Excessive Fund Flows Can Damage Performance

Transcription

1 Daily Mutual Fund Flows and Redemption Policies Abstract We examine how redemption policies affect daily fund flows in open-end mutual funds. Since short-term trading, as manifested in daily funds flows, can have an adverse impact on the passive shareholders of the fund, mutual funds may find it desirable to discourage short-term trading through the use of redemption fees. However, if the daily fund flows are due to fund shareholders legitimate liquidity demands, the redemption fee would have little effect on daily fund flows and possibly an adverse affect on fund shareholders by imposing a liquidity cost on them. We use a cross-sectional comparison to show that daily fund flows are significantly lower for funds with redemption fees. We also use a sample of funds that recently enacted redemption fees to examine whether the distribution of daily fund flows changed after the initiation of the redemption fee. We find that redemption fees are an effective weapon against active traders who attempt to exploit the mutual fund. 1. Introduction In addition to the benefits of professional investment management and diversification, open-end mutual fund shareholders also enjoy the privilege of nearly free and unlimited liquidity. Most funds grant shareholders the right to exchange fund shares for cash at the fund s end-of-day per share net asset value (NAV). Although the fund s costs from transacting shares of the fund s underlying assets are passed on to the shareholders in the form of expenses and lowered realized returns, shareholders pay no direct transaction costs when trading fund shares. While numerous studies explore the performance of the professional managers, more recent scrutiny focuses on how the liquidity feature of mutual funds affects fund performance. 1 This paper examines mutual funds attempts to control the transaction of the fund shares through fund policies, such as redemption fees. 1 Studies of overall fund performance include Carhart (1997), Grinblatt and Titman (1994), Gruber (1996), Hendricks, Patel, and Zeckhauser (1993), Ippolito (1989), Ippolito (1992), and Jensen (1968). Studies that consider the liquidity of fund shares and its impact on performance include Bergstresser and Poterba (2000), Dickson, Shoven, and Sialm (2000), and Edelen (1999). 1

2 Investors redeem shares for several reasons. Typically, redemptions are considered to be motivated by the infrequent liquidity shocks to fund shareholders. These liquidity motives are similar to those of the uninformed liquidity traders in the models of Glosten and Milgrom (1985), Kyle (1985) and Admati and Pfleiderer (1988). Liquidity-motivated and infrequent redemptions that are uncorrelated should not impose a significant constraint on the fund manager s portfolio selection strategy and should not result in significant costs to the fund. However, Edelen (1999) shows that excessive fund flows can have detrimental effects on mutual fund performance, due to transaction costs in adjusting the underlying portfolio holdings and the need to carry cash to fund liquidations. These costs could originate from a small group of shareholders who abuse the liquidity features. Some shareholders might trade fund shares in order to engage in market timing or speculative trading. Among these traders, some might redeem shares in order to exploit a mispricing in the fund's per share NAV. Bhargava and Dubofsky (2000) and Chalmers, Edelen, and Kadlec (2001) show that fund s NAV can be mispriced due to the stale prices of the fund's underlying assets. Greene and Hodges (2001) show that traders do exploit stale-priced mutual fund shares in international funds, significantly diluting fund returns. The gains to the active mutual fund traders come at the expense of the buy-and-hold investors. In addition to the adverse impact of strategically timed fund flows, the shifting of capital-gains tax liabilities to passive investors can lead to even greater dilution of fund returns, as shown by Dickson, Shoven, and Sialm (2000). Managers might attempt to limit the effects of these funds exchanges. Chordia (1996) presents a model in which fund managers choose the fund s policies (an exchange fee in the model) to entice investors to self-select into funds based on their liquidity needs. Nanda, 2

3 Narayanan, and Warther (2000) construct a model in which the heterogeneity of managerial ability and the differences in investor liquidity and marketing needs determine the fees charged by a fund. Managers who can earn higher returns may be able to charge higher fees to deter liquidity traders. Similarly, Nanda and Singh (1998) construct a model in which the endogenous liquidity needs of investors lead them to form a mutual fund, while also determining the fee structure and size of the fund. In their model, funds that are more efficient in managing transaction costs (including the costs arising from fund redemptions and taxes) will have lower penalties for early withdrawal. Because these models depend on the power of transaction fees to discourage fund flows, it is important to understand the impact of redemption fees on investors' liquidity demands. Alternatively, fund managers who are compensated based on the size of the assets under management have an incentive to discourage shareholders from redeeming the fund shares. Since redemption fees could be a way for managers of poor-performing funds to capture investor funds by imposing a fee for exit, the U.S. Securities and Exchange Commission (SEC) generally has been reluctant to encourage widespread use of these fees. For this reason, the SEC encourages mutual funds to set redemption fees based on an estimation of the costs that the redemption imposes on the fund usually a function of the transaction costs the fund must pay to liquidate shares of the fund s underlying assets. Even with these justifications, the SEC all but prohibits redemption fees exceeding 2%. 2 Fund managers even claim in their prospectus that redemption fees exist specifically in order to discourage market timing strategies. For example, the prospectus for the UAM McKee International Equity fund states the fund " charges the 2 For example, Stanton (2000) explains how Fidelity s Small Cap Stock Fund reduced their redemption fee from 3% to 2% at the insistence of the SEC. Fidelity argued that costs imposed by short-term traders justified a 3% fee. 3

4 redemption fee to discourage market timing by those shareholders initiating redemptions to take advantage of short-term market movements (UAM, 2000, p.11). This paper empirically examines these two competing hypotheses of why funds use redemption fees. An agency cost explanation for redemption fees would imply that poor performing funds are more likely to have such fees, while the liquidity cost hypothesis suggests that funds with less liquid holdings would be more likely to employ redemption fees. In addition, we explore the effectiveness of redemption fees in affecting investors exchange behavior. We take two approaches in this examination. First, we examine whether the characteristics of the fund s underlying assets explain the use of redemption fees. To do this, we simply compare the characteristics of funds that have redemption fees with those that do not. Our second approach relates the fund s redemption fee to its shareholders exchange activity. In this context, we examine how redemption fees affect the distribution of frequent (daily) fund flows. In practice, fund managers have several exchange policies at their disposal for inhibiting shareholders ability to exchange fund shares, including prospectus-stated exchange restrictions, market timing language, exchange or redemption fees, minimum holding periods, and load fees. We choose to focus on redemption fees for several reasons. First, redemption fees typically are charged to the person affecting an exchange and are paid directly to the fund. This is in contrast to load fees or some exchange fees that get paid to the fund management company or a third party (e.g., a broker). Thus, redemption fees impose costs on investors leaving the fund and compensate existing shareholders for the costs the fund must pay to rebalance the portfolio. Although many studies have examined the impact of fees and transaction costs on investment performance, there is very little empirical work on how transaction fees actually influence investor behavior. This paper attempts to fill this gap by documenting the impact of redemption 4

5 fee changes on daily mutual fund flows, and thus their effectiveness in controlling the liquidity and performance of mutual funds. The importance of fees in deterring redemptions from investors is also relevant to models of financial intermediation. Several of these models employ production technologies where value is lost if investments are liquidated early. Diamond and Dybvig (1983) note that such a quality would be equivalent to a transaction cost on liquidating the investment. The importance of this cost of withdrawing funds from an investment is a major factor in the models of Diamond and Dybvig (1983), Jacklin (1993), and Chan, Greenbaum, and Thakor (1992), just to name a few. Most of these papers concentrate on bank deposits as the investment examined, but the use of fees to deter frequent trading of mutual funds has many important similarities. Therefore, a better understanding of the influence of redemption fees on investors willingness to withdraw from certain investments is important for the applicability of these models as well. The remainder of this paper is organized as follows. The following section establishes the methodology and key predictions that we examine using a sample of all Morningstar-listed mutual funds. Using this sample, we conduct cross-sectional analysis of funds with and without redemption fees. We examine daily fund flows in section 3. In this analysis, we conduct both cross-sectional analysis and a time-series event study of funds that initiate redemption fees during our sample period. Section 4 summarizes and offers concluding remarks. 2. Cross-Sectional Analysis 2.1 Methodology Fund managers often claim that redemption fees are used in order to decrease the liquidity costs imposed on remaining shareholders when a fund experiences redemptions. 5

6 However, the SEC s reluctance to allow excessive redemption fees implies that there exists the potential for a serious agency problem where fund managers impose unnecessary fees on shareholders in the hopes of controlling investors decisions. 3 Thus, it is important for shareholders, fund managers, and financial regulators to know if funds really do use redemption fees when they are faced with high liquidity costs, or alternatively, when they are faced with high potential agency costs. The agency cost explanation for redemption fees implies that managers use redemption fees to prevent investors from leaving the fund. Managers could be motivated to capture investor dollars in this way whenever there exists conditions that benefit fund managers but hurt investors. Thus such funds could be characterized by high expenses and/or poor performance. High expense ratios would lead to greater payments to fund managers, without directly benefiting shareholders. 4 Alternatively, investors in poorly performing funds would have an incentive to shift assets to a better managed fund, unless prevented from doing so by a cost such as a redemption fee. Since fund manager compensation usually depends on the size of assets under management, they would have a strong incentive to prevent such defections. The ability of redemption fees to prevent defections from poorly performing funds has particular importance in helping to explain the asymmetric relation between mutual fund flow and performance 5. Costs, such as redemption fees, that prevent investors from leaving poorly performing funds could help explain such asymmetries by explaining why investors remain in funds that show repeated low returns. A final consideration in fund managers motivations to use redemption fees is to examine how redemption fees influence the growth of a fund s assets. If redemption 3 In early 2000, the SEC rejected several fund companies attempt to charge redemption fees higher than 2% (see Stanton (2000)). 4 Gruber (1996), Carhart (1997), and Chalmers, Edelen, and Kadlec (1999), all show that funds with higher expense ratios do not necessarily provide investors with higher returns. 5 Sirri and Tufano (1999) find asymmetric flow performance relationship. 6

7 fees lead to slower fund growth then we could expect that smaller funds would be less likely to use redemption fees so as to insure higher growth rates. However, once fund s size reaches a suitable level, then management s concern with asset growth could lessen, and redemption fees may become a more attractive means of controlling investors behavior. A liquidity explanation for redemption fees suggests that mangers use these fees to control liquidity costs that impede investor s returns and thus act in the best interests of longterm shareholders. As Edelen (1999) and Greene and Hodges (2001) show, investors taking advantage of short-term strategies can earn excess returns at the expense of passive shareholders. This occurs because free redemptions allow for unlimited and costless liquidity to be provided to short-term traders at the expense of passive investors. Therefore, if redemption fees are implemented in order to discourage traders from imposing high liquidity costs on a fund, then we would expect funds with high-liquidity costs to make more use of redemption fees. Thus funds with that hold smaller market capitalization stocks, more concentrated holdings, or more volatile holdings to be more likely to employ redemption fees. Similarly, funds operating in sectors characterized by lower liquidity, such as international or specialty equity funds, to depend more on redemption fees. In addition to the direct costs of trading in a fund s holdings, the shifting of capital gains tax liabilities could impose major costs on passive shareholders following frequent fund redemptions (Sialm, Dickson, Shoven, 2000). Thus funds with greater potential capital gains tax liabilities could also be expected to make more use of redemption fees. We employ two separate data sources to examine how mutual funds use redemption fees. First, we examine a broad sample drawn from the September 2001 Morningstar Principia CD. This sample admits an examination of the characteristics of funds that utilize redemption fees. 7

8 To select our sample of equity funds, we require that the Morningstar database have non-missing data for the following mutual fund characteristics: total net assets, historical return standard deviation, Morningstar star rating, median market capitalization of the fund s stocks, number of holdings, percentage of portfolio in the top 10 holdings, turnover, expense ratio, cash, potential capital gains exposure, and prospectus objective. For bond funds we do not include median market capitalization or cash holdings, but we do require non-missing data for the average effective duration of the fund's assets and average credit quality, in addition to the other variables also used for equity funds. The second sample focuses on how redemption fees and daily fund flows interact. We take two approaches in examining the impact of redemption fees on daily fund flows. First, we use cross-sectional analysis to compare the fund flows in funds that have redemption fees to those that do not. As with any cross-sectional comparison, the results might be suspect as to whether the characteristics that might influence the daily fund flows are suitably controlled. Therefore, our second approach relies on an event-study of funds that initiate redemption fees. Our event study sample comprises only seven funds that initiated redemption fees during our sample period. Though the analysis is limited to case studies of these funds, we believe they provide important examples of how redemption fees affect fund flows. An additional benefit of this examination to complement our cross-sectional analysis is that we are certain to have held the characteristics of the fund constant and changed only the fund policy with respect to redemption fees. To determine which funds in both samples have redemption fees, we utilize the Quicken.com interface to the Morningstar database as of September This interface allows us to identify (or sort by) which funds have redemption fees. Out of the universe of 11,264 funds, 8

9 583 (5.2%) have redemption fees. This leaves us with a sample that includes 222 equity funds and 29 bond funds with redemption fees. Matching this information with our daily funds flows sample, we find 27 equity funds and 6 bond funds that have redemption fees. For our event study, we search the prospectus of each fund that has a redemption fee to determine whether the fee is initiated in our sample period. Seven funds initiate redemption fees in our sample period and form the basis for our event-study analysis. 2.2 Cross-Sectional Sample Data Table 1 offers descriptive statistics of the Morningstar sample, with means of the fund characteristics reported separately for funds with and without redemption fees. The overall sample consists of 3,589 equity funds and 1,148 bond funds. We report equity fund characteristics for the overall sample and by four general fund types: domestic Stock, domestic Growth, International, and Specialty. Bond funds are divided into three types: Corporate, US Government, and Municipal. Overall, there are 222 (6.2%) equity funds with redemption fees and 29 (2.5%) bond funds with such fees. For domestic stock and growth funds, less than 5% have redemption fees, while 9% of international funds and nearly 16% of specialty funds have the fees. Less than 1% of US government bond funds have redemption fees, while 2.5% of corporate and 3.4% of municipal bond funds do. Equity funds with redemption fees tend to be insignificantly smaller than non-redemption fee funds, except specialty funds with redemption fees which are nearly twice the average size of non-redemption fee specialty funds. In addition, the Morningstar star rating is significantly lower for funds with redemption fees and the historical standard deviation of returns is significantly higher. For most investment objectives, redemption fee funds are comparable to other funds in 9

10 terms of operations as measured by turnover, expense ratio, total number of holdings, and percentage of cash holdings. Consistent with a liquidity motivation for redemption fees, the median market capitalization of funds underlying assets are significantly smaller and the concentration of holdings in the top ten assets is significantly higher for redemption fee funds. The differences for most variables are generally stronger for funds in the international and specialty investment objectives. Bond funds tend to be insignificantly larger with redemption fees, while there are few significant differences for the historical return standard deviation, concentration in the top ten holdings, average credit quality, or average effective duration. However, the number of holdings is significantly lower for bond funds with redemption fees, offering some support to liquidity costs as a justification for redemption fees. Turnover is also significantly lower for funds with redemption fees, possibly reflecting the effect of the fees on investors' behavior. Furthermore, the finding that expense ratios are significantly lower for funds with redemption fees appears inconsistent with an agency cost hypothesis, as is the finding of marginally significantly higher star ratings for funds with redemption fees. The inconsistencies with the agency cost explanation are especially strong for the municipal bond category. 2.3 Empirical Results Table 2 reports probit model estimates of fund characteristics that determine whether a fund has a redemption fee. The results are generally consistent with funds use of redemption fees to offset transaction costs both for equity and bond funds, although different variables seem to describe the liquidity costs for each type of fund. Funds in less liquid equity categories, such as international and specialty, are much more likely to have redemption fees than domestic stock 10

11 or growth funds. Equity funds with lower median market capitalization of holdings or higher concentrations in their top ten holdings have an increased probability of a redemption fee. Likewise, bond funds with fewer holdings or lower average credit quality are more likely to have redemption fees. This suggests that funds that face high liquidity costs might use redemption fees to help curtail frequent liquidity demands from their fund shareholders. Similarly, cash holdings are inversely related to the incidence of redemption fees for equity funds, while turnover is negatively related to the probability of a redemption fee for bond funds. In this case, the causality possibly is reversed, since redemption fees might cause funds to hold less cash than they otherwise would need to fund redemptions or to suppress turnover. The evidence on the agency cost of redemption fees is mixed. Redemption fees are significantly inversely related to expense ratios in equity funds. This is contrary to the use of redemption fees to keep investors from flocking to lower-expense fee funds. Similarly, equity funds with higher star ratings are more likely to have redemption fees. If Morningstar star ratings are a true indication of fund quality, then this is inconsistent with the use of redemption fees to trap money in poor performing funds. For bond funds, the coefficients for star ratings and expense ratios are of the same sign as for equity funds, but without statistical significance. 3. Daily Fund Flows 3.1 Sample Description We utilize a sample of daily fund flows to capture the effect of redemption fees on frequent exchanges. Our daily fund flows data originate form TrimTabs, Inc. The TrimTabs daily data has daily observations of per share net asset value ( p t ), total net assets (a t ), and a distribution indicator for all funds covered by TrimTabs, Inc. We put the data through rigorous 11

12 screens for errors, focusing our attention on extreme observations. When the NAV appears to be the source of the extreme observation, the NAV is hand-checked against alternative sources. Those that can be corrected and verified are kept, while remaining suspicious observations, such as apparent keying errors, are discarded. We match the TrimTabs data with fund description data from Morningstar, including the prospectus objective of each fund. From the prospectus objective, we use the same four equity style categories as above, plus a bond category. Greene and Hodges (2001) utilize a similar sample from TrimTabs and show that this sample is reasonably representative of the universe of mutual funds. Funds for which we do not have at least 100 days of data or do not have matching Morningstar data are discarded. We also remove funds that average less than $10 million per day in net assets within our sample period. This results in a sample of 433,670 daily observations from 833 mutual funds, which represents approximately 20% of the assets of all mutual funds. Counted among these 833 funds are multiple classes of the same fund (e.g., Class A or B shares). Since there is no a priori reason to assume that the flow to one class would be the same as that to another class of the same fund, we do not aggregate classes into one fund. For example, the distinguishing feature of one class versus the other of the same fund can be an exchange privilege or redemption fee. Therefore, we conduct our analysis of fund flows at the fund class level. From the changes in NAV and total net assets, TrimTabs computes a mutual fund s net fund flow (purchases less sales of fund shares) on date t as pt c t = at at 1. (1) p t 1 This calculation accurately yields date t fund flows as long as the total net asset numbers are post-flow (i.e., reflect date t flow). Following Greene and Hodges (2001), we compare TrimTabs 12

13 data with GAAP-conforming balance sheets filed with the SEC, since GAAP requires total net assets to be reported post-flow. In most cases, TrimTabs receives daily net asset numbers from the mutual fund companies that are actually pre-flow (i.e., do not account for the current-day s flow). In these cases, flow attributed to date t actually occurs on date t-1. For those that we find that flow is reported with a lag, we adjust the flow accordingly. 6 We also adjust fund flows for dividends and distributions. Table 3 reports descriptive statistics for our daily fund flows sample. The average (median) size of the funds in our sample is $954 million ($402 million). Growth funds are significantly larger than the other fund types, with an average (median) of about $1.7 billion ($648 million) in net assets. Domestic stock and bond funds have average (median) net assets of $973 million ($552 million) and $539 million ($303 million), respectively. International funds have average (median) net assets of $688 million ($163 million). Since a few funds enter our sample after the start of our sample period and some exit before the end of our sample period, we annualize every fund s return, assuming daily compounding, for comparability. The average stock fund over the 26-month period has a 14.97% annualized return compared with a 20.5% annualized return for the S&P 500 index over the same period. Bond and growth funds have an average annualized return of 2.51% and 33.70%, respectively, while International funds average a 26.44% annualized return. 3.2 Cross Sectional Analysis of Other Market Timing Restrictions In this section, we replicate the results of Greene and Hodges (2001) by examining how fund flows are affected by other restrictions on market timing contained in the prospectuses of U.S. based mutual funds that invest primarily in international equity securities. Chalmers et al. 6 See Greene and Hodges (2001) for a detailed description of the timing of fund flows in the TrimTabs data. 13

14 (2001) report that approximately 51% of international mutual funds maintain exchange restrictions in their prospectus. Some fund families state that they allow only a limited number of exchanges per year, while others put no limit on the number of exchanges. Virtually all fund families that we are aware of reserve the right to limit exchanges, especially if they suspect market timing activity. Enforcement of limited exchange policies varies widely. 7 To examine these issues, we partition the Trim Tabs international fund sample according to stated exchange restrictions and portfolio composition. For the exchange restrictions, we searched each fund s prospectus for: stated exchange restrictions (e.g., exchanges limited to 4 per year ), market timing language (e.g., do not allow market timing ), exchange fees (e.g., 0.25% per exchange or $10 per exchange ), minimum holding period (e.g., may not exchange within 7 days of last exchange ), and load (e.g., front-end or deferred load). As noted earlier, any stated exchange policy can vary from the effective policy based on the level of monitoring or enforcement. Therefore, we expect the more operational restrictions, such as exchange fees and minimum holding periods, to have a larger impact on exchanges than the other policies, such as timing language and stated exchange restrictions. Table 5 shows the level of fund flows and the dilution impact for the sample funds partitioned by exchange restrictions. Only the minimum holding period results in significantly lower fund flows and less (in magnitude) dilution. 8 Interestingly, front-end load funds have significantly higher daily fund flows and more dilution. Funds with either front-end or deferred loads show no significant 7 We contacted over 30 fund families in late Approximately one-third of those with whom we spoke either stated that there are no restrictions on exchanges or that the restrictions are limits in print, but only sporadically enforced. Some funds, for example, stated that they only restrict large accounts of $1 million or more, but do not monitor smaller accounts systematically. 8 These tests have little power to detect a significant effect of exchange fees, since only 17 of the 109 funds have them as of the end of our sample period and 9 of these funds initiated fees in the middle of our sample period. We have found reductions in fund flows (and dilution) when these funds initiated exchange fees. 14

15 difference from no-load funds. These results suggest that timers are more concentrated in frontend load funds compared with no-load or deferred load funds Event-Study Results of Funds with Redemption Fees Seven mutual funds in our TrimTabs sample implement redemption fees during our sample period (February 1999 to September 2000). We could identify no funds that eliminated redemption fees during this period. The two Bond funds initiating redemption fees are Invesco High Yield Fund and UAM McKee U.S. Government Fund. Four international funds initiate redemption fees: Invesco European, Invesco Pacific Basin, Invesco Latin American Growth, and UAM McKee International Equity. UAM McKee Small Cap Equity is the one domestic Stock fund initiating a redemption fee during our sample period. The fees on Invesco funds are for redemptions within 180 days. Because these fees do not penalize shareholders who redeem after long holding periods, they are aimed at thwarting fund shareholders who trade funds more actively than other shareholders do. Whether daily fund flows reflect the liquidity demands of short-term traders versus long-term traders in each fund is unknown. If daily fund flows typically reflect the long-term-shareholders liquidity demands, then the redemption fees might fail to curb exchanges. This would result in no significant change in fund flows after the initiation of a redemption fee. Alternatively, if daily fund flows originate from traders who engage in market timing or other strategic trading, these redemption fees should raise their costs and significantly affect their trading activity. However, redemption fees might not be enforceable on all shareholders who actively trade fund shares. Some shareholders might exchange fund shares through a third party, such as a fund supermarket or retirement plan. 9 The convention with front-end loads is that an investor is charged the load when entering the fund family. In nearly all cases, investors may exchange into and out of the funds within the same family without paying an additional load 15

16 In these cases, third parties typically batch exchange orders and transmit a single net exchange order to the fund company. This renders detection of individual traders impossible, unless the third party enforces each fund s policy. Whether the redemption fee event causes a decrease in daily fund flows depends on whether fund traders who strategically engage in active trading are able to avoid the redemption fee. Table 5 reports the how daily fund flows compare between the 5-month period before the redemption period versus the 5-month period after the redemption fee. For the seven funds, there is no significant difference in the average daily net fund flow between the two periods. However, there is a significant difference between the level of fund flows in the two periods. We capture the statistical significance in a test of the standard deviation of daily net fund flows between the two periods. To offer a more transparent measure of redemption or exchange activity, we report the average level (i.e., the absolute value) of daily net fund flows in both percentage of assets and dollar terms. In the UAM U.S. Government and the UAM Small Cap Equity funds, the average size of daily net fund flows increases after a redemption fee is initiated. The other five funds reveal a decrease in the size of daily net fund flows. The international funds show the largest change in fund flows, with nearly all experiencing less than 50% trading activity once redemption fees are in place. For example, Figure 1 shows the daily fund flows for the Invesco European fund. In the five months prior to a redemption fee, the level of net fund flows average over $26 million (3.61% of assets) per day. After redemption fees, the daily net fund flows average only $2.8 million (0.51% of assets). The use of redemption fees has recently been targeted at thwarting flows from strategic traders. Greene and Hodges (2001) show that traders of international mutual fund shares exploit stale prices by following a simple trading rule. The trading rule is to exchange into an fee. Furthermore, most funds waive the load if investors return to the fund family within 90 days of a redemption. 16

17 international mutual fund on days in which the U.S. market is up (i.e., has a positive return) and exchange out of the fund when the U.S. market drops. This strategy profits from the correlation of international markets with the U.S. market. Greene and Hodges estimate a significant transfer of wealth from passive, buy-and-hold shareholders to these active market timers. By raising traders costs to this strategy through a redemption fee, funds might be able to reduce this strategic trading. We examine this possibility with those funds in our sample that initiate redemption fees. Consider the following regression model that explains daily fund flows. We use two variables to capture the trading signal in the U.S. market as an indicator of international stale prices. First, we use the S&P500 return, SP500 t, each day. This indicates both the general direction of flows using this strategy and allows larger magnitude returns to motivate increased trading interest. Since a trader would not exchange into a fund on successive positive return days, we also use a trading variable, Signal t, that captures the signal to exchange into or out of a fund only on reversal days. This variable takes on the value 1 (-1) when the S&P500 return is positive (negative) following a negative (positive) return day. We add the dummy variable, PreFee t, to indicate the period prior to the redemption fee and interact this dummy variable with the intercept and the slopes. This results in the following regression model for daily fund flows. f PreFee SP PreFee Signal PreFee SP Signal + e t = β0 t + β1 500t t + β2 t t + γ 0 + γ1 500t + γ 2 t t. (1) If a significant proportion of the fund s daily flows are motivated by strategic traders, this model should explain a significant proportion of the flows. Moreover, if redemption fees are effective at 17

18 limiting strategic trading, then the model parameter estimates should only be significant in the period prior to the redemption fee. While the international fund managers might target the elimination of strategic fund flows, they might also be concerned about frequent liquidity flows. We can interpret the residuals from equation (1) as the noise- or liquidity-components of daily fund flows. We use the Glesjer test to determine whether the redemption fee changes the variance of the error term. If the redemption fee only eliminates strategic fund flows, then we would fail to reject a change in the level of noise trading in the daily fund flows, as measured by the variance of the residuals. Table 6 reports the regression parameter estimates for equation (1). For the four international funds, the model explains a significant amount of the variation in all but the Invesco Latin American Growth fund. As Greene and Hodges (2001) report, Latin American funds could suffer the least from stale prices, since these markets are open at roughly the same time as U.S. markets. Conversely, the model fails to explain any of the variation in daily fund flows for the bond and small-cap funds. For the three international funds, the Glesjer test rejects the null of no heteroskedasticity, indicating that the redemption fee lowers the noise- or liquidity-components of daily fund flows. An alternative explanation is that the model fails to capture all of the strategic fund flows and these are still reflected in the regression residuals. In summary, three of the four international funds had significant daily fund flows that are consistent with a trading rule that exploits stales prices. The redemption fee eliminates these flows. Given the significant size of the flows associated with strategic trading, we might expect those strategic traders to leave these funds and seek out other funds in which to employ their strategy. Thus, redemption fees might also affect the size of the assets of a fund. If many investors deem the short-term liquidity option to be valuable, then the fund might experience net 18

19 flow out of the fund prior to the initiation of the redemption fee. That is, fund shareholders might exit a fund before having a binding redemption fee imposed on their shares. Similarly, shareholders who would have otherwise entered the fund with new money might choose competing funds for their investments. We examine how the redemption fees affect the level of the fund s net assets. For comparison, we match each fund with all other TrimTabs sample funds based on the prospectus objective. Table 7 examines the change in Net Assets for each fund for the 15-day period immediately before and after the fund implementing a redemption fee. Five of the seven funds had redemptions exceeding purchases (i.e., negative fund flows) in the 15 days following the implementation of a redemption fee. In six cases, the net flow is lower than the fund s peer group. This suggests that shareholders view the liquidity option to be valuable and a redemption fee to be a deterrent to investing in a fund. This could explain the apparent reluctance of many firms to adopt redemption fees, despite the evidence that redemption fees appear to thwart trading that has been shown to negatively impact a fund s return. 4. Conclusions In this paper, we examine how redemption policies affect daily fund flows in open-end mutual funds. We also find evidence that funds tend to use redemption fees in order to discourage investor behavior that raises the costs of the mutual fund. In general, funds that hold underlying assets with low liquidity, such as small capitalization stocks, are more likely to have redemption fees. Similarly, specialty and international equity funds and funds with concentrated holdings are more likely to use redemption fees. Although a reverse causality may explain the findings between cash holdings or turnover with redemption fees, we find little evidence that funds use redemption fees in order to capture investor dollars from defecting poor performers. 19

20 In part, we base this on the positive relationship between redemption fees and quality ratings by Morningstar for equity funds. We use a cross-sectional comparison to show that daily fund flows are significantly lower for funds with redemption fees. We exploit the natural experiment of a sample of funds that recently enacted redemption fees to examine whether the distribution of daily fund flows changes after the initiation of the redemption fee. For funds that experienced large turnover from investor exchanges, we find that redemption fees drastically reduce fund flows. For example, several international funds initiated redemption fees in the face of large daily fund flows. Prior to the fee, these funds witnessed an average of 3.4% of their assets turnover per day. After the redemption fees were initiated, fund flows decreased to less than 0.5% of assets per day. Moreover, the fees affected the nature of the redemptions. Prior to the fees, several international fund flows experienced large daily flows consistent with strategic trading that would dilute fund returns. After the redemption fees are initiated, strategic flows of this kind are not evident. In total, our results are consistent with funds use of redemption fees to mitigate shareholders costs. Though redemption fees impose costs on liquidating shareholders, they tend to lower the costs the all of the funds shareholders by reducing both direct costs to transacting in the fund s underlying shares and indirect costs of the dilution of shareholder wealth by strategic, high-frequency exchanges. Though redemption fees succeed in this goal where other exchange policy available to funds fail, we find that surprisingly few funds utilize the fees. 20

21 References: Admati, Anat, and Paul Pfleiderer, 1988, A Theory of Intraday Patterns: Volume and Price Variability, The Review of Financial Studies 1, Bergstresser, Daniel, and James Poterba, 2000, "Do After-Tax Returns Affect Mutual Fund Inflows?" NBER working paper #7595. Bhargava, R., and D. Dubofsky, "A Note on Fair Value Pricing of Mutual Funds," Journal of Banking and Finance, forthcoming. Carhart, Mark, 1997, On the Persistence of Mutual Fund Performance, Journal of Finance 50, Chalmers, John, Roger Edelen, and Gregory Kadlec, 2001, On the Perils of Security Pricing by Financial Intermediaries: The Wildcard Option in Transacting Mutual Fund Shares, Journal of Finance, forthcoming. Chan, Yuk-Shee, Stuart Greenbaum, and Anjan Thakor, 1992, Is Fairly Price Deposit Insurance Possible? Journal of Finance 47, Chordia, Tarun, 1996, The Structure of Mutual Fund Charges, Journal of Financial Economics 41, p Diamond, Douglas, and Philip Dybvig, 1983, Bank Runs, Deposit Insurance, and Liquidity, Journal of Political Economy 91, Dickson, Joel, John Shoven, and Clemens Sialm, 2000, "Tax Externalities of Equity Mutual Funds," NBER working paper #7669. Easley, David, and Maureen O Hara, 1992, Time and the Process of Security Price Adjustment, Journal of Finance 47, Edelen, Roger, 1999, Investor Flows and the Assessed Performance of Open-end Mutual Funds, Journal of Financial Economics 53, Glesjer, H., 1969, A New Test for Heteroscedasticity, Journal of the American Statistical Association, 64, Glosten, L. and P. Milgrom, 1985, "Bid, Ask, and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders," Journal of Financial Economics 13, p Goetzmann, William, Zoran Ivkovic, and K. Geert Rouwenhorst, 2001, Day Trading International Mutual Funds: Evidence and Policy Solutions. Journal of Financial and Quantitative Finance, forthcoming. 21

22 Greene, Jason, and Charles Hodges, 2001, The Dilution Impact of Daily Fund Flows on Openend Mutual Funds. Journal of Financial Economics, forthcoming. Grinblatt, Mark, and Sheridan Titman, 1994, A Study of Monthly Mutual Fund Returns and Performance Evaluation Techniques, Journal of Financial and Quantitative Analysis 29, Gruber, Martin, 1996, Another Puzzle: The Growth in Actively Managed Mutual Funds, Journal of Finance 51, Hendricks, Darryl, Jay Patel, and Richard Zeckhauser, 1993, Hot Hands in Mutual Funds, Journal of Finance 48, Ippolito, Richard, 1989, Efficiency With Costly Information: A Study of Mutual Fund Performance, , The Quarterly Journal of Economics 104, Ippolito, Richard, 1992, "Consumer Reaction to Measures of Poor Quality: Evidence from the Mutual Fund Industry," Journal of Law and Economics 35, Jacklin, Charles, 1993, Market Rate Versus Fixed Rate Demand Deposits, Journal of Monetary Economics 32, Jensen, Michael, 1968, The Performance of Mutual Funds in the Period , Journal of Finance 23, Kyle, A.S., 1985, "Continuous Auctions and Insider Trading," Econometrica 53, Nanda, Vikram, M. P. Naraynan, and Vincent Warther, 2000, Liquidity, Investment Ability, and Mutual Fund Structure, Journal of Financial Economics, forthcoming. Nanda, Vikram, and Rajdeep Singh, 1998, "Mutual Fund Structures and the Pricing of Liquidity," working paper, Washington University and University of Michigan. Stanton, 2000, UAM, 2000, UAM Funds; The McKee Portfolios; Institutional Class Prospectus, 22

23 Table 1 Descriptive Statistics This table reports averages of fund characteristics for funds with and without redemption fees. The sample is drawn from the September 2001 Morningstar database. Panel A contains equity funds and Panel B contains bond funds. * and ** indicate significant differences at the 5- and 1- percent levels, respectively. Panel A: All Equity Funds Stock Growth International Specialty N Without 3,367 1,078 1, With Total Net Assets Without 1,015 1,072 1, With , Return Std Dev Without With 30.26** ** 32.15** Star Rating Without With 3.02* ** 3.04 Median Mkt Cap Without 31,739 30,248 42,729 20,938 25,758 With 19,524** 17,112** 28,230* 12,509** 22,990 Number of Holdings Without With ** % in Top 10 Without With 40.24** ** 53.07** Turnover Without With Expense Ratio Without With * 1.48 % Cash Without With * Panel B: All Bond Funds Corporate US Govt Muni 23

24 N Without With Total Net Assets Without With Return Std Dev Without With Star Rating Without With * Number of Holdings Without With ** ** * % in Top 10 Without With Turnover Without With 35.41** 44.67** ** Expense Ratio Without With 0.75** ** Avg Effective Duration Without With Avg Credit Quality 10 Without With Credit quality is converted to a numeric value by the following algorithm: AAA=7, AA=6, A=5, BBB=4, BB=3, B=2, other=1. 24

25 Table 2 Probit Analysis of Redemption Fees This table reports ordered probit parameter estimates for fund characteristics. The dependent variable is a indicates whether or not a fund has a redemption fee. Panel A presents the estimates for equity funds and Panel B presents the estimates for bond funds. Panel A: Equity Funds Estimate Standard P-value Error Intercept Potential Capital Gains Exposure Year Alpha Year Beta Standard Deviation of Returns Star Rating Expense Ratio Number of Holdings Amount in Top 10 Holdings Turnover Log Total Net Assets Log Median Market Capitalization Cash Holdings Percentage International Fund Dummy Specialty Fund Dummy Growth Fund Dummy Log Likelihood = Pseudo R-square = 10.24% 25

26 Panel B: Bond Funds Estimate Standard P-value Error Intercept Potential Capital Gains Exposure Year Alpha Year Beta Standard Deviation of Returns Star Rating Expense Ratio Number of Holdings Amount in Top 10 Holdings Turnover Log Total Net Assets Credit Quality Duration US Govt Dummy Corporate Dummy Log Likelihood = Pseudo R-square = 15.27% 26

27 Table 3 Cross-sectional Distribution of Total Assets and Annualized Returns Cross-sectional distribution of average daily total assets of each sample mutual fund from February 2, 1998 through March 31, The sample includes 833 mutual funds followed by TrimTabs, Inc. whose objective is categorized as bond fund, growth fund, stock fund, or international fund. Assets are reported in millions of dollars. Panel A: Average daily assets Full Sample Bond Growth Stock International Mean , Std. dev. 2, , , , Median Category Assets 788, , , ,583 68,556 Category Funds (N) Panel B: Annualized returns Full Sample Bond Growth Stock International Mean 14.57%** -2.51%** 33.70%** 14.97%** 26.44%** Std. dev % 7.79% 29.97% 17.52% 21.19% Median 7.24% -3.02% 30.23% 11.44% 23.59% N %Positive 61.7% 4.7% 95.1% 83.4% 97.2% * and ** indicate significant difference from zero at the 5- and 1-percent levels, respectively. and indicate significant difference from 50% at the 5- and 1-percent levels, respectively. 27

28 Table 4 Daily Fund Flows Conditional on Trading Restrictions for International Mutual Funds Daily fund flow statistics are reported for sub-samples portioned by exchange restrictions. Classifications with exchange restrictions and with timing language are assigned if the prospectus states explicit limits on exchanges per year or mentions restrictions on market timing, respectively. Funds that charge redemption or trading fees for exchanges are classified as with exchange fee and funds that prohibit trading within a specific number of days upon entering a fund are classified as with minimum holding period. Funds that have front-end loads are labeled with front-end load. Funds with either a front-end or a deferred load are classified as with front-end or deferred load. We also report the t-statistic on the test of the null hypothesis that the means between the two sub-samples are the same. Mean St. Dev. N With exchange restrictions 0.89% 0.98% 50 Without exchange restrictions 0.93% 0.85% 59 t-test of equality of means With timing language 0.87% 0.88% 68 Without timing language 0.98% 0.97% 41 t-test of equality of means With exchange fee 0.95% 0.97% 17 Without exchange fee 0.91% 0.90% 92 t-test of equality of means 0.16 With minimum holding period 0.49% 0.38% 15 Without minimum holding period 0.98% 0.95% 94 t-test of equality of means -3.54** With front-end load 1.54% 1.14% 30 Without front-end load 0.67% 0.67% 79 t-test of equality of means 3.93** With front-end or deferred load 0.99% 1.00% 61 No-load 0.82% 0.78% 48 t-test of equality of means 0.99 * and ** indicate significant difference from zero at the 5- and 1-percent levels, respectively 28