively. Panel A: ETF Characteristics ETF Held by Mutual Fund ETF Not Held by Mutual Fund (N =567) (N =1,287) Variable Mean Std. Dev. Mean Std. Dev. Difference ETF Size (millions) 1, , *** ETF Age (years) *** Returns (percent) *** Flow Price *** Volume (thousands) , *** Expense Ratio (percent) *** Panel A: Mutual Fund Characteristics ETF User Mutual Fund Non-ETF User Mutual Fund (N = 523) (N = 1,102) Variable Mean Std. Dev. Mean Std. Dev. Difference Expense Ratio (percent) Mutual Fund Size (millions) , , ,

26 Returns (percent) Objective Excess Returns (percent) *** Risk (annualized decimal) *** Load (percent of funds charging) * Mutual Fund Age (years) *** Cash Held (percent) * Flow (percent) ** Abs Flow (percent) ** Turnover (percent) *** 26

27 Table 3 Likelihood of holding an ETF This table presents the results of two logistic regressions. Column 1 utilizes a cross-sectional logistic regression where Columns 2 and 3 tests a panel logistic regression retaining annual observations of just those funds that have ever held an ETF. We obtain coefficients from MF Holds ETF i,t = α i + β j X i,t + ε i,t. j=1 The dependent variable for Column 1 takes a value of one if the mutual fund ever held an ETF, Column 2 (Column3) takes the value of one if that year the ETF holding mutual fund was ranked as a low-user (high-user) mutual fund. Expense Ratio is the fund s expense ratio, Size is the natural log of the total net assets (TNA) of the fund, Return is the average monthly objectiveexcess returns of the fund, Risk is the standard deviation of daily returns over a month, Load takes a value of one if the mutual fund has a load fee and a value of zero otherwise, Age is the natural log of the fund s age, Flow is the absolute value of monthly mutual fund flows, Turnover is the mutual fund turnover ratio, Cash is the percent of the mutual fund s portfolio TNA held in cash. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively and p- values are in parentheses. All Low High Variable [1] [2] [3] Expense Ratio *** 0.174* (0.500) (0.000) (0.089) Size *** (0.180) (0.722) (0.000) Return ** (0.026) (0.716) (0.119) Risk ** (0.873) (0.030) (0.964) Load * (0.300) (0.099) (0.444) Age *** (0.255) (0.531) (0.000) Flow ** * (0.014) (0.116) (0.088) Turnover 0.008*** 0.001* *** (0.000) (0.066) (0.001) Cash *** (0.325) (0.169) (0.004) Intercept *** *** (0.001) (0.001) (0.135) Observations 1,625 3,552 3,552 n 27

28 Table 4 Mutual fund performance This table reports the relation between holding an ETF and mutual fund performance. We obtain coefficients from Objective Excess i,t = α i + β 1 ETF Held i,t + β 2 X i,t 1 + ε i,t. j=2 The dependent variable is average objective excess returns and observations are at annual frequencies. The coefficients of interest are Low User and High User which take a value of one if the mutual fund is in the low or high group of mutual funds that use ETFs that year, respectively, and zero otherwise. Expense Ratio is the fund s expense ratio, Size is the natural log of the total net assets (TNA) of the fund, Risk is the standard deviation of daily returns over a month, Load takes a value of one if the mutual fund has a load fee and a value of zero otherwise, Age is the natural log of the fund s age, Turnover is the mutual fund s turnover ratio, Cash is the percent of the mutual fund s portfolio TNA held in cash. Columns 1 and 2 include all mutual funds and Columns 3 and 4 include only those mutual funds that have held an ETF at some point during the sample period. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively, error terms are clustered by mutual fund and time, and p-values are in parentheses. Variable [1] [2] [3] [4] Low User (0.310) (0.296) (0.800) (0.368) High User *** ** ** * (0.006) (0.023) (0.047) (0.083) Expense Ratio ** ** (0.017) (0.024) Size 0.012** (0.047) (0.947) Risk (0.625) (0.156) Load (0.742) (0.430) Age (0.329) (0.620) Turnover (0.506) (0.597) Cash (0.945) (0.484) Intercept (0.693) (0.697) (0.265) (0.198) Observations 9,885 8,166 3,548 2,988 n 28

29 Table 5 Mutual fund characteristics This table reports the relation between holding an ETF and mutual fund cash, risk, flow and turnover. We obtain coefficients from Characteristic i,t = α i + β 1 ETF Held i,t + β 2 X i,t 1 + ε i,t. j=2 The dependent variable Column 1 is the percent of the mutual fund s portfolio TNA held in cash, Column 2 is the standard deviation of daily returns over a month, Column 3 is the absolute value of monthly mutual fund flows, and Column 4 Turnover is the mutual fund s turnover ratio. The coefficients of interest are Low User and High User which take a value of one if the mutual fund is in the low or high group of mutual funds that use ETFs that year, respectively, and zero otherwise. Control variables include Expense Ratio is the fund s expense ratio, Size is the natural log of the total net assets (TNA) of the fund, Risk is the standard deviation of daily returns over a month, Load takes a value of one if the mutual fund has a load fee and a value of zero otherwise, Age is the natural log of the fund s age, Turnover is the mutual fund s turnover ratio, Cash is the percent of the mutual fund s portfolio TNA held in cash. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively, error terms are clustered by mutual fund and time, and p-values are in parentheses. Cash Risk Flow Turnover Variable [1] [2] [3] [4] Low User * *** (0.915) (0.096) (0.163) (0.002) High User *** (0.458) (0.278) (0.768) (0.001) Expense Ratio 1.277*** *** 0.161** (0.003) (0.564) (0.000) (0.013) Size 0.274*** *** *** (0.008) (0.149) (0.000) (0.007) Risk ** *** (0.020) (0.661) (0.000) Load *** (0.638) (0.213) (0.000) (0.252) Age *** (0.460) (0.382) (0.000) (0.149) Turnover *** 0.047*** 1.609* (0.007) (0.007) (0.057) Cash *** 2.218* ** (0.008) (0.057) (0.016) Intercept 1.609* 0.848*** 5.866*** 0.564*** (0.057) (0.000) (0.000) (0.000) n 29

30 Observations 8,166 8,166 8,166 8,166 30

31 Table 6 Periods ETF held vs. not-held This table reports the mean, standard deviation and t-tests the differences in the means between periods that an ETF was held and periods that the ETF was not held. Panel A presents the results for all low ETF user mutual funds and Panel B presents the results for high users. Expense Ratio is the fund s expense ratio, Mutual Fund Size is the total net assets (TNA) of the fund, Returns are the average monthly returns of the fund, Objective Excess Returns are the difference between the mutual fund s returns and the returns of all other mutual funds in its objective. Risk is the standard deviation of daily returns over a month. Load takes a value of one if the mutual fund has a front or rear load and a value of zero otherwise. Age is the fund s age. Cash Held is the percent of the mutual fund s portfolio. Flow is the monthly percentage into fund. Abs Flow is the absolute value of monthly mutual fund flows. Turnover is the mutual fund s turnover ratio TNA held in cash. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively. Panel A: Low Users Periods ETF Held Periods ETF Not Held Mean Std. Dev. Mean Std. Dev. Difference Expense Ratio (percent) *** Mutual Fund Size (millions) 1, , , , Returns * Objective Excess Returns (percent) Risk (annualized decimal) Load (percent of funds charging) *** Mutual Fund Age (years) Cash Held (percent) ** Flow Abs Flow Turnover (percent) Panel B: High Users Periods ETF Held Periods ETF Not Held Mean Std. Dev. Mean Std. Dev. Difference Expense Ratio (percent) *** 31

32 Mutual Fund Size (millions) , *** Returns (percent) Objective Excess Returns (percent) Risk (annualized decimal) *** Load (percent of funds charging) *** Mutual Fund Age (years) *** Cash Held (percent) *** Flow (percent) *** Abs Flow (percent) Turnover (percent) *** 32

33 Table 7 Mutual fund performance and ETF usage by month This table reports the relation between ETF holdings and mutual fund performance. We obtain coefficients from Objective Excess i,t = α i + β 1 ETF Held i,t + β 2 X i,t 1 + ε i,t. j=2 The dependent variable is average objective excess returns. Columns 1 and 2 look at low ETF user funds while columns 2 and 3 look at high ETF user funds. The coefficients of interest for Columns 1 and 3 is ETF Held which take a value of one for months that the mutual fund holds an ETF and zero otherwise. For Columns 2 and 4 the coefficient of interest are Same Objective ETF, Equity Domestic Obj ETF, and Other ETF Obj which take the value of one if the mutual fund holds most of its ETFs in the same objective, with an equity domestic, or in other objectives, respectively, and is 0 for periods that the mutual fund does not hold an ETF. Expense Ratio is the fund s expense ratio, Size is the natural log of the total net assets (TNA) of the fund, Risk is the standard deviation of daily returns over a month, Load takes a value of one if the mutual fund has a load fee and a value of zero otherwise, Age is the natural log of the fund s age, Turnover is the mutual fund s turnover ratio, Cash is the percent of the mutual fund s portfolio TNA held in cash. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively, error terms are clustered by mutual fund and time, and p-values are in parentheses. Low-users High-users Variable [1] [2] [3] [4] ETF Held (0.892) (0.195) Same Obj ETF * (0.979) (0.061) Equity Domestic Obj ETF (0.895) (0.334) Other ETF Obj (0.555) (0.182) Expense Ratio (0.170) (0.163) (0.155) (0.154) Size (0.777) (0.804) (0.419) (0.408) Risk (0.297) (0.303) (0.549) (0.637) Load (0.286) (0.299) (0.840) (0.825) Age (0.102) (0.107) (0.474) (0.589) Turnover n 33