The Persistence of Fee Dispersion among Mutual Funds

Almost 15 years ago, Elton, Gruber, and Busse (2004) and Hortacsu and Syverson (2004), documented substantial price dispersion for essentially identical S&P 500 index funds. These results were surprising because in competitive markets, prices for close to identical products should have similar prices. In the case of mutual funds, however, substantial deviations in fees might arise because of (i) the inability to arbitrage away such differences (i.e., one cannot short sell open-ended mutual funds whose fees are too high), (ii) investors that do not pay attention to fees, (iii) search frictions, as the number of mutual funds is large, and (iv) nonfinancial fund differentiation. The conclusion of this earlier literature focusing on index funds is that mutual fund markets are not perfectly competitive and that fees do matter to investors.

At the same time, Berk and Green (2004) proposed a partial-equilibrium model of the mutual fund industry that became very influential and was labelled the neoclassical model of mutual funds. This framework argues that percentage fees are irrelevant, as fund size will adjust in equilibrium such that net alphas (i.e., abnormal fund performance after fees) are equal to zero. The apparent conflict between the model’s predictions and the evidence on index funds has usually been attributed to measurement problems of abnormal performance and to the focus of the empirical evidence on a specific subset of funds, namely passive, index funds.

We revisit this question using a sample of all US and international equity funds from 1980 to 2017. In contrast to the earlier empirical literature, this sample is much more comprehensive and covers a much longer time-series. The key advantage of this rich empirical setup is that it allows us to document novel and interesting results regarding the determinants of fund fees in the cross-section and the dynamics of fees and fee dispersion over time for a broad set of funds. Thus, in contrast to these earlier studies, which use a narrowly defined set of funds and a shorter time period, we are able to evaluate the pervasiveness of fee dispersion in a much broader context.

Using this sample, we first assess whether fees matter to investors or not. We regress net alphas on lagged fees and controls in our sample of all US and international equity funds from 1980 to 2017 and find statistically significant negative coefficients on fees. This implies that, on average, funds with high fees tend to show worse performance. Thus, our empirical analysis shows that from an investor’s standpoint, in the cross-section of net-of-fee returns, fees are important.

As a next step, we examine the competitiveness of the mutual fund markets via an examination of mutual fund pricing. Since the publication of Elton, Gruber, and Busse (2004) and Hortacsu and Syverson (2004) mentioned above, the mutual fund markets have experienced dramatic growth, and along with this growth has come the continued debate, on the part of academics, practitioners, the justice system, and regulators, concerning the competitiveness of the mutual fund markets and related questions concerning mutual fund fees and their impact on investor performance.

To address this question, we perform several empirical tests. First, we extend the work of Elton, Gruber, and Busse (2004) and Hortacsu and Syverson (2004) and find large levels of fee dispersion—in the range of 30 to 60 basis points—across similar S&P 500 index funds in the most recent 15 years of data after their studies. These levels of fee dispersion are only slightly smaller than the levels reported in the original studies implying that price efficiency has not improved much during the last 15 years. In a next step, we extend this analysis to all US and International funds. Given the heterogeneity of this sample of funds, we apply a simple linear fund pricing model that accounts for lagged fund characteristics, such as risk and performance, measures of fund manager skill, the extent of active management, service levels, fund size and age. Given this framework that explains more than 40% of the variation in observed fees, we find levels of fee dispersion in the range of 50 to 100 basis points. Thus, this step of our analysis shows that fees vary substantially across otherwise very similar funds.

In a next step, we analyze the dynamics of fee dispersion. Given the enormous growth in the mutual fund industry, we expected to find a drop in fee dispersion, as competition in the industry presumably has increased during the sample period. This, however, is not what we find. In contrast, we find that the phenomenon of fee dispersion is very persistent across time has basically not changed during the sample period we study.

The above findings of economically large and persistent fee dispersion across arguably similar funds carry important investor implications, especially over the long run. To illustrate this, we estimate the returns to two hypothetical investors. One investor purchases low fee funds and the other purchases similar but higher fee funds. The investor purchasing the lowest expense funds (i.e., the bottom quintile of the entire fee distribution) would have earned compounded net-of-fee abnormal returns between 30% and 58% higher, depending on the risk adjustments, than the investor purchasing the most expensive (i.e., the top quintile) funds over our study period.

As a last step, we extend our analysis to the industry level and provide a quantification of misallocated capital. The notion of misallocated capital is model-based and is designed to capture the notion that some funds have grown too big while other funds are not big enough given their net-of-fee performance. This analysis shows that 70% of the funds in our sample are overinvested and, thus, should shrink. The amount of capital overinvested amounts to 1.4 trillion USD and represents an aggregate, life-time measure for all the funds in the sample. Interestingly, while the fraction of overinvested funds does not vary significantly across fee quintiles, the total amount of capital overinvested decreases substantially from 665 billion USD of funds in the lowest fee quintile to 80 billion USD in the most expensive fee quintile. This later pattern is consistent with the observation that, on average, high-fee funds are much smaller than low-fee funds; it is also consistent with the notion that, in terms of cumulative capital, investors do understand that high fee funds often times do not earn their fees.

The complete paper is available for download here.