Another explanation of the mutual fund fee puzzle

Abstract

This paper demonstrates that investor sentiment explains the recent puzzle of the negative relation between fees and before-fee performance of equity mutual funds. Using a composite proxy for investor sentiment, the puzzle can be explained stronger by investor sentiment, compared to the strategic fee-setting explanation discussed in the literature. More-sentiment driven investors would like to select more skilled fund managers, leading to a better future performance in short run. Additionally, when sentiment is high (low), it results in lower (higher) fees. Our results highlight the use of investor sentiment approach in determining mutual fund fees and performance.

Introduction

Mutual fund investors pay fees largely for quality portfolio management provided by the fund. Hence, higher fees should reflect better portfolio management and consequently, translate to better risk-adjusted performance. A negative relation between before-fee risk-adjusted performance and fees would thus be contradictory. This contradiction was highlighted in Gil-Bazo and Ruiz-Verdú (2009) (GR hereafter), in the period from January 1962 to December 2005. Two explanations were offered for their anomalous observation which are cost based explanation and strategic explanation (GR, 2009). However, their empirical analysis shows that although strategic fee setting is a significant contributing factor to the fee–performance anomaly, both cost based explanation and strategic explanation do not explain why managers choose to adjust fees instead of undertaking other means to improve expected performance.

We aim to fill this gap by proposing a behavioural explanation for the fee–performance anomaly. Recent advances in the field of behavioural finance have consistently produced empirical evidences that suggest significant effects of investor sentiment in the mutual fund industry (albeit with much controversy). Baker and Wurgler (2006) (BW hereafter) argue that low (high) future stock returns are associated with high (low) investor sentiment. Gruber (1996) and Barber, Odean, and Zheng (2005) argue that mutual funds picked by retail investors are those with high fees. Bailey, Kumar, and Ng (2011) show that the performance of behaviourally biased investors are poor due to an inclination to make poor decisions when it comes to investments in mutual funds.

Generally, past papers have often used mutual fund flows as a sentiment measure for the mutual fund industry. Fund characteristics such as age, participation costs¹ and flow to past performance sensitivity were identified as determinants of mutual fund flow. These prior studies found a convex relation between flow and past performance. For example, Frazzini and Lamont (2008) describe mutual fund flows as “dumb money” whereby fund inflows (outflows) are correlated with low (high) future returns. However, Spiegel and Zhang (2013) show that evidences of a convex relation between flow and past performance from previous studies may be untrue. These authors argue that prior conclusions of a convex relation between flow and past performance are caused by misspecification of the empirical model and that the relation should instead be linear.

Due to data availability constraint and to avoid producing false convexity estimates as described by Spiegel and Zhang (2013), we opt to use the BW composite sentiment index instead of a fund flow measure. Other proxies for investor sentiment such as overconfidence,² representativeness, and conservatism³ have also been considered but as BW (2007) argue, the use of a few selected bias and trading frictions do not sufficiently describe the complexity of real investors and markets. The BW composite sentiment index also serves as an alternative macro-level approach in measuring investor sentiment. To the best of our knowledge, the negative relation between mutual fund fees and before-fee risk-adjusted performance has yet to be tested with the BW composite sentiment index. According to BW (2006), investor sentiment can affect future stock returns. Arguing this point, it could then be reasonable to suggest that sentiment affects fees through stock returns if fund managers respond to changes in after-fee performance by adjusting fees charged to investors.

We hypothesise that when sentiment is high (low), it results in lower (higher) fees and leads to better (poorer) before-fee risk-adjusted performance. To examine the role of investor sentiment on the before-fee performance and fees of mutual funds, we first confirm and extend GR (2009) demonstration of the negative relation between fees and before-fee risk-adjusted performance. Next, we seek explanations to the puzzle by investigating the role of performance in the determination of fund fees. Lastly, we consider whether better fund governance has a positive or negative influence on the effect of sentiment on fees and performance. If better fund governance reduces the likelihood of fund managers adjusting fees in response to investor sentiment, then it would bring fees more in line with performance. Therefore, we also investigate whether stronger fund governance reduces the effects of investor sentiment on the relation between fees and before-fee performance.

The contribution of this paper differs from GR (2009) in that we show the significance and importance of using a behavioural hypothesis approach to explain mutual fund fees and performance. Expanding research in this field is also important. First, the use of mutual funds as an investment tool is increasingly popular among retail investors in past decades. Second, the number of retail investors holding individual stocks has steadily declined. This trend is observable in French (2008) where he reports that during 1980 to 2007, individual holdings of the market fell by 26.4% whereas open-end mutual fund holdings increased by 27.8%.

This paper is organised as follows. Section 2 describes the critical literature review on mutual fund fees and performance. Section 3 describes the hypothesis development. Section 4 describes the sample and data set. Section 5 presents the methodology and empirical results. Conclusions are drawn in Section 6.