The Only Fund in Town? Geographic Segmentation in the US Mutual Fund Industry.
The question as to whether competitive forces in the mutual fund industry are sufficient to constrain managers from charging excessive fees is an important one as there is considerable evidence that mutual fund investors do not sufficiently respond to product quality (performance) or price (fees) when making their portfolio decisions (Christoffersen and Musto, 2002; Elton, Gruber and Busse, 2004; Gil-Bazo and Ruiz-Verdu, 2009). For instance, active funds charge higher fees and have lower performance than passive index funds (Fama and French, 2010). Even among active funds, Gil-Bazo and Ruiz-Verdu (2009) find that funds with poorer performance actually charge higher fees than their active competitors. These facts suggest that funds engage in monopolistic competition, whereby funds produce differentiated products (at least from the perspective of the consumer) and consumers behave as if they have some degree of brand loyalty and/or high search or switching costs. The fact that mutual fund flows are shown to be fairly insensitive to poor performance supports this view (Christoffersen and Xu, 2017). In addition, Hortacsu and Syverson (2004) determine that despite offering the same underlying portfolio, there is great fee dispersion among competing S&P 500 Index funds. They attribute this dispersion to search costs and differentiation along nonportfolio fund family characteristics.
In this paper, we argue that an important dimension of fund differentiation could be the fund's geographic location, which could determine its likely base of potential investors, as well as its relative set of competitors. If this is true, then geography may play an important role in determining the fees mutual funds charge their investors. Our argument is based on the premise that if investors exhibit a local bias in their portfolio allocation decisions, they may disproportionately favor local versus distant mutual funds. The fact that investors exhibit a strong bias toward investments that are located geographically close to them has been well documented in the literature. For example, investors prefer domestic versus foreign stocks (French and Poterba, 1991), mutual funds disproportionately invest in stocks headquartered nearby (Coval and Moskowitz, 2001), and banks are more likely to lend to local firms (Petersen and Rajan, 2002).
We hypothesize that investors' local bias could result in some degree of geographic segmentation in the mutual fund market. Thus, the intensity of fund competition may not only be determined by market-wide factors, such as the total number of funds in operation, but also by local factors, such as the number of funds that operate nearby. If investors prefer local funds and there are few funds located nearby (e.g., Birmingham, AL), then those funds should have relatively more local market power than funds that are located in regions with more local competitors (e.g., Boston, MA).
We expect that increased local market power should enable funds to earn greater profits (i.e., charge higher management fees) than funds operating in more crowded locations. However, the effect of competition on the total price mutual funds charge (i.e., total expenses) is theoretically ambiguous. That is, the total price a producer (funds) charges consumers (investors) reflects both the cost of producing the product, as well as the profit realized by the producer. If greater competition forces funds to expend greater resources on marketing in order to attract investors, it would raise the fund's production costs and could, as such, raise prices for investors.'
In a similar analysis of the US equity market. Hong, Kubik, and Stein (2008) reason that if investors exhibit a preference for local securities, then the market for those securities could become geographically segmented. They argue that the price of stocks headquartered in regions with relatively few companies could be driven up by the fact that investors have few other local choices to add to their portfolio, a phenomenon they refer to as the "only game in town effect." In this paper, we test for such an effect in the mutual fund industry. Using a sample of 2,682 domestic equity mutual funds from 2000 to 2015, we examine whether mutual fund prices (fees) and quality (performance) are, in part, determined by the degree of local competition those funds face. We define our local marketplaces using US Census Metropolitan Statistical Areas (MSAs), although our results also hold at the state level. We measure the degree of local competition as the total number of funds or fund families in the local area.
We begin by documenting evidence of local bias by mutual fund investors. Our analysis follows that of Sialm, Sun, and Zheng (2014) who document local flow comovement in the hedge fund industry as evidence of local bias in hedge fund investment. They argue that if investors have a local bias in their fund allocations, then flows should be partly driven by shocks to local investment leading to positive comovement in the flows of local funds. We find that a fund's flows are significantly and positively related to the flows of funds in other families in the same local area, even after controlling for performance, local performance, and style effects. This suggests that flows to funds follow a common local factor structure that would be consistent with local bias. While not the main focus of our paper, this finding represents a contribution to the literature as we know of no other study that documents local bias at the mutual fund level.
After establishing evidence of local bias through excess flow comovement, we then examine the relation between mutual fund fees and the degree of competition from other funds in their region. We begin by focusing our attention on the association between local competition and management fees, as the fund manager (i.e., the investment advisor) derives its profit from the management fee, while the 12b-l fees and other expenses are essentially pass-through expenses used to market and operate the fund. (2) We find that management fees are strongly and negatively related to the degree of local fund competition. This relation is robust to a variety of different specifications, subsamples, and ways of measuring geographic competition. This suggests that fund profits deteriorate as local competition intensifies.
Further, as we examine other expense categories, an interesting pattern emerges. Specifically, funds in more competitive areas actually charge higher 12b-l fees and higher other expenses. We also find some evidence that funds in more competitive regions charge higher sales loads than funds in less competitive regions. These results are consistent with the idea that more intense local competition can lead to increased selling expenses as more funds fight for the attention of local investors in order to preserve market share. Although increased local competition can constrain fund profits (by way of the decreased management fees), it also seems to create a negative externality in the fund market by causing funds to spend more to attract investors ultimately reducing the benefit of competition from the perspective investors. Overall, we find some evidence suggesting that total expense ratios are actually slightly higher in more competitive regions. However, funds in competitive regions do not outperform similar funds in less competitive regions suggesting their higher total expenses do not reflect greater managerial skill.
Our paper adds to the literature on mutual fund competition and its impact on fees and asset growth. Our results documenting the countervailing effects of competition on management and sales-related fees are consistent with Wahal and Wang (2011), who find that incumbent funds cut management fees, but increase sales-related expenses when similar funds enter the national market. We contribute to this line of research by analyzing the impact of local competition on fees. In addition, our paper contributes to a growing literature, particularly in the field of banking, suggesting that local bias can result in geographically segmented capital markets. For instance, banks tend to realize a higher proportion of their deposits from local depositors and, in turn, make a majority of their loans locally (Petersen and Rajan, 2002; Becker, 2007). This creates locally segmented banking markets, whereby more competitive local banking markets lead to lower prices and greater availability of credit for small firms (Degryse and Ongena, 2005; Cetorelli and Strahan, 2006). Our paper contributes to this literature by demonstrating that the mutual fund industry also exhibits some of the characteristics of a geographically segmented industry, and that the expenses mutual fund investors pay are partly determined by the degree of local competition.
I. Data and Sample Construction
We construct our sample by considering all US equity mutual funds from 2000 to 2015. We begin with the year 2000 since reliable fimd location information is available in the Center for Research in Security Prices (CRSP) mutual fund database beginning at the end of 1999. We filter our funds using the CRSP objective code, which distills the information in various other codes (Lipper, Strategic Insight, etc.) into a single four-character code that summarizes the objective of the particular fund. We keep all domestic equity funds (first two characters of the CRSP objective code = "ED") and eliminate all funds where the fourth character is "H" or "S" indicating the funds are hedged portfolios or short-bias portfolios.
We take a number of steps to ensure that we correctly identify the geographic location of each fund in our sample. We start by identifying the fund's Metropolitan Statistical Area (MSA) from the zip code given for the fund in the CRSP database. We then supplement the CRSP location data with information from mutual fund NSAR filings (we thank Jon Fulkerson and Chris Clifford for providing this data). First, we eliminate funds where the NSAR indicates the use of subadvisors, as these subadvisors may be located in a different geographic area from the fund itself We then use data from NSAR filings to identify all the zip codes associated with the advisors or administrators of a particular fund. If we find any disagreement among the MSA's implied by the zip codes in the NSAR filings and the CRSP database, we exclude the fund from our sample. This provides an extra measure of confidence that we are correctly identifying the location of the funds we study. We use the WFICN key from the MFLINKS database to aggregate individual share classes to the fund level and eliminate funds for which the WFICN cannot be matched. Finally, we require the fund to have at least $5 million in assets in a given year to be included in that year. We are left with a total of 2,682 unique funds from 2000 to 2015.
Our analysis is at the overall fund level. For fund level returns, expense ratios, etc., we calculate a weighted average measure with weights based on the individual total net assets of each share class. In order to focus on funds where local bias is likely to be more prevalent, we filter out institutional share classes and only include retail share classes in our analysis, however our results are not sensitive to this restriction. To calculate net fund flows each month, we follow the literature by calculating:
[mathematical expression not reproducible] (1) where [TNA.sub.i,t] is the total net assets of fund i in period t and [R.sub.i,t] is the fund's return over the period from t-1 to t. Since net flows are plagued by outliers, we winsorize observations that are in the top or bottom 2.5% of fund flows.
For mutual fund fees, we also calculate similar asset-weighted measures of fees across classes of a given fund. We first consider the management fee, which is the direct compensation to the investment advisor responsible for managing the fund. We also consider the level of 12b-l fees, which are fees charged for marketing and distribution of the fund. Finally, we consider the overall aggregate expense ratio, which is a measure of the total fees paid by investors in a given year, and "other expenses" that include any portion of the total expense ratio outside of the management fee or 12b-l fee. Examples of these other expenses include shareholder service expenses, legal expenses, and accounting expenses. Although we consider all expense measures in our subsequent analysis, we focus much of our attention on the management fee as it reflects compensation to the manager rather than pass-through expenses used to operate and sell the fund to investors.
Table 1 presents summary statistics for the funds in our sample. The average management fee is 0.66%, yet the standard deviation is 0.40% reflecting considerable cross-sectional variation across the funds in our sample. We find that the average fund's expense ratio is about 1.30% suggesting management fees, on average, make up about 51 % of total expenses paid by investors. 12b-l fees, in contrast, make up a smaller proportion of total expenses and average around 25 basis points. The mean fund size is over $1.3 billion with a median size of $187 million.
II. Evidence of Local Bias
Our analysis of local competition is predicated upon the idea that investors exhibit local bias in their mutual fund allocation decisions, which could induce a degree of local segmentation in the mutual fund market. There are many explanations as to why investors may exhibit a preference for local funds. For instance, local bias could be a symptom of familiarity bias whereby investors may be more comfortable making risky decisions in contexts where they consider themselves knowledgeable or more familiar (Heath and Tversky, 1991). Fund investors may be more familiar with local funds and favor them simply because of their increased awareness.
This awareness effect is more likely to impact retail investors who, due to their relative lack of financial sophistication, may be relatively less able to overcome their behavioral biases than institutional investors. Moreover, retail investors have fewer resources and less time to spend considering the set of possible funds making them more likely to simply select local funds they are aware of out of convenience. Thus, we focus our attention on retail funds as they are the funds that are more likely to be susceptible to the effects of local bias. (3)
Though local bias has been documented in many investment contexts, we know of no paper that provides specific evidence of local bias in the mutual fund market.'' This lack of evidence may be partly due to the difficulty in obtaining data on the mutual fund allocations of individual investors. While we lack investor-level data, in this section, we provide suggestive evidence of a local bias in mutual fund investment by examining the comovement of fund flows at the local level. Our analysis follows that of Sialm et al. (2014) who confirm local flow comovement in the hedge fund industry as evidence of local bias in hedge fund investment. They argue that if investors have a local bias in their fund allocations, then flows should be partly driven by shocks to local investment leading to positive comovement in the flows of local funds.
Following Sialm et al. (2014), we estimate local flow comovement in a regression framework. Specifically, we regress a fund's monthly percentage flows on the equally weighted average flows of the funds located in the same area (Local Flow). Since funds from the same family could have correlated flows regardless of local conditions, we exclude same family funds from the calculation of the Local Flow. We estimate fund-level time-series regressions for each fund in our sample and report the average coefficients and t-statistics in Table II. To define Local Flow, we use funds from the same Metropolitan Statistical Area (MSA). We obtain similar results if we define Local Flow at the state level. We present standardized coefficients in the table such that the coefficients can be interpreted as the average standard deviation change in flows for a standard deviation change in the independent variable.
Model 1 reports the results from the univariate model of fund flows regressed on Local Flow. The coefficient on Local Flow is positive and statistically significant (r-stat = 7.50) implying that, on average, a fund's flows positively comove with the flows of other local funds. Economically speaking, a one standard deviation increase in Local Flow corresponds to a 0.0883 standard deviation increase in fund flows. To account for the possibility that local flow comovement could be driven by local funds following similar investment strategies, in Model 2, we control for flows to funds in the same investment style (Style Flow). Style Flow is the equally weighted average percentage flow of all mutual funds in the same CRSP objective code, excluding funds in the same family. The coefficient on Style Flow is also positive and significant, yet the Local Flow coefficient remains positive and significant. In Model 3, we include Market Flow, which is the equally weighted average percentage flow of all mutual funds, excluding funds in the same family, to control for market-wide flow trends throughout the sample period. Including this variable has little impact on the Local Flow coefficient.
Finally, in Model 4, we include two additional sets of return-based controls: three monthly lags of the fund's trailing monthly returns, as well as three monthly lags of monthly Local Returns defined as the equally weighted average return from other funds in the fund's MSA. We include the former to account for the well-known relation between flows and fund performance, and the latter to account for the possibility that local funds invest in similar stocks (perhaps due to their own local bias), which could cause commonalities in flows based on common return shocks rather than local investor bias in mutual fund allocations. As expected, own fund returns are positively related to flows. However, Local Returns are negatively related to flows. One possible explanation for this is that if investors are locally biased, they may benchmark the fund's performance against the performance of other local funds when making investment allocations. This could induce a negative relation between the benchmark returns and flows (when also controlling for own fund returns). All that said, the Local Flow coefficient remains positive and significant in the presence of these return controls. Taken together, the results in Table U suggest fund flows are partly driven by common shocks at the local level. Though indirect, this evidence is suggestive of a local bias in mutual fund investment.
III. Local Competition and Fund Fees
In this section, we examine whether geographic competition is related to fund fees. Note that total mutual fund expenses (expense ratio) consists of both the fund manager's profit (which comes from the management fee), as well as its operating costs (i.e., marketing and distribution costs reflected in 12b-l fees and other administrative expenses). The predictions of the effect of local competition vary for these different components of total fund expenses. Funds operating in less competitive areas should have a greater ability to extract higher profits, but should not necessarily encounter higher operating costs. In fact, funds in more competitive areas are likely to spend more on advertising and other sales related activities as they fight for the attention of local investors. Thus, we expect that increased local competition should have a negative effect on the fund's management fees and a positive effect on the fund's 12b-l fees, which reflect the cost of advertising and client acquisition. We do not have a specific prediction for other expenses, which could be higher or lower depending upon how fund competition impacts the relative supply and demand for fund service providers, such as lawyers, accountants, and fund custodians. For example, increased competition could lead to increased demand for fund service providers, thereby driving up the price of their services. However, it could also be that these providers are relatively scarce in areas with fewer funds, which could make their services more expensive in more remote areas. Due to the countervailing predictions for the components of the expense ratio, the prediction of the relation between competition and total expense ratios is ambiguous.
Our primary measure of local competition is the number of funds operating in the same MSA (though our analyses are largely similar when using the number of fund families). Figure 1 illustrates the average number of funds operating in each of the 374 MSAs defined by the US Census. In our sample of funds, 130 of these MSA's are represented. Not surprisingly, the majority of funds are located in very competitive financial centers, such as New York and Boston, that average well over 100 funds per year. However, the majority of MSAs are not very competitive. Seventy-three percent of MSAs have, at most, 10 funds in any year in our sample.
Figure 2 provides a general perspective as to how the level of mutual fund competition in an area relates to fees. We separate the funds into six groups based on the number of funds in the MSA (< = 5, 6-25, 26-50. 51-100. 101-250. and > 250) and present group averages of the total expense ratio, as well as its various components. We begin by examining our primary fee measure of interest, which is the fund's management fee. We see that there is a nearly monotonic decrease in management fees as geographic competition increases. The average management fee in MSA's with between five or fewer funds is 0.83%, while the average is 0.61% in areas with over 250 funds. This evidence suggests evidence that funds that operate in less competitive areas can extract greater rents in the form of higher management fees. We find nearly the opposite pattern with 12b-l fees, although the pattern is not quite as consistent. Funds in the most competitive MSA's have average 12b-l fees of 0.30%, while funds in the least competitive MSA's have average 12b-l fees of just over 0.17%. Since 12b-l fees are used for advertising and client acquisition costs, it suggests that funds must spend more to attract investor clientele in more competitive areas. Other expenses (i.e., those expenses not included in management fees or 12b-l fees) reflect a V-shaped pattern suggesting that other administrative costs are highest in the least and most competitive areas. When we consider the total expense ratio, we find some evidence that expense ratios are lower in more competitive areas. Of course, at this point, all of this is merely suggestive and unconditional evidence based on univariate statistics. In other words, we are not holding the various fund characteristics constant that also likely differ across MSAs. These characteristics that could be related to fees include fund size, performance, and risk characteristics, as well as fund family characteristics. In what follows, we examine this evidence more systematically using a regression framework.
A. Management Fee Regressions
We now examine the main empirical question of the paper: do mutual funds in less competitive areas charge higher management fees? We test the hypothesis by running regressions of the form:
[mathematical expression not reproducible] (2) where Fe[e.sub.i,t] represents the management fee charged by mutual fund i in year t, and where Control[s.sub.i,t] represents common controls shown in the literature to be related to mutual fund fee structure, such as lagged values of fund return, flow, total net assets, family total net assets, number of funds in the family, fund age, volatility of fund returns over the past year, and turnover of the fund. We also include controls for the fund's Fama and French (1992) four factor betas and fixed effects for the fund's objective code variable from CRSP (crsp_obj_cd) to capture any variation in management fees across different styles that is not captured in other variables. For our Competition Measure, we consider both the log of the number of mutual funds and the log of the number of fund families located in the same metropolitan statistical area (MSA). (5) We run pooled panel regressions with year fixed effects and standard errors clustered by fund family and year.
We present results for the management fee regressions in Table III. In terms of the control variables, lagged ftind returns and net assets are positively related to management fees, while family net assets and the number of funds in the same family are negatively related to fees. Funds with more volatile returns and with higher turnover are associated with higher management fees. From inspecting the four-factor betas, it appears that momentum is the only significant factor. Funds with higher exposure to momentum seem to charge lower management fees. Finally, as expected, we find that index funds have management fees that are about 33 basis points lower, on average, than nonindex funds.
Turning our attention to the main variables of interest, we find a result that is strong and consistent across both of our competition variables. As geographic competition increases, the management fee decreases and all coefficient estimates are significant at the 1% level. For instance, in Model I the coefficient on Ln(# Funds-MSA) is -0.0272 (t-stat = -3.59). This estimate suggests that moving from the 25th to the 75th percentile of # Funds-MSA corresponds to a roughly 7.5% decrease in management fees (decrease of about 5.5 bps).We find similar results if we instead consider the number of fund families per MSA, which also has a coefficient estimate that is significant at the 1% level.
The results thus far indicate a strong relationship between management fees and geographic competition. In Table IV, we provide a series of additional robustness checks on our main results. We use the number of funds in the MSA as our measure of local competition in each model, but the results are all qualitatively similar if we use the number of families instead. In Model 1, we present results excluding the most competitive MSAs (i.e., those with more than 250 funds). In addition to dominating the mutual fund market, these regions are also home to some of the largest and lowest fee fund providers, such as Fidelity. Interestingly, we find very similar results when we exclude these areas suggesting our results are not driven by just the dominant mutual fund markets. (6) In Model 2, we drop the least competitive MSAs, those with less than 25 funds, and continue to find very similar results. Thus, our results also do not appear to be driven by the most remote MSAs with very few competing funds. In Model 3 of Table IV, we include state fixed effects in the model. Here, we are testing whether there is variation in fund management fees that is related to geographic competition, even after controlling for geographic variation at the state level. We again find that our measure of MSA-level competition is negative and statistically significant at the 5% level.
Another concern that we address in Table IV relates to standard error calculations. Thus far, we have clustered our standard errors by both fund family and year. This accounts for the fact that fees are not independent within a fund family and the fact that fees follow time trends and may not be independent within years. However, one could be concerned that our measure of competition is measured at the geographic level and, as such, are the same across funds in the same area. This could lead to excess within-MSA correlations that are not accounted for in our standard error calculations. To address this concern, we re-estimate our main model clustering for MSA and year (Model 4). We find that the results are still significant at conventional levels with a t-stat of-2.61.
In Table V, we examine the effect of competition on management fees across subperiods within our sample. With improvements in investor education, technology, and increasing integration of financial markets, it could be that the effect we are observing is only present in the early years of our sample. To investigate this question, we split our sample into two subsamples: 2000-2007 and 2008-2015. The results suggest that the relation between geographic competition and management fees is very strong and statistically significant in both halves of the sample. (7)
In other robustness tests (not reported), we also consider an alternative form of our management fee regressions and collapse our dataset down to the MSA-year level. We calculate the average management fee across all funds in an MSA in a given year, and also calculate the averages of all the control variables. Whether we calculate equally weighted or value-weighted averages, we find a strong relation between geographic competition in an area and management fees. Our results are also similar if we drop the top 20 fund families in terms of family size. This helps to alleviate the concern that our results may be driven by a handful of fund families with many different mutual funds under their umbrella.
B. 12b-1 Fees, Other Expenses, and Total Expense Ratios
After documenting strong evidence that management fees are significantly and negatively related to local competition, we now turn our attention to the relation between competition and the other expense categories that mutual funds charge their investors (12b-l fees and other expenses), as well as the total expenses investors pay (expense ratio). Regression results for each expense category are presented in Table VI. We use the log of the number of funds in the MSA as our measure of competition and include the same control variables that we used in our management fee regressions. We find qualitatively similar results using the number of families or examining location at the state level.
In Model 1 of Table VI, we consider whether geographic competition is related to the level of 12b-l fees. Outside of management fees, 12b-1 fees are often the largest measurable component of the expense ratio. Wahal and Wang (2011) find that incumbent mutual funds often respond to (nationwide) competition from entrants by waiving management fees and increasing 12b-1 fees thereby hoping to increase awareness of the fund and gain traction with investors in the face of new funds entering the market. Here the results are opposite of the management fee results. The level of 12b-l fees is positively and statistically significantly related to the level of competition in the fund's MSA. This is consistent with the findings of Wahal and Wang (2011) and indicates that funds in more competitive areas spend more on marketing and distribution than funds in less competitive areas.
In Model 2, we examine the "other expenses" category, which is defined as the total expense ratio-management fee-12b-1 fee. This category essentially captures the costs of fund administration that are not captured in the management fee or the 12b-1 fee. According to the SEC, some examples of costs included in the other expenses category include custodial expenses, legal expenses, and shareholder service expenses. The results in Model 2 reveal that other expenses are significantly higher in more competitive locations. One potential explanation for this is that legal and other administrative providers charge higher prices in more competitive areas, perhaps because there is greater demand for their services. It could also be that in more competitive areas, there is more shareholder turnover that could give rise to greater shareholder servicing costs.
Finally, we examine the relation between local competition and total fund expenses (expense ratio). Given that the management fee is negatively related to competition, while the 12b-1 and other expenses are positively related to competition, it is unclear what the net effect of local competition is for investors in terms of the total expenses they pay to mutual fund managers. Model 3 of Table VI presents the results of regressions where the dependent variable is the fund's total expense ratio. The coefficient on Ln(# Funds-MSA) is positive and marginally significant (t-stat = 1.68). This suggests that the higher management fees charged by funds in less competitive areas are more than offset by their lower 12b-1 fees and other expenses. This result is a bit surprising given that the univariate results displayed in Figure 2 suggested a negative relation between expense ratios and competition. In untabulated regressions, we confirm that the univariate relation between total expenses and Ln(# Funds-MSA) is negative, but not statistically significant. The fact that the relation between competition and expense ratios flips sign when we include our set of control variables reflects that other fund and family characteristics are quite different across levels of fund competition.
To examine this issue in more detail, in Table VII, we present the means of all of the control variables in our regressions by levels of MSA competitiveness: <5 funds, 6-25 funds, 26-50 funds, 51-100 funds, 101-250 funds, and >250 funds. We also present results in the final column for tests of statistical significance of the difference in means between the least competitive group <5 funds) and the most competitive group (>250 funds). In examining these results, it quickly becomes clear that there are tremendous differences in several fund characteristics across varying levels of local competition. In particular, fund size generally increases as the number of funds in an MSA increases. Fund families, both in terms of assets under management and the number of funds in the family, are dramatically larger in the top two groups of MSA's compared to any of the smaller MSA groups. In addition, funds in more competitive areas are significantly older than funds in less competitive areas. Finally, funds in more competitive MSA's have different risk exposures (i.e., more exposure to market, large cap, growth, and momentum). However, this difference is not as stark in economic terms as the differences in fund and family size noted above.
C. Propensity Score Matching
Table VII confirms that there are significant differences in the characteristics of funds in remote locations versus those in competitive locations. Although we have previously controlled for these characteristics in linear regressions, these differences in characteristics could have a nonlinear confounding effect on the variables of interest in our study. Thus, in order to ensure that we are capturing the effect of competition differences and not differences in other fund characteristics across MSAs, we now employ a Propensity Score Matching (PSM) technique where we match funds in remote locations to similar funds in competitive locations in order to balance the covariates between the two groups.
In constructing our sample for the PSM approach, we recognize that some fund groups only exist in financial centers and, as such, are not comparable to remote funds. Funds from very large families are almost always located in financial centers. Thus, there is no good way to find a comparable fund in a remote location. To ensure that we can find truly comparable matched pairs of funds from less competitive and more competitive areas, we remove funds from very large families (greater than $100 billion in assets). In addition, we focus on less competitive MSAs that have between 26 and 50 funds to avoid any peculiarities related to extremely sparse MSAs. (8 )We compare these remote funds to funds in competitive MSAs, which we define as having 100 or more funds. We then match each remote fund to a competitive fund with the closest propensity score estimated from a logistic regression predicting the likelihood a fund is located in a remote location as a function of our full set of fund characteristics. By matching in this way, we are able to achieve covariate balance between the samples of remote and competitive funds. The results of covariate balancing tests are shown in Table VIII. Only one of the characteristics is significantly different between the matched samples. After matching, the average fund in less competitive MSAs is larger than the average matched fund in more competitive MSAs. However, the difference is not economically meaningful ($973 million vs. $839 million). Moreover, the likelihood ratio test of the joint insignificance of all covariates is insignificant suggesting good overall covariate balance. We use these matched samples to test for univariate differences in fees, loads, and fund alphas between remote and competitive funds in Tables IX and X.
The results of our PSM tests generally line up with our earlier fee regressions. Funds in competitive locations have significantly lower management fees, but have significantly higher 12b-l fees and other expenses than funds located in remote locations. For example, the first row of Panel A in Table IX indicates that the average management fee of funds in the less competitive areas is 0.73% compared to 0.67% for funds in the more competitive areas. This difference is significant at the 1% level. This is roughly offset by the other expenses category, where funds in more competitive areas have significantly higher fees by about seven basis points. 12b-l fees are also significantly higher by about 10 basis points in the more competitive areas. When all of the components are combined in the total expense ratio, the average expense ratio in the competitive areas is 1.32% compared to 1.21% in the remote areas. This difference is also significant at the 1% level.
To gain a more complete picture of the relation between geographic competition and expenses, we consider fund loads in Panel B of Table IX. To calculate the load for each fund, we take the maximum load (both front and rear load) for each share class. We then calculate the front and rear load for each fund as the asset-weighted average of these maximum loads across share classes. Total load is the sum of these front and rear loads.
If funds in more competitive areas have to spend more on advertising/sales in order to attract funds, then it may be the case that loads are higher in these areas. Our results are consistent with this conjecture. Both front and rear loads are, on average, about 13-14 basis points higher in more competitive areas. This is consistent with the earlier result on 12b-l fees and suggests that local competition may lead to higher total costs for investors, all else being equal.
However, it is important to acknowledge an alternative hypothesis, which is that 12b-1 fees and sales loads may be higher in more competitive regions (i.e., financial centers) as those areas have higher labor costs or costs of advertising leading to higher unit costs for sales-related activities. We argue that one of the reasons that sales costs could be higher in financial centers is that more funds are competing for those resources, thus driving up their price. In this sense, the increased costs would still be a by-product of fund competition. Of course there may be noncompetitive reasons why sales costs are higher in financial centers, which could affect how we interpret the higher sales expenses we find. Although we cannot rule out this alternative hypothesis entirely, we do think that the overall pattern of results we document make it less likely to be the main factor driving our results. In particular, although labor costs are surely higher in financial centers like New York and Boston, they should also be higher for portfolio managers and analysts, which are expenses that impact the management fee. However, we find lower management fees for funds in the most competitive regions. The fact that we find both lower management fees and higher sales-related expenses in more competitive regions supports our competition hypothesis and is inconsistent with the notion that our results are only driven by higher labor costs.
Thus far, we have determined that funds in less competitive areas tend to charge higher management fees than funds in more competitive areas. However, once we consider other fees, it appears that (at least for a subset of funds) mutual funds located in more competitive areas charge total higher expense ratios. At this point, it is not clear from these results whether managers in the more competitive areas are justified in charging higher fees. If they outperform other funds in a manner that justifies the higher expenses, then our results may be perfectly consistent with a competitive market for mutual funds in the United States. To shed light on this question, for each fund we calculate the capital asset pricing model (CAPM), Fama and French (1992) three-factor and four-factor alphas (Fama-French three factor plus momentum) over rolling three-year windows. In Table X, we present matched sample comparisons of remote MSA average alphas with competitive MSA alphas. We use the same propensity score matching procedure discussed in Tables VIII and IX. Regardless as to how we measure alpha, there is no significant difference in performance between the two groups of funds. (9) Keep in mind however, that this is a conditional finding comparing similar funds across regions. There is evidence in the literature that, unconditionally, funds in more competitive regions outperform funds in less competitive regions (Christoffersen and Sarkissian, 2009). Overall, our matched sample results do not suggest that funds in competitive MSA's consistently outperform similar funds in less competitive MS As in a manner that would justify their higher expense ratios. (10)
In this paper, we test the hypothesis that the local bias of mutual fund investors could cause geographic segmentation in the mutual fund market. Consistent with this hypothesis, we document an interesting combination of results concerning the relation between geographic competition and mutual fund fees. We find that funds in less competitive areas charge higher management fees, but lower 12b-l fees and overall expense ratios compared to funds in more competitive areas. In fact, our matched sample approach suggests that when compared to similar funds in less competitive regions, overall expense ratios are 11 basis points higher for funds in competitive MSA's. However, we should caution that this does not mean investors in funds located in financial centers necessarily pay higher fees on average. In fact, unconditional average expense ratios are lower in the most competitive regions as competitive regions are comprised of larger funds that charge lower expense ratios on average, likely due to economies of scale effects. Rather, our results suggest that investors are paying higher expenses for funds located in competitive regions that are otherwise similar (in terms of fund and family characteristics) to funds in less competitive regions.
Our results are also important in understanding the nature of competition in the mutual fund industry. Consistent with the results of Wahal and Wang (2011), we find that competition is related to lower management fees, but higher 12b-l fees. Thus, it seems that competition is effective in driving down fund profits. However, our finding that the total expense ratios of funds in competitive areas are higher than comparable funds in less competitive areas suggests that, in some cases, the cost of attracting customers (through both 12b-1 fees and loads) may more than offset the cost reduction that comes from lower management fees. In other words, increased competition creates a negative externality by way of increased sales expenses that can actually increase the total expenses paid by investors. Thus, our results could help inform investors and regulators about the use of 12b-l and other distribution-related costs. These fees have long been under the microscope as a category of fees that are less salient and more likely to be abused and do not necessarily lead to better investor outcomes.
Bailey, W., A. Kumar, and D. Ng, 2011, "Behavioral Biases of Mutual Fund Investors," Journal of Financial Economics 102, 1-27.
Becker, B., 2007, "Geographical Segmentation of US Capital Markets," Journal of Financial Economics 85, 151-178.
Cetorelli, N. and P.E. Strahan, 2006, "Finance as a Barrier to Entry: Bank Competition and Industry Structure in Local US Markets," Journal of Finance 61, 437-161.
Christoffersen, S.E.K. and D.K Musto, 2002, "Demand Curves and the Pricing of Money Management," Review of Financial Studies 15, 1499-1524.
Christoffersen, S.E.K. and S. Sarkissian, 2009, "City Size and Fund Performance," Journal of Financial Economics 92, 252-275.
Christoffersen, Susan E.K., and Haoyu Xu, 2017, "Investor Attrition and Fund Flows in Mutual Funds," Journal of Financial and Quantitative Analysis 52, 867-893.
Cooper, M., M. Hailing, and W. Yang, 2015, "The Mutual Fund Fee Puzzle," University of Utah Working paper.
Coval, J.D. and T.J. Moskowitz, 2001, "The Geography of Investment: Informed Trading and Asset Prices," Journal of Political Economy 109, 811-841.
Degryse, H. and S. Ongena, 2005, "Distance, Lending Relationships, and Competition," Journal of Finance 60,231-266.
Elton, E.J., M.J. Gruber, and J.A. Busse, 2004, "Are Investors Rational? Choices Among Index Funds," Journal of Finance 59, 261-288.
Fama, E.F. and K.R. French, 1992, "The Cross-Section of Expected Stock Returns," Journal of Finance 47, 427-465.
Fama, E.F. and K.R. French, 2010, "Luck versus Skill in the Cross-Section of Mutual Fund Returns," Journal of Finance 65, 1915-1947.
French, K.R. and J.M. Poterba, 1991, "Investor Diversification and International Equity Markets," National Bureau of Economic Research Working paper.
Gil-Bazo, J. and P. Ruiz-Verdu, 2009, "The Relation Between Price and Performance in the Mutual Fund Industry," Journal of Finance 64, 2153-2183.
Heath, C. and A. Tversky, 1991, "Preference and Belief: Ambiguity and Competence in Choice Under Uncertainty," Journal of Risk and Uncertainty 4, 5-28.
Hong, H., J.D. Kubik, and J.C. Stein, 2008, "The Only Game in Town: Stock-Price Consequences of Local Bias," Journal of Financial Economics 90, 20-37.
Hortacsu, A. and C. Syverson, 2004, "Product Differentiation, Search Costs, and Competition in the Mutual Fund Industry: A Case Study of S&P 500 Index Funds," Quarterly Journal of Economics 119, 403-456.
Petersen, M.A. and R.G. Rajan, 2002, "Does Distance Still Matter? The Information Revolution in Small Business Lending," Journal of Finance 57, 2533-2570.
Satterthwaite, M.A., 1979, "Consumer Information, Equilibrium Industry Price, and the Number of Sellers," Bell Journal of Economics 10, 483-502.
Sialm, C, Z. Sun, and L. Zheng, 2014, "Home Bias and Local Contagion: Evidence from Funds of Hedge Funds," University of Texas Working paper.
Stiglitz, J.E., 1987, "Competition and the Number of Firms in a Market: Are Duopolies More Competitive Than Atomistic Markets?" Journal of Political Economy 95, 1041-1061.
Wahal, S. and A.Y. Wang, 2011, "Competition Among Mutual Funds," Journal of Financial Economics 99, 40-59.
Jesse A. Ellis and Shane Underwood (*)
Much of this work was completed while Underwood was at the University of Alabama. We are grateful to Sergey Chernenko, Wenhao Yang, participants at the 2014 FMA Conference, and seminar participants at the University of Tennessee, Mississippi State University, and Baylor University for helpful comments and suggestions. All errors are our own.
(*) Jesse A. Ellis is an Associate Professor in the Poole College of Management at North Carolina State University in Raleigh, NC. Shane Underwood is an Associate Professor in the Hankamer School of Business at Baylor University in Waco, TX.
(1) There is a great deal of economic theory that suggests various scenarios whereby monopolistic competition could lead to a positive relation between competition and prices, particularly when customers face search or switching costs (Satterthwaite, 1979; Stiglitz, 1987; Hortagsu and Syverson, 2004). If investors have switching or search costs (i.e., perhaps they have brand loyalty or are so uninformed about the set of alternative mutual funds that it is particularly costly for them to search for and evaluate alternative providers), then funds have an incentive to expend resources on advertising and distribution activities in order to capture rents from a set of loyal investors. This process could lead to higher total prices.
(2) Of course, the entire management fee is not profit to the manager (as the manager has other expenses to cover), but it does represent the source of funds that could fall to the manager's bottom line. See https://www.sec.gov/fast-answers/answersmffeeshtm.htmWmanagement for more details on mutual fund fees and their purpose.
(3) We note, however, that institutional investors are not immune to local bias (e.g., Coval and Moskowitz (2001) find evidence of local bias by sophisticated mutual fund managers). Moreover, there are rational reasons why an institutional mutual fund investor (e.g., a pension fund manager) would favor local funds. For example, investing locally could be efficient if proximity could enhance the monitoring capabilities of the investor and facilitate stronger relationship building between the investor and management team. Although we focus on retail funds in our analysis, our results are qualitatively similar if we include institutional share classes.
(4) Note that while Bailey, Kumar, and Ng (2011) consider the local bias of retail mutual fund investors as an explanatory variable in their regressions, they do not provide any direct evidence regarding the overall magnitude or direction of local bias in the investor population.
(5) We also consider measures based on the number of funds within 100 and 250 miles, and the number of funds and families in the same state. The results are qualitatively and quantitatively similar.
(6) We also find very similar results dropping the six financial centers considered by Christoffersen and Sarkissian (2009): New York, Boston, Philadelphia, Chicago, Los Angeles and San Francisco.
(7) We also run the regression over the full sample period with an interaction term testing whether the competition-fee relationship is different over the second half of the sample. We find that there is no significant difference between the 2000-2007 period and the 2008-2015 period.
(8) The results are qualitatively similar if we define less competitive MSA's as all those with fewer than 50 funds.
(9) We use returns net of expenses in our analysis. In untabulated results, we also estimate alphas using a proxy for returns gross of fees and continue to find no significant differences in performance between the two groups.
(10) We should also note that the alphas calculated using net returns here do not include the effects of loads. Incorporating these loads would diminish the alphas of the funds from more competitive areas.
Table I. Summary Statistics This table presents summary statistics for our sample of actively managed mutual funds from 2000 to 2015. The final sample includes 2,682 funds. Percentile Mean 10 50 Expense Ratio 1.30 0.69 1.30 12b-l fee 0.25 0.00 0.25 Management Fee 0.66 0.13 0.71 Other Expenses 0.39 0.07 0.30 Net flow 0.10 -0.26 -0.05 Total Net Assets ($ million) 1,307.61 15.60 187.40 Family Total Net Assets ($ million) 45,609.01 61.60 1,993.80 Fund Age 13.96 3.00 10.00 Standard Deviation of Monthly Return 0.05 0.03 0.05 Turnover Ratio 1.03 0.12 0.60 Percentile 90 Standard Deviation Expense Ratio 1.92 0.49 12b-l fee 0.58 0.24 Management Fee 1.06 0.40 Other Expenses 0.79 0.36 Net flow 0.60 0.50 Total Net Assets ($ million) 2,528.00 5,797.10 Family Total Net Assets ($ million) 145,460.20 119,935.90 Fund Age 28.00 12.97 Standard Deviation of Monthly Return 0.08 0.02 Turnover Ratio 2.02 1.53 Table II. Local Flow Comovement This table provides evidence on flow comovement among local mutual funds. For each mutual fund in the sample, we estimate regressions of monthly percentage flows on Local Flow, Style Flow, Market Flow, and three lags of fund returns and the returns of other funds in the same MSA (Local Return). Local Flow is the equally weighted average percentage flow of all mutual funds in the same metropolitan statiscal area (MSA) excluding funds in the same mutual fund family. Style Flow is the equally weighted average percentage flow of all mutual funds in the same CRSP objective code excluding funds in the same family. Market Flow is the equally weighted average percentage flow of all mutual funds excluding funds in the same family. Funds are only included if they have at least ten years of flow data and an average of at least ten other funds in the same MSA outside of their family. Our final sample consists of 830 funds. We report the average coefficient from the fund-level regressions and t-statistics are in parentheses. (1) (2) (3) (4) Local Flow 0.09 (***) 0.06 (***) 0.06 (***) 0.05 (***) (7.50) (5.52) (4.80) (3.59) Style Flow 0.12 (***) 0.12 (***) 0.10 (***) (26.21) (22.88) (19.69) Market Flow 0.04 (***) 0.02 (***) (6.05) (3.04) Fund Return(t-1) 0.22 (***) (9.63) Fund Return(t-2) 0.15 (***) (5.98) Fund Retum(t-3) 0.16 (***) (4.73) Local Return(t-l) -0.13 (***) (-5.95) Local Return(t-2) -0.15 (***) (-11.80) Local Retum(t-3) -0.14 (***) (-6.15) (***) Significant at the 0.01 level. (**) Significant at the 0.05 level. (*) Significant at the 0.10 level. Table III. Management Fee Regressions This table presents the results from regressions of fund management fees on a number of control variables, as well as measures of geographic competition. Fund Return is the fund's return over the last 12 months. Flow is the net flow over the last 12 months. Size is the fund's total net assets at the end of the prior year, and Family. Size is the total net assets of the fund's family at the end of the prior year. #Funds in Family is the total number of funds in the fund's family. Age is the number of years since the fund's first offer date. Volatility is the standard deviation of the fund's monthly returns over the prior 12 months, and Turnover is the fund's turnover ratio from CRSP. Index Fund Dummy is equal to one if the fund is categorized as an index fund by CRSR Beta (Mkt, SMB, HML, and UMD) represent the sensitivities to the various factors estimated over the prior three years. Ln(# Funds-MSA) is the log of the number of actively managed mutual funds in the MSA in that year. Ln(#Families-MSA) is the log of the total number of fund families in the MSA. Year and CRSP Objective Code fixed effects are included, and standard errors are clustered by fund family and year, t-statistics are in parentheses. (1) (2) Fund Return(t-1) 0.16 (***) 0.16 (***) (4.02) (4.08) Percentage FIow(t-1) 0.00 0.00 (0.35) (0.29) Total Net Assets(t-1) 0.01 (***) 0.01 (***) (3.00) (3.11) Family Total Net Assets(t-1) -0.02 (***) -0.02 (***) (-2.88) (-2.95) #Funds in Family -0.03 (***) -0.04 (***) (-2.95) (-3.40) Fund Age 0.02 (**) 0.02 (**) (2.15) (2.10) Volatility(t-1) 1.46 (***) 1.48 (***) (4.33) (4.27) Fund Turnover(t-1) 0.03 (***) 0.03 (***) (6.47) (6.39) Index Fund Dummy -0.34 (***) -0.34 (***) (-6.76) (-6.80) Beta_Mkt -0.04 -0.04 (-1.09) (-1.12) Beta_SMB 0.03 0.03 (1.55) (1.60) Beta_HML 0.02 0.02 (1.56) (1.55) Beta_UMD -0.05 (**) -0.06 (**) (-2.14) (-2.26) Ln(# Funds-MSA) -0.03 (***) (-3.59) Ln(#Families-MSA) -0.03 (***) (-3.42) Observations 17,399 17,399 [R.sup.2] 0.32 0.32 Year FE YES YES CRSP_OBJ_CD FE YES YES Clustered S.E. Family&Year Family&Year Adjusted [R.sup.2] 0.32 0.32 (***) Significant at the 0.01 level. (**) Significant at the 0.05 level. (*) Significant at the 0.10 level. Table IV. Management Fee Regressions - Robustness Checks This table presents results from regressions of fund management fees on a number of control variables, as well as measures of geographic competition. Fund Return is the fund's return over the last 12 months. Flow is the net flow over the last 12 months. Size is the fund's total net assets at the end of the prior year, and Family Size is the total net assets of the fund's family at the end of the prior year. Age is the number of years since the fund's first offer date. Volatility is the standard deviation of the fund's monthly returns over the prior 12 months, and Turnover is the fund's turnover ratio from CRSP. Beta (Mkt, SMB, HML, and UMD) represent the sensitivities to the various factors estimated over the prior three years. Ln(#Funds-MSA) is the log of the number of actively managed mutual funds in the MSA in that yean Year and CRSP Objective Code fixed effects are included and standard errors are clustered by fund family and year (Models 1-3) and MSA and year (Model 4). t-statistics are in parentheses. (1) (2) Fund Return(t-1) 0.18 (***) 0.15 (***) (3.41) (4.46) Percentage Flow(t-1) -0.00 0.00 (-0.12) (0.23) Total Net Assets(M) 0.01 (**) 0.01 (**) (2.03) (2.27) Family Total Net Assets(t-1) -0.02 (***) -0.02 (***) (-2.81) (-2.71) #Funds in Family -0.01 -0.03 (***) (-0.89) (-2.70) Fund Age 0.02 0.03** (1.29) (2.38) Volatility(t-1) 1.50 (***) 1.07 (***) (2.93) (3.70) Fund Turnover(t-1) 0.04 (***) 0.03 (***) (4.95) (7.07) Index Fund Dummy -0.35 (***) -0.34 (***) (-6.38) (-6.62) Beta_Mkt -0.02 -0.03 (-0.29) (-0.84) Beta_SMB 0.04 (**) 0.02 (2.08) (1.12) Beta-HML 0.02 0.03 (*) (0.91) (1.92) Beta-UMD -0.06 (*) -0.03 (-1.79) (-1.21) Ln(# Funds-MSA) -0.03 (**) -0.03 (**) (-2.49) (-2.37) Observations 10,695 14,779 [R.sup.2] 0.33 0.33 Year FE YES YES CRSP-OBJ.CD FE YES YES Clustered S.E. Family&Year Family&Year State FE Exclude #Funds>250 #Funds<25 Adjusted [R.sup.2] 0.33 0.32 (3) (4) Fund Return(t-1) 0.14 (***) 0.16 (***) (3.64) (4.46) Percentage Flow(t-1) 0.00 0.00 (0.68) (0.34) Total Net Assets(M) 0.01 (***) 0.01 (***) (3.46) (3.20) Family Total Net Assets(t-1) -0.01 (**) -0.02 (***) (-2.48) (-2.65) #Funds in Family -0.03 (***) -0.03 (***) (-3.11) (-2.62) Fund Age 0.03 (***) 0.02 (**) (2.59) (2.37) Volatility(t-1) 1.53 (***) 1.46 (***) (4.47) (4.67) Fund Turnover(t-1) 0.03 (***) 0.03 (***) (4.91) (6.41) Index Fund Dummy -0.33 (***) -0.34 (***) (-8.30) (-5.39) Beta_Mkt -0.05 -0.04 (-1.35) (-1.08) Beta_SMB 0.03 0.03 (*) (1.60) (1.95) Beta-HML 0.02 0.02 (*) (1.53) (1.94) Beta-UMD -0.03 (*) -0.05 (**) (-1.67) (-2.37) Ln(# Funds-MSA) -0.03 (**) -0.03 (***) (-2.03) (-2.61) Observations 17,399 17,399 [R.sup.2] 0.38 0.32 Year FE YES YES CRSP-OBJ.CD FE YES YES Clustered S.E. Family&Year MSA&Year State FE YES Exclude Adjusted [R.sup.2] 0.37 0.32 (***) Significant at the 0.01 level. (**) Significant at the 0.05 level. (*) Significant at the 0.10 level. Table V. Management Fee Regressions--by Subperiods This table presents regressions results from fund management fees on a number of control variables, as well as measures of geographic competition. Fund Return is the fund's return over the last 12 months. Flow is the net flow over the last 12 months. Size is the fund's total net assets at the end of the prior year, and Family Size is the total net assets of the fund's family at the end of the prior year. #Funds in Family is the total number of funds in the fund's family. Age is the number of years since the fund's first offer date. Volatility is the standard deviation of the fund's monthly returns over the prior 12 months, and Turnover is the fund's turnover ratio from CRSP. Index Fund Dummy is equal to one if the fund is categorized as an index fund by CRSR Beta (Mkt, SMB, HML, and UMD) represent the sensitivities to the various factors estimated over the prior three years. Ln (# Funds-MSA) is the log of the number of actively managed mutual funds in the MSA in that year. Year and CRSP Objective Code fixed effects are included, and standard errors are clustered by fund family and year, t-statistics are in parentheses. (1) (2) 2000-2007 2008-2015 Fund Return(t-1) 0.19 (***) 0.06 (***) (3.59) (3.69) Percentage Flow(t-1) -0.00 0.01 (-0.14) (0.56) Total Net Assets(t-1) 0.02 (***) 0.01 (**) (4.03) (2.04) Family Total Net Assets(t-1) -0.02 (***) -0.01 (**) (-3.12) (-2.35) #Funds in Family -0.03 (***) -0.04 (***) (-2.85) (-3.17) Fund Age 0.01 0.03 (**) (1.35) (2.09) Volatility(t-1) 1.68 (***) 2.58 (***) (4.95) (5.15) Fund Turnover(t-1) 0.03 (***) 0.03 (***) (4.09) (6.68) Index Fund Dummy -0.37 (***) -0.32 (***) (-9.84) (-5.70) Beta_Ikt -0.10 (***) -0.03 (-2.75) (-0.80) Beta_SMB 0.06 (**) 0.00 (2.54) (0.23) Beta_HML 0.04 (**) 0.01 (2.02) (0.24) Beta_UMD -0.03 -0.05 (-0.79) (-1.37) Ln(# Funds-MSA) -0.03 (***) -0.02 (***) (-3.88) (-2.97) Observations 7,409 9,990 [R.sup.2] 0.32 0.34 Year FE YES YES CRSP_OBJ_CD FE YES YES Clustered S.E. Family&Year Family&Year Adjusted [R.sup.2] 0.31 0.33 (***) Significant at the 0.01 level. (**) Significant at the 0.05 level. (*) Significant at the 0.10 level. Table VI. Regressions for Expense Ratio and Other Components This table presents results from regressions of several fund expense measures on a number of control variables, as well as measures of geographic competition. Fund Return is the fund's return over the last 12 months. Flow is the net flow over the last 12 months. Size is the fund's total net assets at the end of the prior year, and Family Size is the total net assets of the fund's family at the end of the prior year. #Funds in Family is the total number of funds in the fund's family. Age is the number of years since the fund's first offer date. Volatility is the standard deviation of the fund's monthly returns over the prior 12 months, and Turnover is the fund's turnover ratio from CRSP. Index Fund Dummy is equal to one if the fund is categorized as an index fund by CRSP. Beta (Mkt, SMB, HML, and UMD) represent the sensitivities to the various factors estimated over the prior three years. Ln(# Funds-MSA) is the log of the number of actively managed mutual funds in the MSA in that year. Year and CRSP Objective Code fixed effects are included, and standard errors are clustered by fund family and year. t-statistics are in parentheses. (1) (2) 12b-1 Fee Other Expenses Fund Return(t-1) -0.06 (**) -0.08 (***) (-2.37) (-5.12) Percentage Flow(t-1) -0.02 (***) -0.03 (***) (-2.87) (-5.29) Total Net Assets(t-1) -0.01 (***) -0.03 (***) (-2.73) (-8.85) Family Total Net Assets(t-1) -0.00 -0.02 (***) (-0.10) (-3.69) #Funds in Family 0.01 0.01 (0.49) (0.82) Fund Age -0.02 (*) -0.03 (***) (-1.85) (-3.08) Standard Deviation of Fund Returns(t-1) 0.12 1.08 (***) (0.55) (3.52) Fund Turnover(t-1) 0.00 0.02 (***) (0.90) (2.85) Index Fund Dummy -0.05 (**) -0.06 (**) (-2.13) (-1.96) Beta_Mkt -0.01 0.03 (*) (-0.75) (1.65) Beta_SMB 0.00 0.03 (*) (0.30) (1.67) Beta_HML -0.02 (**) -0.03 (*) (-2.08) (-1.72) Beta_UMD 0.01 0.00 (0.65) (0.22) Ln(# Funds-MSA) 0.02 (***) 0.02 (***) (3.62) (3.52) Observations 17,399 17,399 [R.sup.2] 0.12 0.24 Year FE YES YES CRSP_OBJXD FE YES YES Clustered S.E. Family&Year Family&Year Adjusted [R.sup.2] 0.12 0.24 (3) Total Expense Ratio Fund Return(t-1) 0.01 (0.41) Percentage Flow(t-1) -0.05 (***) (-3.73) Total Net Assets(t-1) -0.03 (***) (-5.83) Family Total Net Assets(t-1) -0.04 (***) (-4.69) #Funds in Family -0.02 (-1.24) Fund Age -0.03 (-1.33) Standard Deviation of Fund Returns(t-1) 2.66 (***) (5.19) Fund Turnover(t-1) 0.05 (***) (5.47) Index Fund Dummy -0.44 (***) (-5.66) Beta_Mkt -0.02 (-0.60) Beta_SMB 0.07 (***) (3.05) Beta_HML -0.03 (-1.22) Beta_UMD -0.04 (*) (-1.70) Ln(# Funds-MSA) 0.02 (*) (1.68) Observations 17,399 [R.sup.2] 0.46 Year FE YES CRSP_OBJXD FE YES Clustered S.E. Family&Year Adjusted [R.sup.2] 0.46 (***) Significant at the 0.01 level. (**) Significant at the 0.05 level. (*) Significant at the 0.10 level. Table VII. Control Variables by Levels of MSA Competitiveness This table presents average values of the control variables from our fund fee regressions presented in Tables III-VI. Funds are separated into categories based on the number of funds in their MSA: <5, 6-25, 26-50, 51-100, 101-250, and >250. The final column reports the difference between the means for MSA's with >250 funds and MSA's with <5 funds. Number of Funds in MSA < = 5 6-25 26-50 51-100 Fund Return 0.07 0.07 0.07 0.08 Flow 0.13 0.08 0.12 0.16 Total Net Assets 343.52 637.74 972.90 1,540.39 Family Total Net Assets 1,168.45 2,486.02 9,222.70 22,099.74 # Funds in Family 4.00 9.55 16.05 17.15 Fund Age 11.04 12.88 14.38 12.66 Volatility 0.05 0.05 0.06 0.06 Fund Turnover 0.88 0.80 0.84 1.50 Index Fund Flag 0.11 0.14 0.10 0.24 Beta_Mkt 0.96 0.97 1.00 1.03 Beta_SMB 0.21 0.21 0.24 0.22 Beta-HML 0.04 0.03 -0.01 -0.00 Beta.UMD 0.01 0.01 0.02 0.02 Number of Funds in MSA 101-250 >250 >250-<5 Fund Return 0.09 0.07 -0.01 Flow 0.07 0.10 -0.03 (***) Total Net Assets 2,312.91 1,176.43 832.91 (***) Family Total Net Assets 200,898.82 62,688.55 61,520.10 (***) # Funds in Family 69.89 29.53 25.53 (***) Fund Age 17.79 14.00 2.95 (***) Volatility 0.06 0.05 -0.00 Fund Turnover 0.94 0.90 0.01 Index Fund Flag 0.12 0.11 -0.00 Beta_Mkt 1.00 0.99 0.04 (***) Beta_SMB 0.15 0.18 -0.02 (**) Beta-HML 0.01 0.01 -0.02 (**) Beta.UMD -0.01 0.03 0.02 (***) (***) Significant at the 0.01 level. (**) significant at the 0.05 level. Table VIII. Covariate Balancing from Propensity Score Matching This table presents results on the covariate balancing from our propensity score matching analysis. Funds from families with assets greater than $100 billion are dropped from the sample. Funds are considered less competitive if they are from an MSA with 26-50 funds and funds are considered more competitive if they are from an MSA with greater than 100 funds. Panel A presents the average values of the various covariates for the less competitive areas and the more competitive areas, as well as the difference in the two means. The final column reports t-statistics for testing the null hypothesis that the two groups have equal means. Panel B presents the results on the overall covariate balancing. "LR chi2" is the likelihood ratio test statistic for the joint insignificance of all the regressors. Less More Competitive Competitive Difference Panel A. Comparison of Matched Sample Means Fund Return 0.07 0.07 0.00 Percentage Flow 0.12 0.13 -0.01 Total Net Assets 972.90 839.18 133.72 Family Total Net Assets 9222.70 9074.40 148.30 # Funds in Family 16.05 16.14 -0.09 Fund Age 14.38 13.95 0.43 Volatility 0.06 0.06 0.00 Fund Turnover 0.84 0.84 -0.00 Index Fund Flag 0.10 0.12 -0.02 Beta_Mkt 1.00 0.99 0.01 Beta_SMB 0.24 0.24 0.00 BetaJHML -0.01 -0.01 0.00 BetaJJMD 0.02 0.01 0.01 t-statistic for Difference of Means Panel A. Comparison of Matched Sample Means Fund Return 0.14 Percentage Flow -1.05 Total Net Assets 2.51(**) Family Total Net Assets 0.31 # Funds in Family -0.20 Fund Age 1.18 Volatility -0.47 Fund Turnover -0.12 Index Fund Flag -1.23 Beta_Mkt 0.82 Beta_SMB 0.24 BetaJHML 0.55 BetaJJMD 0.70 Panel B. Tests of Overall Covariate Balance LR chi2 11.42 (p-value) (0.58) (**) Significant at the 0.05 level. Table IX. Propensity Score Matching Results - Fees This table presents results on the comparison of mean fees from our propensity score matching analysis. Funds from families with assets greater than $ 100 billion are dropped from the sample. Funds are considered less competitive if they are from an MSA with 26-50 funds and funds are considered more competitive if they are from an MSA with greater than 100 funds. Panel A presents the average values of the various components of annual expense ratios for the less competitive areas and more competitive areas, as well as the difference in the two means. The final column provides t-statistics for testing the null hypothesis that the two groups have equal means. Panel B presents similar results for the average values of front end loads, rear loads, and total loads. Less Competitive More Competitive Difference Panel A. Components of Annual Expense Ratio Management Fee 0.73 0.67 0.06 12b-l Fee 0.17 0.27 -0.10 Other Expenses 0.31 0.38 -0.07 Total Expense Ratio 1.21 1.32 -0.11 Panel B. Loads Front Load 1.13 1.27 -0.14 Rear Load 1.09 1.22 -0.13 Total Load 2.22 2.49 -0.27 t-stat Panel A. Components of Annual Expense Ratio Management Fee 4.95 (***) 12b-l Fee -13.51 (***) Other Expenses -6.17 (***) Total Expense Ratio -7.08 (***) Front Load -3.37 (***) Rear Load -4.26 (***) Total Load -4 63 (***) (***) Significant at the 0.01 level. Table X. Propensity Score Matching - Fund Alphas This table presents the results on the comparison of average monthly alphas (in percent) from our propensity score matching analysis. Funds from families with assets greater than $100 billion are dropped from the sample. Funds are considered less competitive if they are from an MSA's with 26-50 funds and funds are considered more competitive if they are from an MSA with greater than 100 funds. Alphas are calculated using monthly returns over three-year rolling windows using the CAPM, Fama French three-factor model, and the Fama and French (1992) three-factor model + momentum models. The final column reports r-statistics for testing the null hypothesis that the two groups have equal means. Less Competitive More Competitive Difference t-Stat CAPM Alpha 0.05 0.05 -0.00 -0.02 3-Factor Alpha -0.06 -0.05 -0.01 -0.37 4-Factor Alpha -0.07 -0.07 -0.00 -0.13
|Printer friendly Cite/link Email Feedback|
|Author:||Ellis, Jesse A.; Underwood, Shane|
|Date:||Sep 22, 2018|
|Previous Article:||Banks and Corporate Decisions: Evidence from Business Groups.|
|Next Article:||Productivity Risk and Industry Momentum.|