Mutual fund expense waivers Jared DeLisle jared.delisle@usu.edu Huntsman School of Business Utah State University Logan, UT 84322 Jon A. Fulkerson * jafulkerson@loyola.edu Sellinger School of Business & Management Loyola University Maryland Baltimore, MD 21210 January 2017 * Contact author. Part of this research was funded by a grant from the Sellinger Fund for Excellence.
Mutual fund expense waivers Abstract Mutual funds regularly change the expense ratios investors pay. Around 70% of these changes are actual changes in the expense ratio, but the remainder represent waiver activity the beginning or end of a temporary waiving of part of the expense ratio. We find that waivers are used by fund managers in a manner distinct from expense changes. Funds introduce waivers more often when performance is poor, and drop them when performance is good. However, investors do not appear to respond as strongly to waivers as actual expense changes, and fund managers do not consistently change their behavior. Our results suggest that not all expense changes are the same, and that waivers are discounted as credible operating changes by both investors and managers.
Mutual fund expense waivers Mutual funds prominently display the net expense ratio in marketing material, prospectuses, and Securities and Exchange Commission (SEC) filings. Widely regarded as the price of the fund, the ratio shows what percentage of the assets is paid annually to the fund sponsor to operate the fund. Expense ratios vary by share class, with institutional and front-load funds receiving lower expense ratios on average. Expense ratios of funds are not set in stone, however. They vary quite often over time. The average share class has about a 40% chance of either an increase or decrease in a given year over the period 2000-2015. There is more than one mechanism by which the expense ratios change, with most examples reflecting actual changes to the share class s expense ratio. Expense ratios have to be approved by shareholders and represent a significant commitment by the fund company. However, many of the changes are due to the introduction or removal of an expense waiver. A waiver occurs when the fund sponsor and the board of directors contractually agree to a specific, lower expense ratio for a specific timeframe. A waiver represents the fund sponsor voluntarily agreeing to lower their revenues from the fund, but is distinct from an actual expense ratio change in that they usually have a specific timeframe and can be negotiated with the board directly. We consider the decisions made around these two types of expense ratio changes. To begin, we consider the fund sponsor s motivations for choosing between the two options. The choice appears to be largely driven by the fund s recent performance. Expense ratios decrease (increase) more frequently when a fund has had high (low) recent returns relative to their peers. In contrast, waivers occur about twice as often for poor performing funds than high performing funds, and are more likely to be ended when a fund has high performance. We interpret the different behavior for waivers as the fund sponsor offering a temporary lower price to entice new investment when the 1
fund has recently underperformed. In contrast, an actual permanent decrease is instituted as the fund, through fund flows and good performance, increases the assets under management and thusly demonstrate a fund s economies of scale. We next consider to what degree investors respond to waivers differently than expense changes. Our results suggest that introducing waivers has a smaller impact on fund cash flow compared to next expense ratio changes for retail investors. Institutional investors appear to completely discount a waiver compared to an actual expense ratio change. For both groups, waivers do not appear to be as desirable as an actual decrease in the expense ratio. Finally, we consider if fund manager behavior changes following a waiver. Lower fees mean lower revenues per dollar invested, and the net impact of the lower fees could lead to more cost control and less effort by the manager. Our results do not suggest consistent changes in effort by the managers, suggesting that they, too, consider waivers to be not as significant as actual expense ratio changes. Our study contributes to the literature in a few ways. We find that waivers are very prominent, with a significant minority of funds waiving a nontrivial amount of their expense ratio every year. However, waivers do not appear to be as credible a signal to investors and managers about the cost of the fund as permanent reductions in expense ratios. We demonstrate that not all expense changes are the same, and that investors put significantly more weight on actual changes in the expense ratio as opposed to waivers. The remainder of our study is organized as follows. Section 1 highlights prior work. Section 2 describes the data and trends in expense waivers. Section 3 models the decisions of the fund sponsor and Section 4 demonstrates how investors respond to this decision. Section 5 examines managerial effort following a waiver. Section 6 concludes. 2
1. Prior work The mutual fund expense ratio provides compensation to the mutual fund sponsor for running the fund. Partly covering costs and partly providing profit to the sponsor, expense ratios have a large impact on investor returns. If a fund charges a 0.80 percent expense ratio and has an expected annual return of 8 percent, the expected cost for the investor is 10% of the expected return per year. Given this outsized impact, expense ratios have received significant research in the literature. The primary line of research focuses on to what degree expense ratios relates to performance. Low expense ratios have a smaller impact on returns, so these funds may have better performance to the investor. However, high expense ratios may give fund managers more resources and greater incentives that could lead to better performance. The research largely concludes expense ratios hurt net performance for the investor: for example, Carhart (1997) finds that the lowest performing funds have the highest fees. Wermers (2000) shows that the average fund outperforms its benchmark in gross returns, but the expense ratio (as well as transaction costs and other factors) leads to underperformance in net returns. Related research considers what determines the size of the expense ratios (i.e., the price). Gil-Bazo and Ruiz-Verdú (2009) document that funds with the worst gross return performance also have higher fees, and attribute this unusual combination as strategic price setting given the funds clienteles. Clientele appears to drive some of the price setting, as differences in search costs (Iannota and Navone, 2012) and sensitivity to prior performance (Christoffersen and Musto, 2002) lead to dispersion in expense ratios. 3
Viewing the expense ratio as the price of the fund, it is relevant to next consider how investors respond to this price. Not surprisingly, increasing the price of the fund decreases demand. Sirri and Tufano (1998) document this inverse relationship, though note that part of the fees expended related to marketing can change the relationship by changing the search costs for the fund. Spiegel and Zhang (2013) find higher fees reduce a fund s market share. In contrast, Barber, Odean, and Zhang (2007) find no relationship between higher expenses and changes in fund flow. Huang, Wei, and Yan (2007) use the expense ratio as a proxy for marketing expenses and attribute these higher marketing expenses to lowering the participation costs of buying a fund. Given that expense ratios matter to investors, one of the tools fund sponsors have is the ability to adjust the expense ratio through an expense waiver. A waiver reduces the expense ratio, usually over some specific time period. Blackman (Wall Street Journal, 2014) outlines several anecdotal motivations for offering a waiver. In money market funds, waivers usually occur in low rate environments or when an individual fund lags its competitors. In equity and bond funds, the claim is that waivers occur mainly in new funds that have yet to achieve sufficient economies of scale to offset the fund s expenditures. The academic research on fund waivers has been modest. Christoffersen (2001) analyzes the practice of waivers in money markets and corroborates the anecdotal claim by Blackman (2014) that money market waivers relate to a fund s relative underperformance. Christoffersen (1998) examines equity funds and finds that funds may introduce a waiver strategically when recent performance puts the fund near a Morningstar rating breakpoint. In this study, we consider how expense ratio changes and waivers differ. 2. Data 4
We construct our sample of mutual fund waivers using share class data from Morningstar Direct for all funds marked with a primary investment style of U.S. Equity and an investment category of Equity or Allocation covering the period December 1999 (when gross expense ratios are first consistently reported in Morningstar) and December 2012. We exclude index funds and funds with less than 80% of assets invested in equities. We also drop share classes with less than $5 million in assets under management and under two years old. The Morningstar data are then matched to the CRSP Survivorship Bias Free Mutual Fund database on ticker, CUSIP, and name by broadly following the matching procedure described in the Data Appendix of Pastor, Stambaugh, and Taylor (2015) and the data match requirements of Berk and van Binsbergen (2016). The waiver variable is constructed as the difference between the gross expense ratio and the net expense ratio. These data are gathered by Morningstar from the annual prospectus. Waivers can be different for different share classes in the same fund, so most of our analysis is performed using share class-year as the unit of observation. Observations without gross expense ratios, or where the calculated waiver is greater than 9% of assets are excluded. Waivers and expense changes less than 5 basis points are often rounding issues in the data, so we round these changes down to zero. We provide summary statistics of our sample in Table 1. The average fund has a net expense ratio of 1.4% and manages $568 million in assets. However, medians suggests these figures are skewed. The median share class has an expense ratio of only 1.3% and the share class includes just $78 million in assets. [Insert Table 1 here] 5
Table 2 directly compares share classes without waivers to those with waivers. The net expense ratio is about the same on average (near 1.4%), but waiver funds have a gross expense ratio of 1.87% on average, implying an average waiver size of 45 basis points. Share classes with waivers tend to be somewhat smaller, with only $172 million in assets compared to $741 million for share classes without waivers. Waiver funds have more cash flow on average at 0.13% of assets compared to 0.09% of assets. Only 5% of share classes with waivers have a five star rating compared to nearly 8% of share classes without a waiver. Share classes with waivers also tend to be younger by around three years (38 months). As waivers are voluntary, the usage can shift through time. Figure 1 shows the percentage of share classes with some expense waiver in each year. Waiver usage peaked in 2009 at 38% of share classes, while 2001 was the lowest with only 24% of share classes. From 2009 to 2010 was the largest drop in waiver usage (down 6% from 38%), and 2008 to 2009 saw the largest increase in waivers (5%). [Insert Figure 1 about here] Together, a simple examination of the data suggests that waivers are common place and correlated with a certain share class characteristics. In the next section, we consider what characteristics driving expense ratio changes, either with or without waivers, and the impact on shareholders. 3. The expense change decision A. Prior returns and expense changes In this section, we explore what motivates a fund to change expense ratios. The discussion in Section 1 suggests that recent performance is a key component in the decision to change the 6
expense ratio. Specifically, in Christoffersen (1998, 2001), a fund s recent returns motivate waivers. We first consider descriptive statistics of expense ratio changes in the context of prior performance, and then model the decision more rigorously in a multivariate setting. Table 3 shows the likelihood of a change conditional on a share class s decile rank based on prior year return. [Insert Table 3 here] Panel A presents the historical frequency of expense ratios decreasing for all share classes by return decile. For share classes in the lowest decile, about 20% reduced expenses, compared to 31% for share classes in the highest decile. While the highest performing share classes are more likely to see an expense decrease, they are far less likely to do so through waivers. Only 14% of expense reductions for the tenth decile are through waivers, representing about 4% of all share classes in the decile. The lowest decile was much more likely to employ waivers. Around 39% of expense decreases were waivers, or 8% of all share classes in a given year. The poorest performing funds were twice as likely to use a waiver than the best performing funds even though these funds were a third less likely to reduce expenses overall. Panel B presents analogous results focusing on cases where expense ratios increased. The worst performing funds were more likely to increase expenses than the best performing funds (26% of share classes compared to 19%). Waivers are most likely to be removed in the highest return decile (accounting for nearly half of all expense increases), yet much less likely for the lowest return decile (only 20% of all expense increases). In absolute turns, about 9% (5%) of the highest (lowest) return share classes had a waiver removed. These simple sorts suggest an important attribute about performance and expense ratios. Expense ratios tend to decrease with good performance and increase with bad performance. However, the use of waivers to accomplish these changes follows the opposite pattern. A low 7
return fund is less likely to lower its expense ratio, but, conditional on lowering it, it will most often use a waiver. This suggests that waivers occur for reasons different than expense ratio changes and that additional factors may be important to managers when choosing which method to employ. We consider these additional factors in the next sub-section by introducing a multivariate model. B. Modeling the expense change decision We model the likelihood of an expense change decision as a function of several factors in addition to prior returns. Some factors control for the general operating expenses of a fund (family assets, turnover, loads, and the current expense ratio). Others capture characteristics about fund attractiveness to investors (volatility and cash flow). Finally, anecdotal evidence suggests that waivers are related to a fund s age, so age dummies are also included as controls. 1 Given these controls, we model the decision using a logit model with time fixed effects. 2 The unit of observation is share class-years. These results are presented in Table 4. [Insert Table 4 here] Panel A considers cases where the expense ratio decreased. The first column provides the marginal effect of each factor on the likelihood of any type of decrease in expenses. Consistent with the univariate results, these multivariate estimates show that high return funds are 24% more likely than funds with average returns to reduce expenses, while low return funds are 9% less 1 All continuous variables have been de-meaned and divided by the full sample standard deviation for that variable. This does not affect the conclusions about statistical significance, but does allow the interpretation of the marginal effects as the impact of a one standard deviation increase in the underlying variable. 2 The results using share class fixed effects are qualitatively the same. 8
likely. Other factors also influence the decision; large and young share classes are less likely to reduce expenses, but more volatile historical returns and turnover lead to more expense decreases. The second and third columns of Panel A consider the decisions to reduce expenses or introduce waivers separately. The likelihood of high return funds reducing expenses is even higher for actual expense ratio increases. However, for waivers, high return funds are 19% less likely to introduce waiver and low return funds are 20% more likely. Comparing the controls between the second and third column, we see that in most cases waivers have similar outcomes. For example, funds in large families are more likely to reduce expenses, but less likely to introduce waivers, all else the same. However high fund flows more often lead to expense changes, yet less often lead to waivers. One concern with modeling these decisions is that the decisions are potentially nested. Waivers and expense changes have the same broad outcome (lower expense ratios), so the fund may first decide to lower expenses in some way, not considering the actual mechanism until later. Once this first decision is made, the second decision would be choosing between changing the expense ratio and changing the waiver. We model this decision process using a Heckman procedure. The first stage uses a model identical to column 1; the second stage models the decision to introduce a waiver conditional on reducing expenses with no time fixed effects. We present the results of this model in the fourth column. With respect to return rankings, the Heckman results are broadly consistent with the waiver logit results and suggest a statistically significant difference between the expense ratio decrease and waiver decision. Several controls also show statistically different outcomes. For example, a high volatility fund is 14% more likely to introduce a waiver than reduce expenses conditional on the expenses changing at all. An interesting exception is that 9
while age seems to matter for reducing expense ratios, it is not as statistically relevant for distinguishing between waivers and expense decreases. In Panel B, we repeat the same tests except for expense increases instead of decreases. We find results that are qualitatively similar with those of Panel A. For example, low return funds are more likely to increase expenses, but less likely to drop a waiver. One difference is that young funds are much more likely to remove a waiver. This is consistent with anecdotal evidence and our own unreported analysis that waivers are very common among recently introduced share classes. We also repeat the Heckman analysis. In that case (column 4), the coefficients stay largely the same compared to the basic waiver results, though the effect of high return goes from 29% to 51%. Overall, in this section we find that waivers and expense ratio changes are viewed differently by the fund sponsor. A non-trivial factor is prior performance, with poor performing funds introducing waivers and high performing funds dropping waivers. However, the net impact on shareholders is effectively the same between the two outcomes both change the revenues for the sponsor and both require interaction with the board. In the next section, we consider if the investor response to these changes explains why fund sponsors pursue different strategies. 4. Investor reaction to waivers The prior section considered what factors motivate a fund to change their expense structure. This decision likely depends both on the fund characteristics considered in that section and what fund managers believe will be the investor response. Investors reveal preferences quite directly with mutual funds by buying or selling new shares. Data about monthly cash flows into funds is readily available, and the prior academic literature has explored many of the characteristics that 10
determine how investors decide which mutual funds to buy and sell. Relevant to our study, Sirri and Tufano (1998) find that cash flow decreases following expense increases. In this section we consider how these waivers affect investor decisions. We model cash flows broadly following the prior literature. Specifically: CF t = α + β 1 ExpenseDecrease t 1 + β 2 NewWaiver t 1 + β 3 ExpenseIncrease t 1 (1) + β 2 WaiverDropped t 1 + γ X t 1 + ε t where: ExpenseDecrease is a dummy variable equal to one if there is any type of expense decrease in the prior year, NewWaiver is a dummy variable equal to one if ExpenseDecrease was a newly introduced waiver, ExpenseIncrease is a dummy variable equal to one if there is any type of expense increase in the prior year, WaiverDropped is a dummy variable equal to one if ExpenseIncrease was due to the fund dropping a waiver, and, X t is a vector of variables controlling for prior returns, volatility, age, total net assets (TNA), turnover, net expense ratio, and loads. We estimate Eq. (1) using a panel regression with share class and time fixed effects, and standard errors clustered by share class. Cash flow is measured by calendar year as a percentage of beginning-of-year TNA, with all independent variables measured as of the end of the prior year. The results for this estimation are presented in Table 5. We also provide results separately for retail and institutional funds, as these two groups may assess waivers differently given the separate marketing channels. 11
[Insert Table 5 here] As expected, expense decreases lead to higher cash flow in the following year (3.91%). To get a sense of the magnitude, a fund earned an additional 11.62% in assets on average if it was in the top 30% of returns, so the effect from an expense decrease is about a third the magnitude of the outperformance effect. However, investors appear to react differently to a decrease if it resulted from a waiver. In those cases, the impact is reduced by 3.25%, for a net impact indistinguishable from zero (p=0.48). Investors appear to discount expense changes from waivers as a credible signal of improvements to future fund performance. Also, as expected, expense increases reduce subsequent cash flows, with an average decrease of 1.74% per year. If the expense increase was due to a waiver being removed or expiring, the effect is opposite, with a net effect around 3% on average, suggesting that funds benefit when waivers end. We speculate that a fund would only allow a waiver to end if the manager anticipates a positive reception by investors, so this is possible correlated with other positive characteristics of the fund. Part of these results may be driven by the clientele, so we next consider retail and institutional share classes separately. Retail share classes and institutional share classes react differently to the instituting of waivers. For both groups, an expense decrease yields approximately 4% more cash flow in the following year. Retail investors respond less when the decrease comes from a waiver, but the net effect is still a gain. Institutional investors appear to withdraw money; these share classes have an average loss in assets of 2.28% in the following year. For expense increases, institutional investors have effectively no response to an expense increase, though dropping waivers leads to very large subsequent cash flows. Retail share classes have losses 12
following an expense increase of 2%, but if it was from a waiver removal the fund had gains of around 2.43%. In general, investors appear to treat waivers differently from the typical expense change. Institutional share classes have a stronger response to changes due to a waiver, perhaps because they represent the decisions of high worth investors, or etail funds see a larger increase in flows, while institutional share classes show only a modest increase in flows following a waiver. The time frame of the waiver effects retail investors more, while institutional investors make no distinction between long-running and new waivers. B. Saliency of waivers to investors The prior sub-section assumes that waivers are salient to investors. However, investors may instead be reacting to the observed net expense ratio without considering the waiver itself. This section considers if investors distinguish between actual changes in the net expense ratio and the introduction of waivers. We make this comparison by adopting the model in Eq. (1) as follows: CF t = α + β 1 NetExpenseRatio t 1 + β 2 WaiverDV t 1 (2) + β 3 NetExpenseRatio t 1 WaiverDV t 1 + β 4 WaiverYears t 1 2 + β 4 WaiverYears t 1 + γ X t 1 + ε t 5. Manager reaction to waivers While investors receive obvious benefits from lower expense ratios, fund managers and families appear to have a less obvious benefit. As seen in the prior section, waivers do not increase cash flows, so there is revenue lost through the waiver that is not offset by a larger asset base. If 13
the average fund appears to lose revenues with a waiver, could the fund be changing operations to lower their costs simultaneously? In this section, we consider cost control as a way managers turn the waiver into a net profit for the fund family. If the old saying you get what you pay for is true in fund management, lower net expense ratios through waivers should imply less active management by the fund manager. Measuring management effort is difficult in general, so we consider two proxies. First, the turnover ratio represents actual trading by the fund manager. If fund managers put forth less effort by, for example, decreasing the amount of information they gather, or deferring trades to get better prices in the market, the amount of trading will decrease. A second proxy is the Active Share (AS) measure of Cremers and Petajisto (2009). AS captures how different a fund s portfolio is from the fund s benchmark index. High AS suggests the manager has positions quite different from the portfolio and is therefore actively pursuing a strategy distinct from tracking the benchmark index. In the context of effort, we assume that a decrease in effort would be accompanied by a decrease in AS as the manager engages less in information production and instead moves the fund s assets closer its benchmark index. Both proxies are incomplete; managers could pursue low turnover and low AS strategies for reasons other than cost control. However, the degree to which waivers are correlated with changes in these variables could be suggestive of how managers respond to changes in their fund expense structures. We model the impact of waivers on managerial effort as follows: Effort t = α + β 1 Waiver t 1 + β 2 WaiverDV t 1 + γ X t 1 + ε t (3) Effort includes the current year turnover ratio or active share depending on the model. Waiver represents the amount of the waiver, as larger waivers may have a bigger impact than small waivers. WaiverDV is a dummy variable taking on a value of one if the fund currently has a waiver 14
and zero otherwise. The Active Share measures come from Petajisto (2013), though it is only available through 2012. 3 The output for this model is presented in Table 6, though we suppress control variables for the sake of brevity. [Insert Table 6 here] Considering turnover first, we see in Panel A that the average fund has a nontrivial decrease in trading of 12% in the year following a waiver. The larger the waiver, the more turnover decreases. Considering retail and institutional share classes separately, retail share classes appear to have the largest decrease in turnover. Institutional share classes show little effect from waivers in any way. This suggests that manager effort decreases following a waiver. AS is shown in Panel B. AS appears to increase with waivers, with waiver funds having an AS 3.8 percentage points higher. The effect is similar for retail funds, but, again, institutional funds do not appear to change with managerial effort. Broadly, it suggests that waiver funds increase their AS following a waiver. This suggests that manager effort increases following a waiver. Considering both proxies, we do not have a clear signal that manager effort changes following a waiver introduction. 6. Conclusion Expense changes occur on a regular basis and often represent a nontrivial change in the cost of investing in a fund. Changes are quite common with roughly 40% of share classes having some change in a given year. A manager may implement a change in two ways: actual changes in the expense ratio and temporary waivers. We consider the motivations behind these changes, and 3 We thank Petajisto for making the Active Share measures from this article available at petajisto.net. 15
particularly focus on why a waiver may be more or less meaningful than an actual expense ratio change. Actual expense ratio decreases occur in the best performing funds, and increases occur in the worst performing funds. In contrast, waivers occur most often in the worst performing funds, and are removed most often in the best performing funds. We speculate that waivers are used to manage the perception of a fund s short term performance, improving the desirability of bad performing funds and profiting from the desirability of good performing funds. Our results are also consistent with waivers being viewed as temporary by both investors and management. This is suggested in two ways. First, new waivers do not lead to higher cash flow. Investors appear to not consider waivers to be a big component of the fund s anticipated future returns, and distinguish them from expense ratio changes. Second, waivers do not have a clear impact on managerial effort. While turnover appears to decrease, so does the activeness of the portfolio. Overall our results suggest that waivers are a temporary tool for fund sponsors to manage a fund s position in the market and do not carry as much importance as actual expense ratio changes. 16
References Barber, Brad, Terrance Odean, and Lu Zheng, 2005, Out of sight, out of mind: The effects of expenses on mutual fund flows, Journal of Business 78, 2095-2119. Berk, Johnathan, and Jules van Binsbergen, 2016, Assessing asset pricing models using revealed preferences, Journal of Financial Economics 119, 1-23. Blackman, Andrew, 2014, Funds fees discounts aren t forever, Wall Street Journal, http://www.wsj.com/articles/sb10001424052702304014504579248542262335648, published 1/4/14, accessed 8/25/16. Carhart, Mark. 1997. On persistence in mutual fund performance. Journal of Finance 52, 57-82. Christoffersen, Susan E. K., 1998, Fee waivers in mutual funds, Unpublished dissertation, ProQuest Paper AAI9913445, http://repository.upenn.edu/dissertations/aai9913445, accessed 8/25/16. Christoffersen, Susan E. K., 2001, Why do money fund managers voluntarily waive their fees?, Journal of Finance 56, 1117-1140. Christoffersen, Susan E. K., and David K. Musto, 2002, Demand curves and the pricing of money management, Review of Financial Studies 15, 1499-1524. Cremers, Martijn, and Antti Petajisto, 2009, How active is your fund manager? A new measure that predicts performance, Review of Financial Studies 22, 3329-3365. Del Guercio, Diane, and Paula Tkac, 2008, Star power: The effect of Morningstar ratings on mutual fund flow, Journal of Financial and Quantitative Analysis 43, 907-936. Gil-Bazo, Javier, and Pablo Ruiz-Verdú, 2008, When cheaper is better: Fee determination in the market for equity mutual funds, Journal of Economic Behavior & Organization 67, 871-885. Gil-Bazo, Javier, and Pablo Ruiz-Verdú, 2009, The relation between price and performance in the mutual fund industry, Journal of Finance 64, 2153-2183. Huang, Jennifer, Kelsey D. Wei, and Hong Yan, 2007, Participation costs and the sensitivity of fund flows to past performance, Journal of Finance 62, 1273-1311. Iannotta, Giuliano, and Marco Navone, 2012, The cross-section of mutual fund fee dispersion, Journal of Banking & Finance 36, 846-856. Pastor, Lubos, Robert Stambaugh, and Lucian Taylor, 2015, Scale and skill in active management, Journal of Financial Economics 116, 23-45. 17
Petajisto, Antti, 2013, Active share and mutual fund performance, Financial Analysts Journal 69, 73-93. Sialm, Clemens, Laura Starks, and Hanjiang Zhang, 2015, Defined contribution pension plans: Sticky or discerning money?, Journal of Finance 70, 805-838. Sirri, Erik, and Peter Tufano, 1998, Costly search and mutual fund flows, Journal of Finance 53, 1589-1622. Spiegel, Matthew, and Hong Zhang, 2013, Mutual fund risk and market share-adjusted fund flows, Journal of Financial Economics 108, 506-528. Wermers, Russ. 2000. Mutual fund performance: An empirical decomposition into stock-picking talent, style, transaction costs, and expenses. Journal of Finance 55, 1655-1703. 18
Figure 1 Mutual fund expense ratio waivers through time The figure below shows the percentage of share classes with an expense ratio waiver for each year 1999-2012. The sample includes only actively managed US equity funds. 40% Percentage of funds with waiver 35% 30% 25% 20% 15% 10% 5% 0% 19
Table 1 Summary statistics The table below presents summary statistics for a sample of equity mutual fund share classes obtained from Morningstar Direct covering the period 1999-2012. It is joined to the CRSP Mutual fund database on ticker, CUSIP, and name. Net expense ratio is the prospectus net expense ratio as a percentage of net assets; Gross expense ratio is the prospectus gross expense ratio as a percentage of assets; Equity is the percentage of assets invested in equities; Cash is the percentage of assets invested in cash; Closed is a dummy variable with a value of one if the share class is closed to new investment; Retail is a dummy variable equal to one if the share class is designated as a retail share class in CRSP; Institutional is a dummy variable equal to one if the share class is designated as an institutional share class in CRSP; TNA is the fund s total net assets as of the end of the year of the observation; Family assets is the total investment company assets managed by the fund s family in CRSP as of the end of the year of the observation; Twelve month return is the return for the calendar year; Twelve month std. dev. Is the standard deviation of monthly returns during the year; Age is the time in months since the oldest observation for that share class in CRSP; Turnover is the fund s report turnover ratio; Front-end load is a dummy variable that takes on a value of one if the share class has a front-end load; Back-end load is a dummy variable that takes on a value of one if the share class has a back-end load; Any load is a dummy variable that takes on a value of one if the share class has any load; Star fund indicates that the share class has a five-star rating from Morningstar; Star family indicates that another share class in the fund family has a five-star rating from Morningstar; and Annual flow is the sum of the implied monthly flows for the share class for the calendar year as percentage of beginning of year s TNA. 20
N Mean 50th Percentile 25th Percentile 75th Percentile Std. Dev. Net expense ratio (%) 58,767 1.40 1.30 0.65 1.81 0.54 Gross expense ratio (%) 58,767 1.54 1.40 0.69 1.94 0.76 Equity (%) 54,352 95.20 96.30 85.93 98.41 4.44 Cash (%) 54,352 2.83 1.99-0.03 4.10 3.64 Closed (DV) 54,352 0.07 0.00 0.00 0.00 0.25 Retail (DV) 58,761 0.68 1.00 0.00 1.00 0.47 Institutional (DV) 58,761 0.31 0.00 0.00 1.00 0.46 TNA 55,493 568.55 78.40 7.20 302.50 2700.53 Family Assets 58,688 149,809 43,969 314 119,579 302,735 Twelve month return 54,621 0.07 0.09-0.39 0.20 0.23 Twelve month std. dev. 50,843 0.05 0.05 0.02 0.06 0.02 Age 58,767 114.21 89.00 16.00 144.00 110.84 Turnover 57,361 0.85 0.65 0.12 1.10 0.79 Front-end load (DV) 58,767 0.21 0.00 0.00 0.00 0.41 Back-end load (DV) 58,767 0.39 0.00 0.00 1.00 0.49 Any load (DV) 58,767 0.54 1.00 0.00 1.00 0.50 Star Fund 58,767 0.07 0.00 0.00 0.00 0.26 Star Family 58,767 0.67 1.00 0.00 1.00 0.47 Annual Flow (%) 50,525 0.10-0.05-0.39 0.19 0.53 21
Table 2 Characteristics of waiver and non-waiver share classes The table below compares summary statistics for funds with a waiver and those without a waiver. Waiver % is defined as the size of the waiver as a percentage of the gross expense ratio. The sample and all other variables are as defined in Table 1. The t-stat is from a two-sample test of equality. Sub-group Mean No waiver Waiver t-stat Waiver (%) 0.01 0.45 Net expense ratio (%) 1.38 1.42-7.35 Gross expense ratio (%) 1.39 1.87-73.86 Equity (%) 95.04 95.55-12.31 Cash (%) 2.91 2.67 6.89 Closed (DV) 0.07 0.05 11.45 Retail (%) 0.69 0.68 2.94 Institutional (%) 0.30 0.31-2.27 TNA 740.67 171.67 22.91 Family Assets 169,733 106,284 23.68 Twelve month return 0.07 0.09-8.11 Twelve month return std. dev. 0.05 0.05-2.08 Age 126.02 88.40 38.67 Turnover 0.82 0.90-11.17 Front-end load (DV) 0.20 0.22-4.33 Back-end load (DV) 0.39 0.37 4.72 Any load (DV) 0.54 0.54-0.37 Star Fund 0.08 0.05 12.94 Star Family 0.70 0.60 25.95 Annual Flow (%) 0.09 0.13-8.03 22
Table 3 Expense changes conditional on prior return The table below presents the likelihood of expense ratio changes conditional on prior performance. Every year, all funds in the sample are sorted by prior year returns into deciles. Panel A presents the probability of an expense decrease occurring and panel B presents the probability of an expense increase occurring. % of changes that are waivers and % of changes that are waiver removal represent the percentage of the observations in the Yes column that are related to waivers. Panel A - Expense decreases: Any expense decrease? Return Rank No Yes % of changes that are waivers Worst - 1 79.66% 20.34% 39.27% 2 79.69% 20.31% 34.46% 3 80.77% 19.23% 32.01% 4 81.42% 18.58% 29.72% 5 78.91% 21.09% 29.30% 6 78.34% 21.66% 26.95% 7 77.48% 22.52% 25.84% 8 76.64% 23.36% 19.80% 9 74.68% 25.32% 20.98% Best - 10 68.67% 31.33% 13.96% Panel B - Expense increases: Any expense increase? Return Rank No Yes % of changes that are waiver removal Worst - 1 73.86% 26.14% 21.15% 2 78.31% 21.69% 23.85% 3 80.26% 19.74% 27.57% 4 80.03% 19.97% 31.36% 5 81.68% 18.32% 32.32% 6 81.32% 18.68% 37.86% 7 83.04% 16.96% 37.98% 8 82.23% 17.77% 37.59% 9 83.04% 16.96% 41.31% Best - 10 81.34% 18.66% 49.08% 23
Table 4 Likelihood of expense ratio changes The table below compares provides the marginal effect on the unconditional probability of offering specific expense events occurring using a logit model or a Heckman model. Top 30% return (DV) and Bottom 30% return (DV) are dummy variables equal to one if the share class has a return in the prior year in the top and bottom 30%, respectively. Young is a dummy variable equal to one if the share class is less than five years old, while Middle-age equals one if the share class is five to ten years old. Other variables are defined as in Table 1. All independent variables come from the prior calendar year, while the dependent variable takes on a value of one if the fund if the event occurred in the current year. The model also includes year and standard errors clustered by share class. Panel A focuses on expense ratio decreases and Panel B focuses on increases. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level. 24
Panel A: Expense ratio decreases VARIABLES Logit Logit Logit Heckman Dependent variable Any expense decrease Only expense ratio decrease Only waivers Only waivers Top 30% return (DV) 1.2393*** 1.4054*** 0.8061*** 0.7544*** [0.000] [0.000] [0.000] [0.005] Bottom 30% return (DV) 0.9062*** 0.7958*** 1.1960*** 1.2083*** [0.001] [0.000] [0.000] [0.007] Twelve month volatility 1.0874*** 1.0559*** 1.0799*** 1.1400*** [0.000] [0.008] [0.005] [0.005] Young (< 5 years, DV) 0.8303*** 0.8153*** 0.9095 0.8228* [0.000] [0.000] [0.158] [0.068] Middle-age (5-10 years, DV) 0.8604*** 0.8399*** 0.9543 0.9125 [0.000] [0.000] [0.346] [0.262] Log(TNA) 0.9525** 1.0019 0.8502*** 0.8540*** [0.011] [0.933] [0.000] [0.001] Log(Family Assets) 1.0041 1.0287* 0.9497** 0.9176*** [0.759] [0.065] [0.022] [0.001] Turnover 1.0608*** 1.0558*** 1.0434* 1.0074 [0.000] [0.000] [0.054] [0.837] Closed (DV) 0.9340 0.8969* 1.0534 0.9276 [0.183] [0.067] [0.543] [0.485] Load (DV) 1.0259 1.0216 1.0555 1.0495 [0.373] [0.517] [0.271] [0.435] Share class net cash flow 1.1713*** 1.2362*** 0.9349*** 0.8881* [0.000] [0.000] [0.008] [0.070] Fund net cash flow 0.9010 1.2687 0.7289** 0.5669* [0.376] [0.215] [0.028] [0.073] Net expense ratio 1.5950*** 1.6717*** 1.1940*** 1.1606 [0.000] [0.000] [0.000] [0.373] Share class fixed effect No No No No Year fixed effect Yes Yes Yes Yes Observations 45,499 45,499 45,449 45,499 R-squared 0.0600 0.0903 0.0312 0.116 25
Panel B: Expense ratio increases VARIABLES Logit Logit Logit Heckman Dependent variable Any expense increase Only expense ratio increase Only waiver removals Only waivers Top 30% return (DV) 1.0288 0.9003*** 1.2858*** 1.5081*** [0.352] [0.005] [0.000] [0.000] Bottom 30% return (DV) 1.1652*** 1.2636*** 0.9252 1.2501 [0.000] [0.000] [0.125] [0.197] Twelve month volatility 1.2277*** 1.2729*** 0.9957 1.3851 [0.000] [0.000] [0.886] [0.104] Young (< 5 years, DV) 0.9971 0.7904*** 1.4968*** 1.5377*** [0.942] [0.000] [0.000] [0.000] Middle-age (5-10 years, DV) 0.8915*** 0.8372*** 1.0685 0.8421 [0.000] [0.000] [0.192] [0.213] Log(TNA) 0.8875*** 0.9248*** 0.8483*** 0.7231** [0.000] [0.001] [0.000] [0.017] Log(Family Assets) 0.9796 0.9811 0.9718 0.9214** [0.134] [0.250] [0.188] [0.026] Turnover 1.0632*** 1.0361** 1.0983*** 1.2228*** [0.000] [0.032] [0.000] [0.004] Closed (DV) 0.9187 0.9789 0.8137** 0.6612*** [0.119] [0.736] [0.035] [0.006] Load (DV) 1.3873*** 1.5306*** 1.0752 1.8700* [0.000] [0.000] [0.128] [0.078] Share class net cash flow 0.9052*** 0.7723*** 1.0916*** 1.0013 [0.000] [0.000] [0.000] [0.991] Fund net cash flow 1.3181 1.7959** 0.9233 1.0915 [0.154] [0.043] [0.599] [0.901] Net expense ratio 1.0126 0.9634* 1.1156*** 1.3276*** [0.439] [0.055] [0.000] [0.000] Share class fixed effect No No No No Year fixed effect Yes Yes Yes Yes Observations 45,053 45,053 45,003 45,499 R-squared 0.0667 0.135 0.0378 0.211 26
Table 5 Investor response to expense waivers The table provides the results of estimating Eq. (2) in a panel regression setting with share class and year fixed effects, and errors clustered by share class. The dependent variable is a fund s implied net cash flow for the year as a percentage of beginning of year TNA. Any expense decrease (DV) is a dummy variable equal to one if there was any type of expense decrease, and New waiver (DV) is equal to one if Any expense decrease (DV) occurred because of a waiver. Any expense increase (DV) and Dropped waiver (DV) are defined analogously. All other variables and the sample are defined as in Table 4. All independent variables come from the prior calendar year. Retail and Institutional are defined using the CRSP Retail and Institutional indicators. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level. VARIABLES Full sample Retail Institutional only Any expense decrease (DV) 0.0391*** 0.0365*** 0.0427** [0.000] [0.000] [0.017] New waiver (DV) -0.0325*** -0.0195* -0.0655** [0.003] [0.091] [0.017] Any expense increase (DV) -0.0174*** -0.0200*** -0.0091 [0.010] [0.005] [0.598] Dropped waiver (DV) 0.0488*** 0.0443*** 0.0609** [0.000] [0.000] [0.015] Top 30% return (DV) 0.1162*** 0.1200*** 0.1002*** [0.000] [0.000] [0.000] Bottom 30% return (DV) -0.0886*** -0.0898*** -0.0821*** [0.000] [0.000] [0.000] Twelve month volatility 0.0197*** 0.0158*** 0.0275** [0.000] [0.001] [0.038] Young (< 5 years, DV) 0.1697*** 0.1941*** 0.0867*** [0.000] [0.000] [0.005] Middle-age (5-10 years, DV) 0.0280*** 0.0414*** -0.0200 [0.000] [0.000] [0.330] Ln(Size) -0.3851*** -0.3613*** -0.4658*** [0.000] [0.000] [0.000] Turnover 0.0017 0.0024-0.0048 [0.766] [0.704] [0.719] Net expense ratio -0.0155-0.0132-0.0114 [0.247] [0.350] [0.780] Load (DV) 0.0546*** 0.0791*** -0.0384 [0.000] [0.000] [0.143] Share class fixed effect Yes Yes Yes Year fixed effect Yes Yes Yes Observations 44,014 31,390 12,396 R-squared 0.206 0.225 0.188 27
Table 6 Waiver impact on managerial effort The table provides the results of estimating Eq. (3) in a panel regression setting. The panel regression uses style and year fixed effects and standard errors clustered by share class. All variables are defined the same as in Table 5. All independent variables come from the prior calendar year, while the dependent variable represents the size of the waiver as a percentage of the gross expense ratio. For ease of reading, control variables have been excluded from the table. Panel A includes turnover ratio as the dependent variable, while Panel B includes Active Share as the dependent variable. Panel A: Turnover as measure of effort Full Sample Retail Funds Institutional Funds Net Expense Ratio -0.0130-0.0116-0.00601 (-1.217) (-0.948) (-0.168) Waiver (DV) -0.120*** -0.125*** -0.0775 (-2.995) (-3.500) (-0.512) Waiver (%) -0.136*** -0.170*** -0.0605 (-3.227) (-2.961) (-0.944) Net Expense Ratio*Waiver (%) 0.109*** 0.150*** -0.0639 (5.420) (4.900) (-1.157) # Years Waivers Offered 0.00369 0.00912** -0.00109 (1.087) (2.273) (-0.210) Square of # Years Waivers Offered 0.00311 0.00826* -0.00110 (0.891) (1.942) (-0.208) Style Fixed Effects Yes Yes Yes Year Fixed Effects Yes Yes Yes Additional controls Yes Yes Yes Observations 10408 6845 2156 R2 0.727 0.722 0.703 Standard Errors Clustered by Fund & Year Yes Yes Yes 28
Panel B: Active share as a measure of managerial effort Full Sample Retail Funds Institutional Funds Net Expense Ratio 0.008*** 0.009*** 0.0006 (4.64) (3.01) (0.11) Waiver (DV) 0.038*** 0.035** 0.004 (2.93) (2.40) (0.13) Waiver (%) 0.007 0.019-0.030 (0.47) (0.83) (-0.65) Net Expense Ratio*Waiver (%) -0.004-0.010 0.018 (-0.35) (-0.59) (0.42) # Years Waivers Offered 0.0002 0.0003-0.0006 (0.33) (0.40) (-0.45) Square of # Years Waivers Offered -0.00002-0.00000 0.00000 (-0.34) (-0.15) (0.03) Style Fixed Effects Yes Yes Yes Year Fixed Effects Yes Yes Yes Additional controls Yes Yes Yes Observations 10,049 6,638 2,098 R2 0.892 0.893 0.907 Standard Errors Clustered by Fund & Year Yes Yes Yes 29