The Persistence of Mutual Fund Performance in Europe

The starting point of our analysis is the model of Berk and Green (BG), who characterize the competitive provision of capital to mutual funds. In their model, investors learn about managerial ability from past returns and demand shares of all funds with positive expected risk-adjusted performance net of fees and other costs. If there are diseconomies of scale in portfolio management, the flows of money into (out of) outperforming (underperforming) funds drive their performance down (up) to zero. In equilibrium, all funds deliver zero net expected performance. Therefore, fund performance is not predictable from fund characteristics or past performance.

BG’s influential work has changed the prevalent view on mutual fund performance persistence by showing that lack of predictability in mutual fund performance is consistent with a market populated by competing rational investors, even if fund managers possess skill. However, there exists abundant empirical evidence that underperforming US equity funds continue to underperform in the long term. The model cannot explain, either, why performance persists for winners in the short term. Ferreira show that fund performance persistence is a widespread phenomenom throughout the world. Under the framework of BG, such persistence in mutual fund performance is an anomaly that needs to be explained.

One possible explanation for the discrepancy between the model’s implication of performance unpredictability and the empirical evidence on performance persistence is that the assumption of diseconomies of scale in asset management is not a good characterization of the mutual fund industry. However, the available empirical evidence suggests that US equity fund performance 2 decreases with size. Chen show that, conditional on other fund characteristics, performance decreases with lagged assets under management, especially for funds investing in small-cap growth stocks, suggesting that liquidity is a source of diseconomies of scale portfolio management. Yan confirm these findings using more direct measures of portfolio liquidity.

Our paper is related to the empirical study of Glode, who investigate time variation in performance persistence. Glode find evidence of more persistence after up-markets, which they attribute to a larger presence of unsophisticated investors in the market. Our model provides a precise mechanism through which the entry of less sophisticated investors in the market results in more persistence in performance.

In particular, we find evidence that a number of investment strategies generate alphas in excess of 7%/year (after fund-level trading costs, but before other fund expenses), when measured with a single-factor model or, alternatively, with a four-factor model that controls for fund exposures to size, value, and momentum. These results are generated by an out-of-sample exercise in which investors use the first five years of our sample to obtain initial estimates, then revise their beliefs recursively as new data arrive, using Bayesian updating rules. Moreover, the results are robust to the choice of sample period, and hold in separate out-of-sample portfolio selection experiments conducted on the periods 1993-2000 and 2001-2008.

Our data was obtained from Lipper and comprises monthly returns on European equity mutual funds over the period from June 1988 to February 2008, a total of 237 observations. It includes funds that were alive at the end of the sample as well as dead funds. We include actively managed funds as well as specialist funds with a more passive investment objective. The number of funds over time by investment objective. The number of funds in our sample rose sharply from just over 200 in 1988 to more than 4,400 at the end of the sample, doubling or more than doubling during the first three five-year periods. A similar, if less pronounced, pattern has been observed in the U.S. fund industry.

Following the large literature on mutual fund performance, we control for risk exposure in measuring the funds’ ability to outperform. In particular, we adopt the four factor approach advocated by Carhart. The four factors are a market risk factor, measured here by the MSCI Europe total return index; a size factor (small minus big, or SML) which captures the difference between returns on the Europe STOXX Small Cap Return Index and the Europe STOXX Large Cap Return Index; a value factor (high minus low, or HML) computed as the difference between European value and European growth portfolios. Finally, a momentum factor constructed from the return difference between the top and bottom six sectors from the Dow Jones STOXX 600 Super Sector Indices is included. For comparison, we also report results from a more conventional single-factor approach that only includes the market factor

Recent studies suggest that funds’ ability to generate alpha varies over time, in a way that can be tracked by means of macroeconomic or financial state variables. Moreover, fund exposure to risk factors may also be state- and time-dependent. To capture such effects we consider the following state variables. First, we use the slope of the term structure of interest rates, measured as the difference between the yield of a 10-year Euro area government bond and the 1-month Euribor yield. Second, we consider the dividend yield for a portfolio of European stocks. Third, we use the default spread on European bonds, calculated as the difference between the yields of corporate bonds and yields on government debt. Fourth, we consider the level of the short risk-free rate, measured as the 1-month Euribor. These variables have been widely used in the literature on time-varying investment opportunities and played a key role in the study on U.S. mutual funds by Avramov and Wermers.

Panel A lists performance results for the equal-weighted universe of funds in our sample and the benchmark MSCI Europe index. Over the full sample, 1988-2008, the equal-weighted portfolio of funds returned 10.2% per annum, 85 basis points below the benchmark which returned 11.05% per annum. This average underperformance reflects very different performance of the funds in the first part of the sample, 1988-1998, a period during which they significantly trailed the benchmark, against a period of outperformance relative to the benchmark during the latter sample from 1999 to 2008.

These results are consistent with findings reported for the U.S.. It is generally found that mutual funds on average underperform by between 50 and 100 basis points per annum. These results also indicate that survivorship bias is not overly important in our sample. To further explore this point, we also report quantiles for the alpha distribution. If survivorship bias was a key concern, we would expect the left-tail quantiles to be much smaller than those observed in the right tails (as under-performing funds are excluded from the sample). This is not what we observe. In fact, the cross-sectional distribution of single-factor alphas, which arguably is the most relevant comparison, is largely symmetric.

To fix terminology, a single-period survival rule means that a fund with current-period performance less than some threshold disappears at the end of the period, while a multiperiod survival rule means that a fund disappears if its past n-period performance is less than some threshold. Some of the important theoretical insights about survivor biases pertain to a single-period rule. However, our theoretical work and that of others indicates that the effects of survivor conditioning depend critically on the nature of the survival rule. Evidence that lagged performance predicts survival, even in the presence of the most recent year’s performance, suggests a multiperiod survival rule for U.S. mutual funds.

We examine empirically the impact of survivor conditioning on persistence tests and find that the conditioning attenuates performance persistence relative to the full sample. This empirical evidence supports the theoretical predictions in Brown, Grinblatt and Titman, and Carpenter and Lynch for mutual funds. Myers finds that survivor conditioning empirically reduces performance persistence for pension funds as well.

Several recent articles have constructed mutual fund databases that attempt to control for survivor biases. Elton, Gruber, and Blake (1996) follow the cohort of funds listed in Wiesenberger’s 1977 volume from 1976 until 1993, constructing complete return histories up to the date of merger for funds with assets of more than $15 million. Brown and Goetzmann use annual returns from 1977 to 1988 estimated from Wiesenberger’s Investment Companies, while Malkiel uses quarterly returns from 1971 to 1991, obtained from Lipper Analytical Services. Myers (1999) and Coggin and Trzcinka examine survivor biases associated with U.S. pension funds. Our dataset includes all known diversified equity mutual funds monthly from January 1962 to December 1995. However, even this database may impose a small degree of survivor conditioning, since Elton, Gruber, and Blake find that it contains some errors and missing returns.

The persistence of mutual fund performance has become one of the main topic for performance measurement of a fund portfolio. The reason it is so widely researched is that the answer will assist not only investors with a more informed decision but also help academic researchers confirm the efficient market hypothesis. For instance, the semi-strong form of the hypothesis implies that abnormal returns can only be obtained from insider information or luck, not from technical or fundamental analysis.

For such importance, many studies have been conducted to measure performance persistence using different performance measures. Grinblatt and Titman from testing alpha persistence and multiple benchmarks concluded that performance persists; Hendricks, Patel, Zeckhauser showed short-term persistence and “cold hands” effect; Elton, Gruber, and Blake, Carhart , and Droms and Walker concluded persistence to a certain degree using Alpha persistence and investing in winners analysis. The studies from Brown and Goetzmann, Malkiel, and Droms and Walker (2001a) also indicated persistence while showing that persistence may be sensitive to the tested period, i.e. the observed persistence was much stronger in 1970s than during 1980s. Most research was conducted on U.S. funds using data from the years of 1960s to early years of 1990s.

In this paper examine the persistence of mutual fund performance using some European large cap mutual funds from 2014 to 2018. From the stock price we calculate Jensen’s alpha and use this measure in multiple tests. The results show significant persistence in short term period and insignificant persistence in longer term. This pattern is consistent with the consensus of aforementioned studies, which consistently show the short-term persistence through alpha test, while seeing little evidence of persistence in the long term.

To test the performance of a mutual fund in the very short run, in this test - daily performance, and see how that measure persists through a month’s period, we need a risk adjusted measure. We choose Jensen’s alpha for this task. To see how a winning or losing fund perform, we choose a well performing fund and an underperforming fund from a temporary period - called ranking period. 

First, a ranking period of 1 month (consisting of 20 trading days) is selected, then we calculate the performance of the funds based on Jensen’s alpha in the sample and sort out the top ranking (best performing fund base on Jensen’s alpha for the ranking period) and the bottom ranking fund (worst performing fund based on jensen’s alpha for the ranking period). Then to see how well the performance persist on a daily basis with risk adjusted, daily alpha is needed for the next period - post-ranking period. We calculate daily alpha using the formulamRanking period (1/1/2018 - 31/1/2018): we run the regression for one month (January 2018) and sort out one firm with highest alpha and one with lowest alpha for this month. We assume that those are the current best and worst performing fund of the period being observed (January 2018)

Beta estimation window (1/1/2014 - 31/12/2017): we calculate the beta coefficient for each fund over a 4 years period (January 2014 - December 2017) and assume that it is the fixed “true beta” for the next short-term period. With the already available input of market premium and risk premium (calculated from market index, risk free rate, and observed returns), the “true” beta input will help calculate the projected alpha that is volatility adjusted. With this assumption, the daily Jensen’s alphas projected in the post-ranking period will be highly comparable. Thus we can make day to day observation in the post-ranking period, compute cumulative abnormal returns and observe whether where is any lingering effect of good or bad performance.

Post-ranking period (1/2/2018 - 28/2/2018): the alpha coefficient calculated in this phase is already beta-adjusted and thus serves as a measure of abnormal return. For economic meaning, the alpha indicates how well a fund/ an investment has performed after accounting for the risk involved. We use the estimated beta, together with the daily risk premia, to get daily alphas from February 1st 2018 to February 28th 2018. By calculating cumulative abnormal return difference at the end of the period, one can see the persistence in performance of a top rank and bottom rank fund.

 If cumulative abnormal returns of the top ranking fund is significantly larger than that of the bottom ranking fund then there is persistence in their performance. Bollen and Busse did a similar ranking and comparing process using abnormal returns from market timing and net selectivity; albeit using a larger sample of funds and running a longer post-ranking period.

Besides the test for alpha persistence in short term, we also want to test how daily returns of an average fund in our sample depend on its own historical data. To do this we coefficient for lagged return to the simple regression with risk premium, market premium, Jensen’s alpha, and beta. The coefficient for lagged return shows how the excess return of a fund on a given day is correlated with its excess return the day before. 

If this lagged return coefficient is positive and significant, then there is some persistence in performance relative to the previous day, i.e. if the previous day return is positive then on average the next day return is positive and vice versa; If this lagged return coefficient is negative and significant then we conclude there is high volatility in return on a daily basis, i.e. if the previous day is positive then the return of the next day would likely be negative and vice versa; If the coefficient is insignificant then one can conclude there is no relation between two consecutive days returns.

In this regression we attempt to find persistence of performance in very short periods. Although daily excess return is not a risk adjusted measure, this test is simple and utilize the readily available stock price and premia that we have at hand. Therefore, we think it is worth looking into. In this test we split the 5-year data set into 2 parts: the first half part consisting of data from January 2014 to June 2016 (2,5 years), the second half consisting of data from July 2016 to December 2018 (2,5 years). In each sub datasets we run regression for each fund to find the Jensen’s alpha. This Jensen’s alpha measure the general performance of that fund, accounted for market risk, for the whole sub period.