Active Fundamental Performance

Active Fundamental Performance Abstract We propose a new measure, active fundamental performance (AFP), to identify skilled mutual fund managers. AFP evaluates fund investment skills conditioned on the release of firms’ fundamental information. For each fund, we examine the covariance between deviations of its portfolio weights from a benchmark portfolio and the underlying stock performance on days when firms publicize fundamental information. Because asset prices on these information days better reflect firm fundamentals, AFP can more effectively identify investment skills. From 1984 to 2014, funds in the top decile of high AFP subsequently outperformed those in the bottom decile by 2% to 3% per year. Received August 25, 2016; editorial decision December 17, 2017 by Editor Andrew Karolyi. Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online. Despite a large body of literature examining performance evaluation, identifying skilled mutual fund managers remains challenging. Noise and random shocks to asset returns make it difficult to differentiate skills from luck. Because observed mutual fund alphas typically are small and noisy, an evaluator would need an unfeasibly long return series to identify a skilled manager reliably.1 Recent studies show that asset prices reflect fundamental information to a greater extent on news release days: the risk-return relationship tends to manifest more clearly, and stock-level mispricing tends to be corrected with the release of important fundamental information (e.g., Savor and Wilson 2013, 2014, 2016; Lucca and Moench 2015; Engelberg, McLean, and Pontiff 2015). Such findings have important implications for performance evaluation. When stock prices are more informative of their fundamental values, fund performance is more indicative of the manager’s investment skill. However, no existing performance measures have exploited this window of opportunity. With the current study, we intend to fill this gap. We propose a measure of fund investment skills conditioned on the release of firm fundamental information. We focus on price movements driven by fundamental information, and our measure uses available fund performance and stock return data more efficiently to increase the power of performance evaluation. Specifically, the proposed performance measure, which we refer to as “active fundamental performance” (AFP), captures the covariance between a fund’s portfolio weight deviation from a passive benchmark (or from previous holdings) and the stocks’ performance during a 3-day window surrounding subsequent earnings announcements. We choose this short window as a proxy for the release of fundamental information because earnings announcements, as major company information events, are associated with a substantial “correction” in stock prices. For instance, Sloan (1996) documents that approximately 40% of the profits from accrual strategies cluster in a 3-day earnings announcement window. La Porta et al. (1997) report that between 25% and 30% of the returns to various value strategies considered by Lakonishok, Shleifer, and Vishny (1994) are concentrated on the 3 days around earnings announcements. Jegadeesh and Titman (1993) estimate that approximately 25% of momentum profits are concentrated on the 3 days surrounding earnings announcements. More recently, Engelberg, McLean, and Pontiff (2015) investigate 97 stock return anomalies and find that anomaly returns are 7 times higher on earnings announcement days. This evidence suggests that the short window around earnings announcements is a period in which stock prices converge to fundamental values. Therefore, we posit that a fund’s underlying stock performance around earnings announcements can be particularly revealing in identifying active managers skilled in selecting mispriced securities and successfully predicting the correction in their prices. We analyze quarterly holdings data for 2,538 unique, actively managed U.S. equity funds over the period 1984–2014. For each fund in each quarter, we compute the index-based AFP (the covariance between a fund’s active weights, or portfolio weight deviations from its benchmark index, and the underlying stocks’ performance during subsequent earnings announcements) and the trade-based AFP (the covariance between changes in a fund’s portfolio weights and the underlying stocks’ performance during subsequent earnings announcements). We find that the index-based (trade-based) AFP measure is positive on average, with a cross-fund mean of 9.95 (5.40) basis points (bps) per 3-day window and a standard deviation of 32.46 (22.09) bps. These results suggest substantial cross-sectional heterogeneity in mutual funds’ performance; at least some managers exhibit skill. In addition, the measure shows a positive, albeit only moderate, correlation with other commonly used performance measures. For example, the index-based (trade-based) AFP has average cross-sectional correlations of 25%, 24%, 32%, and 17% (6%, 6%, 8%, and 23%) with raw fund returns, the four-factor alpha, Daniel et al.’s (1997) characteristic selectivity (CS) measure, and the Grinblatt and Titman (1993) measure, respectively. Thus, compared with other performance measures, AFP captures unique fund characteristics and substantial incremental information. We find that AFP strongly predicts subsequent fund performance. In univariate sorts by index-based (trade-based) AFP, mutual funds in the top decile with the highest AFP outperform those in the bottom decile with the lowest AFP by 2.76% (1.44%) per year. This outperformance cannot be accounted for by their different exposures to risk or style factors. For example, after adjusting for their differential loadings on the market, size, value, and momentum factors, mutual funds in the top decile with the highest index-based (trade-based) AFP continue to outperform those in the bottom decile by 2.40% (1.44%) per year. In other tests, we control for the effects of liquidity and post-earnings announcement drifts, and AFP remains a strong predictor of future fund performance. Note that our fund portfolio strategy is based on information about fund holdings lagged by at least 2 months. The U.S. Securities and Exchange Commission (SEC) requires mutual funds to disclose their portfolio composition within 2 months, making this strategy implementable for mutual fund investors or funds of mutual funds aiming to improve their selection performance. In double sorts, we find that the strong predictive power of AFP for future fund returns is incremental beyond that of various return- and holding-based performance measures, such as CS (Daniel et al. 1997), return gap (Kacperczyk, Sialm and Zheng 2008), reliance on public information (RPI; Kacperczyk and Seru 2007), and active share (Cremers and Petajisto 2009). We also find that mutual fund investors can further improve their returns if they combine the information contained in AFP with signals from return gap. For example, mutual funds in the top quartiles of the index-based AFP and return gap outperform those in the bottom quartiles by 3.36% per year based on the four-factor model. We also consider multiple regressions that jointly control for alternative measures of fund investment skill such as CS, return gap, RPI, and active share, together with other fund characteristics, including fund age, size, expense ratios, turnover, past flow, and past performance. We find that AFP remains statistically significant in predicting future fund performance in these regressions. In both the portfolio and the regression analyses, AFP significantly predicts future fund performance more consistently than the alternative performance measures. We conduct further analyses to shed light on the information sources for AFP. Using stock-level analyses, we find that both active weights and changes in portfolio weights can predict abnormal return around future earnings announcements and exhibit stronger predictability for companies with more information asymmetry (i.e., those with lower analyst coverage, higher idiosyncratic volatilities, and higher analyst disagreement). Moreover, high-AFP funds’ superior performance is more pronounced when financial analysts’ forecasts diverge. These results support the notion that the ability to form a superior estimate of firms’ fundamental values drives AFP and its performance forecasting power. The AFP measures also exhibit time-series persistence; for instance, mutual funds in the top decile with the highest index-based AFP continue to exhibit significantly higher AFP than those in the bottom decile in the subsequent 2 years. This persistence is largely due to the superior AFP of skilled funds in the top decile. The current study contributes to the broad literature on mutual fund performance and market efficiency by introducing a new performance measure as well as a new and more efficient way of using available data.2 We show that focusing on important data points (information days) and integrating information on the extent and quality of active management into a single measure yield substantially more power in identifying skilled managers. The AFP measure is related to Grinblatt and Titman’s (1993) benchmark-free performance measure and Daniel et al.’s (1997) characteristic-based performance measure, which are based on fund holdings and subsequent stock returns. Our innovation of conditioning on a fundamental information release goes further and exploits the high information-to-noise ratio during the time period in which prices tend to converge to their fundamental values. The conditional approach sharpens signals from subsequent stock performance and substantially improves the power to evaluate stock-picking skills. In unreported tests, we find that when we replace earnings announcement returns with stock returns in the entire subsequent quarter, AFP’s forecasting power for future fund performance disappears. Our study has implications for the growing literature that quantifies the level of activeness in a managed portfolio (e.g., Cremers and Petajisto 2009; Cremers et al. 2013; Petajisto 2013; Stambaugh 2014). We contribute to this literature stream by nesting the degree of activeness into a tight framework for performance evaluation. Our approach focuses on signed active portfolio weights (or changes in weights) and examines the covariance between the weights and the underlying stock returns on information days. Empirically, we show that the AFP measure helps separate skilled versus unskilled active funds with high active shares, which makes it particularly useful for mutual fund investors. Ali et al. (2004) and Baker et al. (2010) document evidence that trades by aggregate institutional investors and aggregate mutual funds forecast subsequent earnings surprises, respectively. Cohen, Frazzini, and Malloy (2008) track the shared educational network between mutual fund and corporate managers and find that connected stock holdings achieve high average returns, particularly during earnings announcements. In combination with studies documenting earnings announcements as the short period when stock prices quickly converge to the fundamental values, these papers provide a useful foundation for us to create a powerful fund-level performance measure that identifies skilled fund managers. Although the current study is closely related to Baker et al. (2010), it differs in several key aspects. First, whereas Baker et al. (2010) focus on whether mutual fund managers on average can pick stocks, we investigate how to identify individual managers who can select stocks. Second, in terms of empirical tests, Baker et al. (2010) focus on stock-level analyses, assessing earnings announcement returns of all fund trades or trades of a certain group of funds. In contrast, the current study focuses on fund-level analyses by proposing a new fund performance measure. Third, the emphasis in Baker et al. (2010) is on the ability of fund managers to forecast a firm’s future earnings number. An important contribution of our study is to introduce the notion of performance evaluation, conditional on the release of fundamental information. We use earnings announcements as a proxy and starting point, but our measure has broad implications and can include other fundamental information events. Another literature stream examines how investors use publicly available information and explores the potential performance implications. For instance, Kacperczyk and Seru (2007) examine mutual fund managers’ reliance on public information as reflected in analyst stock recommendations, and Fang, Peress, and Zheng (2014) investigate their reliance on mass media. Both studies show that relying on public information has negative consequences for mutual fund managers’ performance. In a different context, Engelberg, Reed, and Ringgenberg (2012) investigate the source of short sellers’ skills and show that short sellers have a superior ability to interpret and process publicly available information, which translates into investment skills. In contrast with these studies focusing on how fund managers make use of public information, which is backward looking, we evaluate fund performance conditional on the release of fundamental information with a forward-looking focus. The investment ability of fund managers includes but is not limited to their ability to process public information. The key to our identification is that the realization of superior fundamental performance tends to be catalyzed when important firm-level information such as earnings news is released to the public domain. 1. Methodology: AFP In this section, we develop our new measure of performance evaluation, AFP. The starting point is the covariance between portfolio weights and future asset returns, Cov($$w_{i,t}$$,$$ R_{i,t})$$, as shown in Equation (1). As Grinblatt and Titman (1993) and Lo (2008) point out, aggregating the covariance over all investments is an intuitive way to examine managers’ skill to forecast asset returns. The portfolio weights reflect managers’ strategic decisions to buy, sell, or avoid an asset in anticipation of its future returns. A skilled manager whose percentage holdings of assets increase in future asset returns will on average exhibit a positive covariance. The covariance measure is theoretically appealing, in that it captures fund performance due to active portfolio management (Grinblatt and Titman 1993; Lo 2008): \begin{align} &\underbrace {\sum\limits_{i=1}^N {Cov(w_{i,t} ,R_{i,t} )} }_{Active}=\sum\limits_{i=1}^N {E(w_{i,t} \times R_{i,t} )} -\sum\limits_{i=1}^N {E(w_{i,t} )\times E(R_{i,t} )}\notag \\ &\quad=\underbrace {E(R_{p,t} )}_{Total}-\underbrace {\sum\limits_{i=1}^N {E(w_{i,t} )\times E(R_{i,t} )} }_{Passive}, \end{align} (1) where $$w_{i,t}$$ is the weight of security $$i$$ in the fund’s portfolio at the beginning of period $$t$$, $$R_{i,t}$$ is security $$i$$’s return during period $$t$$, and $$N$$ is the number of securities in the fund’s portfolio. Equation (1) shows that the sum of the covariances across securities captures the difference between total portfolio returns $$R_{p,t}$$ and the passive return to the portfolio. We propose two innovations to the covariance measure. Motivated by recent studies on information releases and asset prices (see, e.g., Savor and Wilson 2013, 2014, 2016; Lucca and Moench 2015; Engelberg, McLean, and Pontiff 2015), we measure future asset returns within a short window during which price movements are driven by firm fundamental information. This approach helps mitigate the influence of noise in asset returns and increases the power of the covariance measure to identify investment skills. Specifically, we examine stock returns during 3-day windows around earnings announcements to focus on time periods during which stock prices tend to reflect this fundamental information. The advantage of focusing on earnings announcement returns lies in the high information-to-noise ratio on earnings announcement days and the large cross-section of earnings announcement events, both of which increase the power of performance evaluation. A disadvantage arises because firms release earnings numbers relatively infrequently, typically once per quarter. Although we choose a 3-day event window—a standard practice in event studies of earnings announcements—our results also are robust to using a 5-day event window. Our second innovation is to use benchmark-adjusted portfolio weights, instead of raw portfolio weights, to capture a manager’s active investment decisions. Cremers and Petajisto (2009) pioneered the idea of using benchmark-adjusted portfolio weights to measure a fund manager’s active investment decisions. We provide a theoretical illustration of this approach in Internet Appendix A. Because an actively managed mutual fund’s performance is typically benchmarked against a passive index, the passive index is a natural choice for the fund’s benchmark portfolio. Alternatively, we could use the fund’s own past holdings as a benchmark, thereby using changes in portfolio weights (i.e., fund trades) to compute AFP. Active weights, or deviations in portfolio weights from the benchmark, contain the fund manager’s information revealed through both recent and longer-term holdings. Changes in portfolio weights reflect recent portfolio adjustments and thus recent information about security returns. Ex ante, both active weights and changes in portfolio weights can be useful for assessing the manager’s skill, though the former may be more comprehensive in capturing the manager’s information set. Herein, we compute both index-based and trade-based AFP, and we report both sets of results. As Equation (2) indicates, our covariance measure reflects the expected abnormal fund returns during earnings announcements, attributable to the fund’s active portfolio decisions: \begin{align} \textit{Cov}(w_{i,t} - w_{i,t}^b, \textit{CAR}_{i,t}) &= E[(w_{i,t} - w_t^b) \times \textit{CAR}_{i,t} ] - E(w_{i,t} - w_{i,t}^b) \times E(\textit{CAR}_{i,t})\notag\\ &= E[(w_{i,t} - w_{i,t}^b) \times \textit{CAR}_{i,t} ] - 0 \times E(\textit{CAR}_{i,t})\notag\\ &= E[(w_{i,t} - w_{i,t}^b) \times \textit{CAR}_{i,t} ]. \end{align} (2) We develop an empirical analog of the sum of this covariance measure across securities, termed AFP, in Equation (3): \begin{align} \textit{AFP}_{j,t} = \sum\limits_{i=1}^{N_j} (w_{i,t}^j - w_{i,t}^{b_j})\textit{CAR}_{i,t}, \end{align} (3) where $$\textit{AFP}_{j,t}$$ is mutual fund $$j$$’s AFP based on its portfolio selection in quarter $$t$$, $${\boldsymbol{w}_{\boldsymbol{i,t}}^{\boldsymbol{j}}}$$ is the weight of stock $$i$$ in fund $$j$$’s portfolio at the start of quarter $$t$$, $${\boldsymbol{w}_{\boldsymbol{i,t}}^{\boldsymbol{b}_{\boldsymbol{j}}}}$$ is the weight of stock $$i$$ in fund $$j$$’s benchmark portfolio at the start of quarter $$t$$, $$\textit{CAR}_{i,t}$$ is stock $$i$$’s 3-day cumulative abnormal return surrounding its quarterly earnings announcement during quarter $$t$$, and $$N_{j}$$ is the number of stocks in fund $$j$$’s investment universe (i.e., the union of stocks held by the fund and those in the fund’s benchmark portfolio). The daily abnormal returns refer to the difference in daily returns between a stock and its size and book-to-market matched portfolio. We sum the daily abnormal returns from 1 day before to 1 day after earnings announcements to obtain the 3-day abnormal return. 2. Computing AFP 2.1 Sample construction We obtain the portfolio holdings for actively managed equity mutual funds from Thomson Reuters’s CDA/Spectrum Mutual Fund Holdings Database and returns for the individual mutual funds and other fund characteristics from the Center for Research in Security Prices (CRSP) Survivor-Bias-Free U.S. Mutual Fund Database. We then use the MFLINKS data set to merge the two databases, excluding balanced, bond, money market, international, index, and sector funds, as well as funds not invested primarily in equity securities. After applying this filter, the sample consists of 2,538 unique funds, ranging in time from the first quarter of 1984 to the second quarter of 2014. To select the benchmark index for fund managers, we follow Cremers and Petajisto’s (2009) procedure. The universe of benchmark indexes includes 19 widely used benchmark indexes: S&P 500, S&P 400, S&P 600, S&P 500/Barra Value, S&P 500/Barra Growth, Russell 1000, Russell 2000, Russell 3000, Russell Midcap, value and growth variants of the four Russell indexes, Wilshire 5000, and Wilshire 4500. For each fund in each quarter, we select an index that minimizes the average distance between the fund portfolio weights and the benchmark index weights. Data on the index holdings of the 12 Russell indexes since their inception come from the Frank Russell Company, and data on the S&P 500, S&P 400, and S&P 600 index holdings since December 1994 are provided by COMPUSTAT. For the remaining indexes and periods, we use the index funds holdings to approximate index holdings. The information on daily stock prices and returns for common stocks traded on the NYSE, AMEX, and NASDAQ comes from the CRSP daily stock files. We obtain firms’ announcement dates for quarterly earnings from COMPUSTAT and analysts’ consensus earnings forecasts from I/B/E/S. Panel A of Table 1 shows the summary statistics for mutual funds in our sample. An average fund in our sample manages $\$$ 1.39 billion of assets and is 15.70 years old. Their mutual fund investors achieve an average return of 2.39% per quarter. The net percentage fund flow is skewed to the right; the quarterly fund flow has a mean of 1.35% but a median of only $$-$$1.05%. On average, mutual funds in our sample incur an annual expense ratio of 1.23% and turn over their portfolios by 86.08% per year. These numbers are in line with the previous literature. Table 1 Descriptive statistics A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 This table presents descriptive statistics for our sample of mutual funds. The sample consists of 2,538 distinct mutual funds from 1984Q1 to 2014Q2. Panel A presents the summary statistics for fund characteristics. TNA is the quarter-end total net fund assets in millions of dollars; age is the fund age in years; quarterly return is the quarterly net fund return as a percentage; flow is the quarterly growth rate of assets under management as a percentage after adjusting for the appreciation of the fund’s assets; expense is the fund expense ratio as a percentage, and turnover is the turnover ratio of the fund as a percentage. Panel B shows the time-series average of the cross-sectional Spearman correlation coefficients for the variables of interest. Table 1 Descriptive statistics A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 This table presents descriptive statistics for our sample of mutual funds. The sample consists of 2,538 distinct mutual funds from 1984Q1 to 2014Q2. Panel A presents the summary statistics for fund characteristics. TNA is the quarter-end total net fund assets in millions of dollars; age is the fund age in years; quarterly return is the quarterly net fund return as a percentage; flow is the quarterly growth rate of assets under management as a percentage after adjusting for the appreciation of the fund’s assets; expense is the fund expense ratio as a percentage, and turnover is the turnover ratio of the fund as a percentage. Panel B shows the time-series average of the cross-sectional Spearman correlation coefficients for the variables of interest. Panel B of Table 1 shows the average Spearman cross-sectional correlation coefficients among fund characteristics. The results confirm our intuition: the average 29% correlation coefficient between fund size and age indicates that large funds tend to have a longer track record, and the correlation between fund size and expense ratio of $$-$$22% indicates that large funds tend to incur lower expense ratios. We also find a negative correlation of $$-$$13% between fund age and fund flow, consistent with the idea that established mutual funds with longer life spans tend to be stable, with a smaller percentage of inflows. In the next subsection, we analyze AFP. 2.2 Computing AFP For each fund in each quarter, we compute both index-based and trade-based AFP. For the trade-based AFP, we measure changes in portfolio weights and account for mechanical changes in portfolio weights due to stock price changes.3 Most earnings announcements (more than 89% in our sample) occur in the first 2 months after the quarter ends, so we use the time line in Figure 1: the active weights (changes in portfolio holdings) for fund $$j$$ are measured at the end of month $$t$$ (e.g., end of March), and the earnings announcements are observed in the first 2 months in the following quarter (e.g., April and May). We use Equation (3) to compute the AFP for fund $$j$$, or AFP$$_{j,t}$$. In the analysis of fund performance in the next section, we track the performance of fund $$j$$ for the subsequent 3 months, including months $$t+$$3, $$t+$$4, and $$t+$$5 (e.g., June–August). Figure 1 View largeDownload slide Time line for AFP This figure illustrates the measurement of AFP. In this illustration, AFP is measured at the end of May. The portfolio compositions of mutual funds and their benchmarks (lagged portfolio weights adjusted for price changes for trade-based AFP) are measured in March and observable by the end of May. The earnings announcements arrive in April and in May. Mutual fund portfolios are formed using AFP (available in real time) at the beginning of June and held until the end of August. We then recompute AFP using new information; doing so initiates portfolio rebalancing. Figure 1 View largeDownload slide Time line for AFP This figure illustrates the measurement of AFP. In this illustration, AFP is measured at the end of May. The portfolio compositions of mutual funds and their benchmarks (lagged portfolio weights adjusted for price changes for trade-based AFP) are measured in March and observable by the end of May. The earnings announcements arrive in April and in May. Mutual fund portfolios are formed using AFP (available in real time) at the beginning of June and held until the end of August. We then recompute AFP using new information; doing so initiates portfolio rebalancing. We choose this time line to balance two key considerations. On the one hand, the closer earnings announcement dates are to the reported holdings dates, the more accurate are the AFP measures due to unobserved trades by mutual funds within the quarter. On the other hand, a longer time span between the announcement dates and the reported holdings dates allows us to include more firms in the analysis. Using the first 2 months of a quarter balances these two considerations: most firms report their quarterly performance in the first 2 months of a quarter, and the time span between the reported portfolio composition and earnings announcement performance is reasonably close. Moreover, from a practical perspective, because mutual funds are required to report their holdings within 2 months, the information required to compute AFP is available to investors in real time. Thus, the trading strategy is implementable. To provide further justification for our time line, in Figure 2 we plot the average AFP value for a median fund over the 13 weeks following a typical quarter end. It indicates that for an average fund, the value of AFP stabilizes during the eighth or ninth week after the quarter ends, when we observe the fund’s portfolio composition. Thus, we conclude that incorporating earnings events that occur after the first 2 months do not improve the information content of a fund’s AFP. Figure 2 View largeDownload slide Cumulative AFP over the weeks following quarter ends This figure plots the AFP for a median mutual fund in our sample, cumulative over the weeks following quarter ends, when the active fund weights or changes in portfolio weights are measured. Like in Equation (3), AFP is the sum of the product of the fund’s active portfolio weights (panel A) or changes in portfolio weights (panel B) and subsequent 3-day abnormal returns surrounding earnings announcements. The value of cumulative AFP at the end of week 13 is scaled to equal 1. Figure 2 View largeDownload slide Cumulative AFP over the weeks following quarter ends This figure plots the AFP for a median mutual fund in our sample, cumulative over the weeks following quarter ends, when the active fund weights or changes in portfolio weights are measured. Like in Equation (3), AFP is the sum of the product of the fund’s active portfolio weights (panel A) or changes in portfolio weights (panel B) and subsequent 3-day abnormal returns surrounding earnings announcements. The value of cumulative AFP at the end of week 13 is scaled to equal 1. For an average fund in a typical quarter, the index-based (trade-based) AFP is equal to 10.28 (5.30) bps, with a standard deviation of 84.72 (97.10) bps.4 A substantial proportion of the high variability of AFP comes from cross-fund dispersion. For each fund, we compute the average AFP over its entire life. The cross-fund standard deviation for index-based (trade-based) AFP is 32.46 (22.09) bps, which is 3.26 (4.09) times the cross-section mean of 9.95 (5.40) bps. This high cross-fund dispersion in AFP is the main focus of the current research. 3. Predicting Mutual Fund Performance by AFP 3.1 Portfolio sorts Using a portfolio-based analysis, we examine the profitability of a strategy that invests in mutual funds according to their AFP. Specifically, at the end of each May, August, November, and February, we sort mutual funds into ten portfolios by their AFP and rebalance them quarterly. We compute equally weighted returns for each decile portfolio over the following quarter, both after and before expenses. In addition, we estimate the risk-adjusted returns on the portfolios as intercepts from time-series regressions, according to the Capital Asset Pricing Model (CAPM); Fama and French’s (1993) three-factor model (market, size, and value); Carhart’s (1997) four-factor model, which augments the Fama-French factors with Jegadeesh and Titman’s (1993) momentum factor; and the five-factor model, which adds Pastor and Stambaugh’s (2003) liquidity risk factor to the four-factor model. For instance, the Carhart four-factor alpha is the intercept from the following time-series regression: \begin{align} R_{p,t} -R_{f,t}&= \alpha_p + \beta_m (R_{m,t}-R_{f,t}) + \beta_{\textit{smb}}\textit{SMB}_t +\beta_{\textit{hml}}\textit{HML}_t +\beta_{\textit{umd}}\textit{UMD}_t +\varepsilon_{p,t},\notag\\ \end{align} (4) where $$R_{p,t}$$ is the return in month $$t$$ for fund portfolio $$p$$, $$R_{f,t}$$ is the 1-month Treasury-bill rate in month $$t$$, $$R_{m,t}$$ is the value-weighted stock market return in month $$t$$, SMB$$_{t}$$ is the difference in returns between small and large capitalization stocks in month $$t$$, HML$$_{t}$$ is the return difference between high and low book-to-market stocks in month $$t$$, and UMD$$_{t}$$ is the return difference between stocks with high and low past returns in month $$t$$. Table 2 presents the portfolio results, which indicate that AFP predicts future fund performance. Panel A shows the net returns for portfolios of funds sorted by their AFP. In the quarter following portfolio formation, mutual funds with high index-based (trade-based) AFP in decile 10 outperform the funds with the lowest AFP in decile 1 by 23 (12) bps per month, or 2.76% (1.44%) per year. This superior performance cannot be attributed to their high propensity to take risks or different investment styles; the differences in the alphas obtained from the CAPM, Fama-French three-factor, Carhart four-factor, and Pastor-Stambaugh five-factor models are 21, 27, 20, and 22 (12, 15, 12, and 12) bps per month for index-based (trade-based) AFP, respectively, and all of these differences are statistically significant. Table 2 AFP and mutual fund returns: Decile portfolios A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net returns Average return 0.87 0.90 0.93 0.91 0.93 0.92 0.96 1.02 1.00 1.10 0.23*** (3.42) (3.74) (3.95) (3.89) (3.96) (3.90) (4.01) (4.18) (4.06) (4.16) (3.14) CAPM $$\alpha $$ –0.18 –0.12 –0.08 –0.10 –0.08 –0.09 –0.06 –0.01 –0.03 0.03 0.21*** (–2.79) (–2.49) (–1.79) (–2.26) (–2.1) (–2.3) (–1.52) (–0.3) (–0.58) (0.37) (2.97) Fama-French $$\alpha$$ –0.19 –0.14 –0.09 –0.11 –0.09 –0.11 –0.07 –0.01 –0.02 0.07 0.27*** (–3.52) (–3.28) (–2.52) (–3.27) (–2.87) (–3.28) (–2.13) (–0.26) (–0.39) (1.38) (3.59) Carhart $$\alpha $$ –0.16 –0.12 –0.08 –0.10 –0.09 –0.11 –0.08 –0.03 –0.03 0.04 0.20*** (–2.95) (–2.84) (–2.29) (–2.8) (–2.39) (–3.16) (–2.24) (–0.72) (–0.76) (0.82) (3.23) Pastor-Stambaugh $$\alpha $$ –0.17 –0.13 –0.09 –0.11 –0.09 –0.12 –0.08 –0.03 –0.03 0.05 0.22*** (–3.17) (–3.06) (–2.49) (–3.08) (–2.64) (–3.43) (–2.4) (–0.79) (–0.79) (0.93) (3.49) Gross returns Average return 0.97 1.00 1.03 1.00 1.02 1.01 1.05 1.11 1.10 1.20 0.23*** (3.83) (4.14) (4.35) (4.29) (4.35) (4.29) (4.41) (4.57) (4.46) (4.56) (3.14) CAPM $$\alpha $$ –0.08 –0.02 0.02 0.00 0.01 0.00 0.03 0.08 0.07 0.13 0.21*** (–1.2) (–0.47) (0.38) (–0.03) (0.35) (0.02) (0.79) (1.74) (1.18) (1.74) (2.97) Fama-French $$\alpha $$ –0.09 –0.04 0.00 –0.02 0.00 –0.01 0.02 0.09 0.08 0.18 0.27*** (–1.6) (–0.92) (0.11) (–0.57) (–0.04) (–0.42) (0.72) (2.44) (2.00) (3.33) (3.59) Carhart $$\alpha $$ –0.05 –0.02 0.01 –0.01 0.01 –0.02 0.02 0.07 0.07 0.15 0.20*** (–0.98) (–0.54) (0.29) (–0.24) (0.23) (–0.47) (0.46) (1.88) (1.69) (2.84) (3.23) Pastor-Stambaugh $$\alpha $$ –0.06 –0.03 0.00 –0.02 0.00 –0.02 0.01 0.07 0.07 0.15 0.22*** (–1.19) (–0.74) (0.05) (–0.51) (–0.01) (–0.7) (0.28) (1.79) (1.62) (2.89) (3.49) B. Decile portfolios formed according to the trade-based AFP Net returns Average return 0.87 0.92 0.92 0.90 0.93 0.88 0.88 0.93 0.95 0.99 0.12*** (3.23) (3.73) (3.80) (3.79) (3.88) (3.72) (3.67) (3.77) (3.82) (3.63) (2.98) CAPM $$\alpha $$ –0.16 –0.06 –0.05 –0.05 –0.03 –0.07 –0.08 –0.05 –0.03 –0.04 0.12*** (–2.29) (–1.15) (–1.03) (–1.22) (–0.76) (–1.83) (–2.01) (–0.99) (–0.63) (–0.55) (2.94) Fama-French $$\alpha$$ –0.14 –0.07 –0.08 –0.09 –0.05 –0.10 –0.11 –0.06 –0.03 0.01 0.15*** (–3.18) (–1.99) (–2.07) (–2.35) (–1.57) (–2.87) (–3.04) (–1.45) (–0.76) (0.11) (3.93) Carhart $$\alpha $$ –0.15 –0.07 –0.07 –0.07 –0.04 –0.08 –0.10 –0.07 –0.04 –0.03 0.12*** (–3.28) (–1.63) (–1.89) (–1.92) (–1.15) (–2.32) (–2.62) (–1.85) (–0.96) (–0.7) (3.25) Pastor-Stambaugh $$\alpha $$ –0.15 –0.07 –0.08 –0.08 –0.05 –0.09 –0.11 –0.08 –0.04 –0.03 0.12*** (–3.23) (–1.76) (–2.06) (–2.18) (–1.33) (–2.62) (–2.93) (–2.03) (–1.14) (–0.62) (3.34) Gross returns Average return 0.98 1.02 1.01 1.00 1.02 0.97 0.97 1.03 1.05 1.09 0.12*** (3.62) (4.13) (4.20) (4.18) (4.27) (4.12) (4.06) (4.16) (4.21) (4.03) (3.02) CAPM $$\alpha $$ –0.05 0.04 0.05 0.04 0.06 0.02 0.01 0.05 0.07 0.06 0.12*** (–0.76) (0.85) (1.05) (0.93) (1.45) (0.52) (0.27) (1.00) (1.29) (0.84) (2.98) Fama-French $$\alpha $$ –0.03 0.03 0.02 0.01 0.04 –0.01 –0.01 0.04 0.07 0.11 0.15*** (–0.79) (0.68) (0.54) (0.21) (1.14) (–0.15) (–0.36) (1.08) (1.92) (2.47) (3.98) Carhart $$\alpha $$ –0.05 0.03 0.02 0.02 0.05 0.01 0.00 0.02 0.06 0.07 0.12*** (–1.03) (0.78) (0.57) (0.57) (1.37) (0.32) (–0.02) (0.63) (1.63) (1.53) (3.29) Pastor-Stambaugh $$\alpha $$ –0.05 0.03 0.01 0.01 0.04 0.00 –0.01 0.02 0.06 0.08 0.12*** (–1.01) (0.62) (0.38) (0.29) (1.13) (0.01) (–0.33) (0.46) (1.47) (1.59) (3.39) A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net returns Average return 0.87 0.90 0.93 0.91 0.93 0.92 0.96 1.02 1.00 1.10 0.23*** (3.42) (3.74) (3.95) (3.89) (3.96) (3.90) (4.01) (4.18) (4.06) (4.16) (3.14) CAPM $$\alpha $$ –0.18 –0.12 –0.08 –0.10 –0.08 –0.09 –0.06 –0.01 –0.03 0.03 0.21*** (–2.79) (–2.49) (–1.79) (–2.26) (–2.1) (–2.3) (–1.52) (–0.3) (–0.58) (0.37) (2.97) Fama-French $$\alpha$$ –0.19 –0.14 –0.09 –0.11 –0.09 –0.11 –0.07 –0.01 –0.02 0.07 0.27*** (–3.52) (–3.28) (–2.52) (–3.27) (–2.87) (–3.28) (–2.13) (–0.26) (–0.39) (1.38) (3.59) Carhart $$\alpha $$ –0.16 –0.12 –0.08 –0.10 –0.09 –0.11 –0.08 –0.03 –0.03 0.04 0.20*** (–2.95) (–2.84) (–2.29) (–2.8) (–2.39) (–3.16) (–2.24) (–0.72) (–0.76) (0.82) (3.23) Pastor-Stambaugh $$\alpha $$ –0.17 –0.13 –0.09 –0.11 –0.09 –0.12 –0.08 –0.03 –0.03 0.05 0.22*** (–3.17) (–3.06) (–2.49) (–3.08) (–2.64) (–3.43) (–2.4) (–0.79) (–0.79) (0.93) (3.49) Gross returns Average return 0.97 1.00 1.03 1.00 1.02 1.01 1.05 1.11 1.10 1.20 0.23*** (3.83) (4.14) (4.35) (4.29) (4.35) (4.29) (4.41) (4.57) (4.46) (4.56) (3.14) CAPM $$\alpha $$ –0.08 –0.02 0.02 0.00 0.01 0.00 0.03 0.08 0.07 0.13 0.21*** (–1.2) (–0.47) (0.38) (–0.03) (0.35) (0.02) (0.79) (1.74) (1.18) (1.74) (2.97) Fama-French $$\alpha $$ –0.09 –0.04 0.00 –0.02 0.00 –0.01 0.02 0.09 0.08 0.18 0.27*** (–1.6) (–0.92) (0.11) (–0.57) (–0.04) (–0.42) (0.72) (2.44) (2.00) (3.33) (3.59) Carhart $$\alpha $$ –0.05 –0.02 0.01 –0.01 0.01 –0.02 0.02 0.07 0.07 0.15 0.20*** (–0.98) (–0.54) (0.29) (–0.24) (0.23) (–0.47) (0.46) (1.88) (1.69) (2.84) (3.23) Pastor-Stambaugh $$\alpha $$ –0.06 –0.03 0.00 –0.02 0.00 –0.02 0.01 0.07 0.07 0.15 0.22*** (–1.19) (–0.74) (0.05) (–0.51) (–0.01) (–0.7) (0.28) (1.79) (1.62) (2.89) (3.49) B. Decile portfolios formed according to the trade-based AFP Net returns Average return 0.87 0.92 0.92 0.90 0.93 0.88 0.88 0.93 0.95 0.99 0.12*** (3.23) (3.73) (3.80) (3.79) (3.88) (3.72) (3.67) (3.77) (3.82) (3.63) (2.98) CAPM $$\alpha $$ –0.16 –0.06 –0.05 –0.05 –0.03 –0.07 –0.08 –0.05 –0.03 –0.04 0.12*** (–2.29) (–1.15) (–1.03) (–1.22) (–0.76) (–1.83) (–2.01) (–0.99) (–0.63) (–0.55) (2.94) Fama-French $$\alpha$$ –0.14 –0.07 –0.08 –0.09 –0.05 –0.10 –0.11 –0.06 –0.03 0.01 0.15*** (–3.18) (–1.99) (–2.07) (–2.35) (–1.57) (–2.87) (–3.04) (–1.45) (–0.76) (0.11) (3.93) Carhart $$\alpha $$ –0.15 –0.07 –0.07 –0.07 –0.04 –0.08 –0.10 –0.07 –0.04 –0.03 0.12*** (–3.28) (–1.63) (–1.89) (–1.92) (–1.15) (–2.32) (–2.62) (–1.85) (–0.96) (–0.7) (3.25) Pastor-Stambaugh $$\alpha $$ –0.15 –0.07 –0.08 –0.08 –0.05 –0.09 –0.11 –0.08 –0.04 –0.03 0.12*** (–3.23) (–1.76) (–2.06) (–2.18) (–1.33) (–2.62) (–2.93) (–2.03) (–1.14) (–0.62) (3.34) Gross returns Average return 0.98 1.02 1.01 1.00 1.02 0.97 0.97 1.03 1.05 1.09 0.12*** (3.62) (4.13) (4.20) (4.18) (4.27) (4.12) (4.06) (4.16) (4.21) (4.03) (3.02) CAPM $$\alpha $$ –0.05 0.04 0.05 0.04 0.06 0.02 0.01 0.05 0.07 0.06 0.12*** (–0.76) (0.85) (1.05) (0.93) (1.45) (0.52) (0.27) (1.00) (1.29) (0.84) (2.98) Fama-French $$\alpha $$ –0.03 0.03 0.02 0.01 0.04 –0.01 –0.01 0.04 0.07 0.11 0.15*** (–0.79) (0.68) (0.54) (0.21) (1.14) (–0.15) (–0.36) (1.08) (1.92) (2.47) (3.98) Carhart $$\alpha $$ –0.05 0.03 0.02 0.02 0.05 0.01 0.00 0.02 0.06 0.07 0.12*** (–1.03) (0.78) (0.57) (0.57) (1.37) (0.32) (–0.02) (0.63) (1.63) (1.53) (3.29) Pastor-Stambaugh $$\alpha $$ –0.05 0.03 0.01 0.01 0.04 0.00 –0.01 0.02 0.06 0.08 0.12*** (–1.01) (0.62) (0.38) (0.29) (1.13) (0.01) (–0.33) (0.46) (1.47) (1.59) (3.39) This table presents the performance of decile fund portfolios formed according to their AFP. Like in Equation (3), AFP is the sum of the product of active portfolio weights (panel A) or changes in portfolio weights (panel B) and subsequent 3-day abnormal returns surrounding earnings announcements. We formed and rebalanced the decile portfolios at the end of 2 months after each quarter from 1984Q1 to 2014Q2 and the return series ranges from June 1984 to November 2014. Decile 10 is the portfolio of funds with the highest AFP value. We compute monthly equally weighted percentage net and gross (net plus expense ratio) returns on the portfolios, as well as risk-adjusted returns based on the CAPM, the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model, and the Pastor and Stambaugh (2003) five-factor model. We report the alphas as a monthly percentage. $$t$$-statistics are shown in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 2 AFP and mutual fund returns: Decile portfolios A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net returns Average return 0.87 0.90 0.93 0.91 0.93 0.92 0.96 1.02 1.00 1.10 0.23*** (3.42) (3.74) (3.95) (3.89) (3.96) (3.90) (4.01) (4.18) (4.06) (4.16) (3.14) CAPM $$\alpha $$ –0.18 –0.12 –0.08 –0.10 –0.08 –0.09 –0.06 –0.01 –0.03 0.03 0.21*** (–2.79) (–2.49) (–1.79) (–2.26) (–2.1) (–2.3) (–1.52) (–0.3) (–0.58) (0.37) (2.97) Fama-French $$\alpha$$ –0.19 –0.14 –0.09 –0.11 –0.09 –0.11 –0.07 –0.01 –0.02 0.07 0.27*** (–3.52) (–3.28) (–2.52) (–3.27) (–2.87) (–3.28) (–2.13) (–0.26) (–0.39) (1.38) (3.59) Carhart $$\alpha $$ –0.16 –0.12 –0.08 –0.10 –0.09 –0.11 –0.08 –0.03 –0.03 0.04 0.20*** (–2.95) (–2.84) (–2.29) (–2.8) (–2.39) (–3.16) (–2.24) (–0.72) (–0.76) (0.82) (3.23) Pastor-Stambaugh $$\alpha $$ –0.17 –0.13 –0.09 –0.11 –0.09 –0.12 –0.08 –0.03 –0.03 0.05 0.22*** (–3.17) (–3.06) (–2.49) (–3.08) (–2.64) (–3.43) (–2.4) (–0.79) (–0.79) (0.93) (3.49) Gross returns Average return 0.97 1.00 1.03 1.00 1.02 1.01 1.05 1.11 1.10 1.20 0.23*** (3.83) (4.14) (4.35) (4.29) (4.35) (4.29) (4.41) (4.57) (4.46) (4.56) (3.14) CAPM $$\alpha $$ –0.08 –0.02 0.02 0.00 0.01 0.00 0.03 0.08 0.07 0.13 0.21*** (–1.2) (–0.47) (0.38) (–0.03) (0.35) (0.02) (0.79) (1.74) (1.18) (1.74) (2.97) Fama-French $$\alpha $$ –0.09 –0.04 0.00 –0.02 0.00 –0.01 0.02 0.09 0.08 0.18 0.27*** (–1.6) (–0.92) (0.11) (–0.57) (–0.04) (–0.42) (0.72) (2.44) (2.00) (3.33) (3.59) Carhart $$\alpha $$ –0.05 –0.02 0.01 –0.01 0.01 –0.02 0.02 0.07 0.07 0.15 0.20*** (–0.98) (–0.54) (0.29) (–0.24) (0.23) (–0.47) (0.46) (1.88) (1.69) (2.84) (3.23) Pastor-Stambaugh $$\alpha $$ –0.06 –0.03 0.00 –0.02 0.00 –0.02 0.01 0.07 0.07 0.15 0.22*** (–1.19) (–0.74) (0.05) (–0.51) (–0.01) (–0.7) (0.28) (1.79) (1.62) (2.89) (3.49) B. Decile portfolios formed according to the trade-based AFP Net returns Average return 0.87 0.92 0.92 0.90 0.93 0.88 0.88 0.93 0.95 0.99 0.12*** (3.23) (3.73) (3.80) (3.79) (3.88) (3.72) (3.67) (3.77) (3.82) (3.63) (2.98) CAPM $$\alpha $$ –0.16 –0.06 –0.05 –0.05 –0.03 –0.07 –0.08 –0.05 –0.03 –0.04 0.12*** (–2.29) (–1.15) (–1.03) (–1.22) (–0.76) (–1.83) (–2.01) (–0.99) (–0.63) (–0.55) (2.94) Fama-French $$\alpha$$ –0.14 –0.07 –0.08 –0.09 –0.05 –0.10 –0.11 –0.06 –0.03 0.01 0.15*** (–3.18) (–1.99) (–2.07) (–2.35) (–1.57) (–2.87) (–3.04) (–1.45) (–0.76) (0.11) (3.93) Carhart $$\alpha $$ –0.15 –0.07 –0.07 –0.07 –0.04 –0.08 –0.10 –0.07 –0.04 –0.03 0.12*** (–3.28) (–1.63) (–1.89) (–1.92) (–1.15) (–2.32) (–2.62) (–1.85) (–0.96) (–0.7) (3.25) Pastor-Stambaugh $$\alpha $$ –0.15 –0.07 –0.08 –0.08 –0.05 –0.09 –0.11 –0.08 –0.04 –0.03 0.12*** (–3.23) (–1.76) (–2.06) (–2.18) (–1.33) (–2.62) (–2.93) (–2.03) (–1.14) (–0.62) (3.34) Gross returns Average return 0.98 1.02 1.01 1.00 1.02 0.97 0.97 1.03 1.05 1.09 0.12*** (3.62) (4.13) (4.20) (4.18) (4.27) (4.12) (4.06) (4.16) (4.21) (4.03) (3.02) CAPM $$\alpha $$ –0.05 0.04 0.05 0.04 0.06 0.02 0.01 0.05 0.07 0.06 0.12*** (–0.76) (0.85) (1.05) (0.93) (1.45) (0.52) (0.27) (1.00) (1.29) (0.84) (2.98) Fama-French $$\alpha $$ –0.03 0.03 0.02 0.01 0.04 –0.01 –0.01 0.04 0.07 0.11 0.15*** (–0.79) (0.68) (0.54) (0.21) (1.14) (–0.15) (–0.36) (1.08) (1.92) (2.47) (3.98) Carhart $$\alpha $$ –0.05 0.03 0.02 0.02 0.05 0.01 0.00 0.02 0.06 0.07 0.12*** (–1.03) (0.78) (0.57) (0.57) (1.37) (0.32) (–0.02) (0.63) (1.63) (1.53) (3.29) Pastor-Stambaugh $$\alpha $$ –0.05 0.03 0.01 0.01 0.04 0.00 –0.01 0.02 0.06 0.08 0.12*** (–1.01) (0.62) (0.38) (0.29) (1.13) (0.01) (–0.33) (0.46) (1.47) (1.59) (3.39) A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net returns Average return 0.87 0.90 0.93 0.91 0.93 0.92 0.96 1.02 1.00 1.10 0.23*** (3.42) (3.74) (3.95) (3.89) (3.96) (3.90) (4.01) (4.18) (4.06) (4.16) (3.14) CAPM $$\alpha $$ –0.18 –0.12 –0.08 –0.10 –0.08 –0.09 –0.06 –0.01 –0.03 0.03 0.21*** (–2.79) (–2.49) (–1.79) (–2.26) (–2.1) (–2.3) (–1.52) (–0.3) (–0.58) (0.37) (2.97) Fama-French $$\alpha$$ –0.19 –0.14 –0.09 –0.11 –0.09 –0.11 –0.07 –0.01 –0.02 0.07 0.27*** (–3.52) (–3.28) (–2.52) (–3.27) (–2.87) (–3.28) (–2.13) (–0.26) (–0.39) (1.38) (3.59) Carhart $$\alpha $$ –0.16 –0.12 –0.08 –0.10 –0.09 –0.11 –0.08 –0.03 –0.03 0.04 0.20*** (–2.95) (–2.84) (–2.29) (–2.8) (–2.39) (–3.16) (–2.24) (–0.72) (–0.76) (0.82) (3.23) Pastor-Stambaugh $$\alpha $$ –0.17 –0.13 –0.09 –0.11 –0.09 –0.12 –0.08 –0.03 –0.03 0.05 0.22*** (–3.17) (–3.06) (–2.49) (–3.08) (–2.64) (–3.43) (–2.4) (–0.79) (–0.79) (0.93) (3.49) Gross returns Average return 0.97 1.00 1.03 1.00 1.02 1.01 1.05 1.11 1.10 1.20 0.23*** (3.83) (4.14) (4.35) (4.29) (4.35) (4.29) (4.41) (4.57) (4.46) (4.56) (3.14) CAPM $$\alpha $$ –0.08 –0.02 0.02 0.00 0.01 0.00 0.03 0.08 0.07 0.13 0.21*** (–1.2) (–0.47) (0.38) (–0.03) (0.35) (0.02) (0.79) (1.74) (1.18) (1.74) (2.97) Fama-French $$\alpha $$ –0.09 –0.04 0.00 –0.02 0.00 –0.01 0.02 0.09 0.08 0.18 0.27*** (–1.6) (–0.92) (0.11) (–0.57) (–0.04) (–0.42) (0.72) (2.44) (2.00) (3.33) (3.59) Carhart $$\alpha $$ –0.05 –0.02 0.01 –0.01 0.01 –0.02 0.02 0.07 0.07 0.15 0.20*** (–0.98) (–0.54) (0.29) (–0.24) (0.23) (–0.47) (0.46) (1.88) (1.69) (2.84) (3.23) Pastor-Stambaugh $$\alpha $$ –0.06 –0.03 0.00 –0.02 0.00 –0.02 0.01 0.07 0.07 0.15 0.22*** (–1.19) (–0.74) (0.05) (–0.51) (–0.01) (–0.7) (0.28) (1.79) (1.62) (2.89) (3.49) B. Decile portfolios formed according to the trade-based AFP Net returns Average return 0.87 0.92 0.92 0.90 0.93 0.88 0.88 0.93 0.95 0.99 0.12*** (3.23) (3.73) (3.80) (3.79) (3.88) (3.72) (3.67) (3.77) (3.82) (3.63) (2.98) CAPM $$\alpha $$ –0.16 –0.06 –0.05 –0.05 –0.03 –0.07 –0.08 –0.05 –0.03 –0.04 0.12*** (–2.29) (–1.15) (–1.03) (–1.22) (–0.76) (–1.83) (–2.01) (–0.99) (–0.63) (–0.55) (2.94) Fama-French $$\alpha$$ –0.14 –0.07 –0.08 –0.09 –0.05 –0.10 –0.11 –0.06 –0.03 0.01 0.15*** (–3.18) (–1.99) (–2.07) (–2.35) (–1.57) (–2.87) (–3.04) (–1.45) (–0.76) (0.11) (3.93) Carhart $$\alpha $$ –0.15 –0.07 –0.07 –0.07 –0.04 –0.08 –0.10 –0.07 –0.04 –0.03 0.12*** (–3.28) (–1.63) (–1.89) (–1.92) (–1.15) (–2.32) (–2.62) (–1.85) (–0.96) (–0.7) (3.25) Pastor-Stambaugh $$\alpha $$ –0.15 –0.07 –0.08 –0.08 –0.05 –0.09 –0.11 –0.08 –0.04 –0.03 0.12*** (–3.23) (–1.76) (–2.06) (–2.18) (–1.33) (–2.62) (–2.93) (–2.03) (–1.14) (–0.62) (3.34) Gross returns Average return 0.98 1.02 1.01 1.00 1.02 0.97 0.97 1.03 1.05 1.09 0.12*** (3.62) (4.13) (4.20) (4.18) (4.27) (4.12) (4.06) (4.16) (4.21) (4.03) (3.02) CAPM $$\alpha $$ –0.05 0.04 0.05 0.04 0.06 0.02 0.01 0.05 0.07 0.06 0.12*** (–0.76) (0.85) (1.05) (0.93) (1.45) (0.52) (0.27) (1.00) (1.29) (0.84) (2.98) Fama-French $$\alpha $$ –0.03 0.03 0.02 0.01 0.04 –0.01 –0.01 0.04 0.07 0.11 0.15*** (–0.79) (0.68) (0.54) (0.21) (1.14) (–0.15) (–0.36) (1.08) (1.92) (2.47) (3.98) Carhart $$\alpha $$ –0.05 0.03 0.02 0.02 0.05 0.01 0.00 0.02 0.06 0.07 0.12*** (–1.03) (0.78) (0.57) (0.57) (1.37) (0.32) (–0.02) (0.63) (1.63) (1.53) (3.29) Pastor-Stambaugh $$\alpha $$ –0.05 0.03 0.01 0.01 0.04 0.00 –0.01 0.02 0.06 0.08 0.12*** (–1.01) (0.62) (0.38) (0.29) (1.13) (0.01) (–0.33) (0.46) (1.47) (1.59) (3.39) This table presents the performance of decile fund portfolios formed according to their AFP. Like in Equation (3), AFP is the sum of the product of active portfolio weights (panel A) or changes in portfolio weights (panel B) and subsequent 3-day abnormal returns surrounding earnings announcements. We formed and rebalanced the decile portfolios at the end of 2 months after each quarter from 1984Q1 to 2014Q2 and the return series ranges from June 1984 to November 2014. Decile 10 is the portfolio of funds with the highest AFP value. We compute monthly equally weighted percentage net and gross (net plus expense ratio) returns on the portfolios, as well as risk-adjusted returns based on the CAPM, the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model, and the Pastor and Stambaugh (2003) five-factor model. We report the alphas as a monthly percentage. $$t$$-statistics are shown in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. We conjecture that the relatively weaker results using trade-based measure may be due to confounding empirical factors, such as the fact that fund flows often induce substantial trading activities by mutual funds (e.g., Lou 2012; Vayanos and Woolley 2013). Moreover, active weights capture a broader information set of fund managers revealed through both recent and longer-term holdings, whereas changes in portfolio weights are driven by their recent portfolio adjustments and thus reflect only their recent information. Panel B shows the results based on gross fund returns by adding back expenses, which potentially provide a clearer picture of the value in terms of the alpha created by fund managers. These results indicate that fund managers with high index-based (trade-based) AFP produce a monthly gross Carhart four-factor alpha of 15 (7) bps, with a $$t$$-statistic of 2.84 (1.53), whereas managers with low AFP produce a monthly gross four-factor alpha of $$-$$5 ($$-$$5) bps that is statistically indistinguishable from 0, before expenses. This finding lends further support to the notion that fund managers with high AFP tend to be skilled. 3.1.1 Accounting for the post-earnings announcement drift Researchers have documented the post-earnings announcement drift (PEAD), a tendency for stock prices to drift in the direction of earnings surprises in the weeks following earnings announcements (e.g., Ball and Brown 1968; Bernard and Thomas 1989). Although we could also view exploiting PEAD as an indicator of investment skill, we aim to distinguish between a fund manager’s ability to forecast firms’ fundamental values and his or her ability to trade on released earnings news. To examine whether the forecasting power of AFP is driven by PEAD, we form hedge portfolios that replicate the payoffs of strategies exploiting PEADs. Specifically, we follow Livnat and Mendenhall (2006) and compute the standardized earnings surprise (SUE) for each stock in each quarter: \begin{align} SUE_{i,t}=\frac{X_{i,t}-E(X_{i,t})}{P_{i,t}}, \end{align} (5) where $$X_{i,t}$$ is earnings per share for stock $$i $$ in quarter $$t$$, E($$X_{i,t})$$ is expected earnings per share for stock $$i $$ in quarter $$t$$, and $$P_{i,t}$$ is the price for stock $$i $$ at the end of quarter $$t$$. We use the forecasts of a seasonal random walk model and consensus analyst earnings forecasts as proxies for expected earnings per share. Our primary measure of quarterly earnings is based on the primary earnings per share before extraordinary items. As a robustness check, we also consider the earnings surprises after excluding special items. We label the SUE based on the seasonal random walk model as SUE1, the SUE after excluding special items as SUE2, and the SUE based on consensus analyst forecasts as SUE3. At the end of each month, we form decile portfolios based on the SUE in the previous month. We then compute the equally weighted returns on a strategy that buys stocks in the top three deciles (high SUE) and short sells stocks in the bottom three deciles (low SUE). In turn, PEAD1, PEAD2, and PEAD3 refer to the returns on the three strategies based on alternative calculations of SUE. The results in Table 3 show that, even after we control for the effect of PEAD, the superior performance of high AFP funds remains economically and statistically significant. This finding suggests that the superior performance of high AFP funds cannot be explained by a strategy to exploit PEAD.5 Table 3 AFP and mutual fund returns: Controlling for the influence of the post-earnings announcement drift A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net return Six-factor: PEAD1 –0.30 –0.26 –0.20 –0.22 –0.20 –0.21 –0.18 –0.10 –0.11 –0.01 0.29*** (–4.14) (–4.86) (–4.66) (–4.75) (–4.37) (–4.68) (–3.75) (–2) (–1.74) (–0.09) (3.42) Six-factor: PEAD2 –0.32 –0.29 –0.22 –0.24 –0.22 –0.22 –0.19 –0.11 –0.11 –0.02 0.30*** (–4.46) (–5.32) (–5.16) (–5.23) (–4.83) (–5.13) (–4.09) (–2.22) (–1.84) (–0.21) (3.43) Six-factor: PEAD3 –0.23 –0.17 –0.13 –0.15 –0.12 –0.16 –0.12 –0.06 –0.08 0.01 0.24*** (–4.38) (–3.93) (–3.6) (–4.16) (–3.42) (–4.54) (–3.57) (–1.58) (–1.9) (0.20) (3.79) Gross return Six-factor: PEAD1 –0.19 –0.17 –0.11 –0.13 –0.11 –0.11 –0.09 –0.01 –0.01 0.10 0.29*** (–2.69) (–3.07) (–2.51) (–2.76) (–2.36) (–2.58) (–1.81) (–0.14) (–0.15) (1.23) (3.43) Six-factor: PEAD2 –0.22 –0.19 –0.13 –0.15 –0.12 –0.13 –0.10 –0.02 –0.01 0.09 0.30*** (–3.01) (–3.51) (–2.96) (–3.2) (–2.77) (–2.98) (–2.09) (–0.33) (–0.23) (1.11) (3.43) Six-factor: PEAD3 –0.13 –0.07 –0.04 –0.06 –0.03 –0.06 –0.03 0.04 0.02 0.12 0.24*** (–2.4) (–1.62) (–1) (–1.61) (–0.86) (–1.83) (–0.89) (1.02) (0.49) (2.06) (3.79) Net return Six-factor: PEAD1 –0.24 –0.17 –0.19 –0.20 –0.15 –0.19 –0.21 –0.17 –0.14 –0.08 0.16*** (–3.56) (–2.97) (–4.02) (–4.46) (–2.84) (–4.28) (–4.91) (–3.12) (–3) (–1.11) (4.08) Six-factor: PEAD2 –0.26 –0.19 –0.21 –0.21 –0.16 –0.20 –0.22 –0.18 –0.15 –0.09 0.17*** (–3.86) (–3.33) (–4.48) (–4.93) (–3.24) (–4.66) (–5.37) (–3.35) (–3.32) (–1.25) (4.19) Six-factor: PEAD3 –0.22 –0.13 –0.12 –0.13 –0.09 –0.13 –0.15 –0.12 –0.09 –0.07 0.14*** (–4.53) (–3.17) (–3.11) (–3.43) (–2.29) (–3.79) (–4.22) (–3.01) (–2.33) (–1.52) (3.50) Gross return Six-factor: PEAD1 –0.13 –0.07 –0.10 –0.10 –0.05 –0.09 –0.12 –0.07 –0.04 0.03 0.16*** (–1.99) (–1.27) (–2.01) (–2.34) (–1.04) (–2.15) (–2.71) (–1.34) (–0.86) (0.39) (4.12) Six-factor: PEAD2 –0.15 –0.09 –0.11 –0.12 –0.07 –0.11 –0.13 –0.08 –0.05 0.02 0.17*** (–2.27) (–1.6) (–2.44) (–2.76) (–1.38) (–2.48) (–3.12) (–1.54) (–1.12) (0.25) (4.23) Six-factor: PEAD3 –0.11 –0.03 –0.02 –0.04 0.01 –0.04 –0.05 –0.02 0.01 0.03 0.15*** (–2.35) (–0.74) (–0.64) (–0.99) (0.15) (–1.07) (–1.55) (–0.59) (0.23) (0.67) (3.54) A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net return Six-factor: PEAD1 –0.30 –0.26 –0.20 –0.22 –0.20 –0.21 –0.18 –0.10 –0.11 –0.01 0.29*** (–4.14) (–4.86) (–4.66) (–4.75) (–4.37) (–4.68) (–3.75) (–2) (–1.74) (–0.09) (3.42) Six-factor: PEAD2 –0.32 –0.29 –0.22 –0.24 –0.22 –0.22 –0.19 –0.11 –0.11 –0.02 0.30*** (–4.46) (–5.32) (–5.16) (–5.23) (–4.83) (–5.13) (–4.09) (–2.22) (–1.84) (–0.21) (3.43) Six-factor: PEAD3 –0.23 –0.17 –0.13 –0.15 –0.12 –0.16 –0.12 –0.06 –0.08 0.01 0.24*** (–4.38) (–3.93) (–3.6) (–4.16) (–3.42) (–4.54) (–3.57) (–1.58) (–1.9) (0.20) (3.79) Gross return Six-factor: PEAD1 –0.19 –0.17 –0.11 –0.13 –0.11 –0.11 –0.09 –0.01 –0.01 0.10 0.29*** (–2.69) (–3.07) (–2.51) (–2.76) (–2.36) (–2.58) (–1.81) (–0.14) (–0.15) (1.23) (3.43) Six-factor: PEAD2 –0.22 –0.19 –0.13 –0.15 –0.12 –0.13 –0.10 –0.02 –0.01 0.09 0.30*** (–3.01) (–3.51) (–2.96) (–3.2) (–2.77) (–2.98) (–2.09) (–0.33) (–0.23) (1.11) (3.43) Six-factor: PEAD3 –0.13 –0.07 –0.04 –0.06 –0.03 –0.06 –0.03 0.04 0.02 0.12 0.24*** (–2.4) (–1.62) (–1) (–1.61) (–0.86) (–1.83) (–0.89) (1.02) (0.49) (2.06) (3.79) Net return Six-factor: PEAD1 –0.24 –0.17 –0.19 –0.20 –0.15 –0.19 –0.21 –0.17 –0.14 –0.08 0.16*** (–3.56) (–2.97) (–4.02) (–4.46) (–2.84) (–4.28) (–4.91) (–3.12) (–3) (–1.11) (4.08) Six-factor: PEAD2 –0.26 –0.19 –0.21 –0.21 –0.16 –0.20 –0.22 –0.18 –0.15 –0.09 0.17*** (–3.86) (–3.33) (–4.48) (–4.93) (–3.24) (–4.66) (–5.37) (–3.35) (–3.32) (–1.25) (4.19) Six-factor: PEAD3 –0.22 –0.13 –0.12 –0.13 –0.09 –0.13 –0.15 –0.12 –0.09 –0.07 0.14*** (–4.53) (–3.17) (–3.11) (–3.43) (–2.29) (–3.79) (–4.22) (–3.01) (–2.33) (–1.52) (3.50) Gross return Six-factor: PEAD1 –0.13 –0.07 –0.10 –0.10 –0.05 –0.09 –0.12 –0.07 –0.04 0.03 0.16*** (–1.99) (–1.27) (–2.01) (–2.34) (–1.04) (–2.15) (–2.71) (–1.34) (–0.86) (0.39) (4.12) Six-factor: PEAD2 –0.15 –0.09 –0.11 –0.12 –0.07 –0.11 –0.13 –0.08 –0.05 0.02 0.17*** (–2.27) (–1.6) (–2.44) (–2.76) (–1.38) (–2.48) (–3.12) (–1.54) (–1.12) (0.25) (4.23) Six-factor: PEAD3 –0.11 –0.03 –0.02 –0.04 0.01 –0.04 –0.05 –0.02 0.01 0.03 0.15*** (–2.35) (–0.74) (–0.64) (–0.99) (0.15) (–1.07) (–1.55) (–0.59) (0.23) (0.67) (3.54) This table presents the performance of decile fund portfolios formed according to their AFP, controlling for the influence of PEAD. Specifically, we construct hedge portfolios that seek to replicate the payoffs of strategies exploiting the post-earnings announcement drift. We follow Livnat and Mendenhall (2006) and compute the SUE for each stock in each quarter: we use the seasonal random walk model and consensus analyst earnings forecast to proxy for expected earnings per share. We use the primary earnings per share before extraordinary items as our primary measure of quarterly earnings, and we consider the earnings surprise after the exclusion of special items. We label the SUE according to the seasonal random walk model as SUE1, the SUE after the exclusion of special items as SUE2, and the SUE based on consensus analyst forecasts as SUE3. At the end of each month, we form decile portfolios based on the SUE in the previous month and compute the equally weighted returns on a strategy that buys stocks in the top three deciles with high SUE and shorts stocks in the bottom three deciles with low SUE. The returns on the three strategies based on three SUEs are called returns to PEAD1, PEAD2, and PEAD3. We report the alphas as a monthly percentage using three versions of the six-factor models that augment the five-factor model in Table 3 ($$\alpha )$$ with the return to a strategy that exploits PEAD. The $$t$$-statistics are shown in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Table 3 AFP and mutual fund returns: Controlling for the influence of the post-earnings announcement drift A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net return Six-factor: PEAD1 –0.30 –0.26 –0.20 –0.22 –0.20 –0.21 –0.18 –0.10 –0.11 –0.01 0.29*** (–4.14) (–4.86) (–4.66) (–4.75) (–4.37) (–4.68) (–3.75) (–2) (–1.74) (–0.09) (3.42) Six-factor: PEAD2 –0.32 –0.29 –0.22 –0.24 –0.22 –0.22 –0.19 –0.11 –0.11 –0.02 0.30*** (–4.46) (–5.32) (–5.16) (–5.23) (–4.83) (–5.13) (–4.09) (–2.22) (–1.84) (–0.21) (3.43) Six-factor: PEAD3 –0.23 –0.17 –0.13 –0.15 –0.12 –0.16 –0.12 –0.06 –0.08 0.01 0.24*** (–4.38) (–3.93) (–3.6) (–4.16) (–3.42) (–4.54) (–3.57) (–1.58) (–1.9) (0.20) (3.79) Gross return Six-factor: PEAD1 –0.19 –0.17 –0.11 –0.13 –0.11 –0.11 –0.09 –0.01 –0.01 0.10 0.29*** (–2.69) (–3.07) (–2.51) (–2.76) (–2.36) (–2.58) (–1.81) (–0.14) (–0.15) (1.23) (3.43) Six-factor: PEAD2 –0.22 –0.19 –0.13 –0.15 –0.12 –0.13 –0.10 –0.02 –0.01 0.09 0.30*** (–3.01) (–3.51) (–2.96) (–3.2) (–2.77) (–2.98) (–2.09) (–0.33) (–0.23) (1.11) (3.43) Six-factor: PEAD3 –0.13 –0.07 –0.04 –0.06 –0.03 –0.06 –0.03 0.04 0.02 0.12 0.24*** (–2.4) (–1.62) (–1) (–1.61) (–0.86) (–1.83) (–0.89) (1.02) (0.49) (2.06) (3.79) Net return Six-factor: PEAD1 –0.24 –0.17 –0.19 –0.20 –0.15 –0.19 –0.21 –0.17 –0.14 –0.08 0.16*** (–3.56) (–2.97) (–4.02) (–4.46) (–2.84) (–4.28) (–4.91) (–3.12) (–3) (–1.11) (4.08) Six-factor: PEAD2 –0.26 –0.19 –0.21 –0.21 –0.16 –0.20 –0.22 –0.18 –0.15 –0.09 0.17*** (–3.86) (–3.33) (–4.48) (–4.93) (–3.24) (–4.66) (–5.37) (–3.35) (–3.32) (–1.25) (4.19) Six-factor: PEAD3 –0.22 –0.13 –0.12 –0.13 –0.09 –0.13 –0.15 –0.12 –0.09 –0.07 0.14*** (–4.53) (–3.17) (–3.11) (–3.43) (–2.29) (–3.79) (–4.22) (–3.01) (–2.33) (–1.52) (3.50) Gross return Six-factor: PEAD1 –0.13 –0.07 –0.10 –0.10 –0.05 –0.09 –0.12 –0.07 –0.04 0.03 0.16*** (–1.99) (–1.27) (–2.01) (–2.34) (–1.04) (–2.15) (–2.71) (–1.34) (–0.86) (0.39) (4.12) Six-factor: PEAD2 –0.15 –0.09 –0.11 –0.12 –0.07 –0.11 –0.13 –0.08 –0.05 0.02 0.17*** (–2.27) (–1.6) (–2.44) (–2.76) (–1.38) (–2.48) (–3.12) (–1.54) (–1.12) (0.25) (4.23) Six-factor: PEAD3 –0.11 –0.03 –0.02 –0.04 0.01 –0.04 –0.05 –0.02 0.01 0.03 0.15*** (–2.35) (–0.74) (–0.64) (–0.99) (0.15) (–1.07) (–1.55) (–0.59) (0.23) (0.67) (3.54) A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net return Six-factor: PEAD1 –0.30 –0.26 –0.20 –0.22 –0.20 –0.21 –0.18 –0.10 –0.11 –0.01 0.29*** (–4.14) (–4.86) (–4.66) (–4.75) (–4.37) (–4.68) (–3.75) (–2) (–1.74) (–0.09) (3.42) Six-factor: PEAD2 –0.32 –0.29 –0.22 –0.24 –0.22 –0.22 –0.19 –0.11 –0.11 –0.02 0.30*** (–4.46) (–5.32) (–5.16) (–5.23) (–4.83) (–5.13) (–4.09) (–2.22) (–1.84) (–0.21) (3.43) Six-factor: PEAD3 –0.23 –0.17 –0.13 –0.15 –0.12 –0.16 –0.12 –0.06 –0.08 0.01 0.24*** (–4.38) (–3.93) (–3.6) (–4.16) (–3.42) (–4.54) (–3.57) (–1.58) (–1.9) (0.20) (3.79) Gross return Six-factor: PEAD1 –0.19 –0.17 –0.11 –0.13 –0.11 –0.11 –0.09 –0.01 –0.01 0.10 0.29*** (–2.69) (–3.07) (–2.51) (–2.76) (–2.36) (–2.58) (–1.81) (–0.14) (–0.15) (1.23) (3.43) Six-factor: PEAD2 –0.22 –0.19 –0.13 –0.15 –0.12 –0.13 –0.10 –0.02 –0.01 0.09 0.30*** (–3.01) (–3.51) (–2.96) (–3.2) (–2.77) (–2.98) (–2.09) (–0.33) (–0.23) (1.11) (3.43) Six-factor: PEAD3 –0.13 –0.07 –0.04 –0.06 –0.03 –0.06 –0.03 0.04 0.02 0.12 0.24*** (–2.4) (–1.62) (–1) (–1.61) (–0.86) (–1.83) (–0.89) (1.02) (0.49) (2.06) (3.79) Net return Six-factor: PEAD1 –0.24 –0.17 –0.19 –0.20 –0.15 –0.19 –0.21 –0.17 –0.14 –0.08 0.16*** (–3.56) (–2.97) (–4.02) (–4.46) (–2.84) (–4.28) (–4.91) (–3.12) (–3) (–1.11) (4.08) Six-factor: PEAD2 –0.26 –0.19 –0.21 –0.21 –0.16 –0.20 –0.22 –0.18 –0.15 –0.09 0.17*** (–3.86) (–3.33) (–4.48) (–4.93) (–3.24) (–4.66) (–5.37) (–3.35) (–3.32) (–1.25) (4.19) Six-factor: PEAD3 –0.22 –0.13 –0.12 –0.13 –0.09 –0.13 –0.15 –0.12 –0.09 –0.07 0.14*** (–4.53) (–3.17) (–3.11) (–3.43) (–2.29) (–3.79) (–4.22) (–3.01) (–2.33) (–1.52) (3.50) Gross return Six-factor: PEAD1 –0.13 –0.07 –0.10 –0.10 –0.05 –0.09 –0.12 –0.07 –0.04 0.03 0.16*** (–1.99) (–1.27) (–2.01) (–2.34) (–1.04) (–2.15) (–2.71) (–1.34) (–0.86) (0.39) (4.12) Six-factor: PEAD2 –0.15 –0.09 –0.11 –0.12 –0.07 –0.11 –0.13 –0.08 –0.05 0.02 0.17*** (–2.27) (–1.6) (–2.44) (–2.76) (–1.38) (–2.48) (–3.12) (–1.54) (–1.12) (0.25) (4.23) Six-factor: PEAD3 –0.11 –0.03 –0.02 –0.04 0.01 –0.04 –0.05 –0.02 0.01 0.03 0.15*** (–2.35) (–0.74) (–0.64) (–0.99) (0.15) (–1.07) (–1.55) (–0.59) (0.23) (0.67) (3.54) This table presents the performance of decile fund portfolios formed according to their AFP, controlling for the influence of PEAD. Specifically, we construct hedge portfolios that seek to replicate the payoffs of strategies exploiting the post-earnings announcement drift. We follow Livnat and Mendenhall (2006) and compute the SUE for each stock in each quarter: we use the seasonal random walk model and consensus analyst earnings forecast to proxy for expected earnings per share. We use the primary earnings per share before extraordinary items as our primary measure of quarterly earnings, and we consider the earnings surprise after the exclusion of special items. We label the SUE according to the seasonal random walk model as SUE1, the SUE after the exclusion of special items as SUE2, and the SUE based on consensus analyst forecasts as SUE3. At the end of each month, we form decile portfolios based on the SUE in the previous month and compute the equally weighted returns on a strategy that buys stocks in the top three deciles with high SUE and shorts stocks in the bottom three deciles with low SUE. The returns on the three strategies based on three SUEs are called returns to PEAD1, PEAD2, and PEAD3. We report the alphas as a monthly percentage using three versions of the six-factor models that augment the five-factor model in Table 3 ($$\alpha )$$ with the return to a strategy that exploits PEAD. The $$t$$-statistics are shown in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. 3.2 Double sorts In this subsection, we evaluate the robustness of the performance predictive power of AFP using double sorts, for two main reasons. First, researchers have reported a number of fund characteristics that are related to fund skill; it is therefore important to assess in detail the incremental forecasting power of AFP for fund performance relative to these characteristics. Second, if different proxies of fund skill contain complementary information, it is possible to find potential interactions between AFP and these fund characteristics. We examine CS (Daniel et al. 1997), the return gap (Kacperczyk, Sialm and Zheng 2008), RPI (Kacperczyk and Seru 2007), and active shares (Cremers and Petajisto 2009).6 The CS measure is the characteristic-adjusted return on a fund’s stock holdings. The characteristic benchmarks are formed on the basis of size, industry-adjusted book-to-market, and momentum, following Daniel et al. (1997). The return gap (Kacperczyk, Sialm and Zheng 2008) reflects the difference between a fund’s realized return and the hypothetical fund return implied by the fund’s stock holdings. It captures the value created by the fund manager due to unobserved actions, such as interim trades. A fund’s reliance on public information (Kacperczyk and Seru 2007) measures the extent to which a fund’s trades can be explained by analyst stock recommendations. The active share variable (Cremers and Petajisto 2009) measures the extent to which a fund manager’s portfolio holdings deviate from those of the benchmark index. We provide detailed descriptions of these measures in Internet Appendix B. The results in Table 4 indicate that AFP has substantial incremental forecasting power for future fund performance. Among the five variables we consider, AFP and return gap significantly predict future fund performance.7 AFP also shows strong performance forecasting power after we control for alternative measures of skill. For instance, Table 4, panel A.1, provides results using double sorts on AFP and the CS measure. The findings indicate that AFP predicts future fund performance, after controlling for CS. The return spread between funds with high and low AFP is large and statistically significant among funds in the top three quartiles. Table 4 Performance predictive power of AFP: Double sorts A. Double sorts on index-based AFP and fund characteristics AFP All Low 2 3 High High-low A.1. CS All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low CS –0.02 –0.05 –0.05 –0.01 0.00 0.04 (–0.34) (–0.68) (–0.82) (–0.25) (–0.04) (0.83) 2 –0.06 –0.10 –0.05 –0.08 –0.01 0.10** (–1.59) (–2.26) (–1.23) (–2.01) (–0.12) (2.28) 3 –0.10 –0.13 –0.15 –0.08 0.01 0.13*** (–3.08) (–3.12) (–3.83) (–2.2) (0.16) (2.87) High CS –0.13 –0.19 –0.14 –0.13 –0.02 0.18*** (–2.31) (–3.3) (–2.18) (–2.56) (–0.29) (3.05) High-low –0.10 –0.15* –0.09 –0.12 –0.02 0.13* (–1.21) (–1.72) (–0.88) (–1.43) (–0.17) (1.82) A.2 Return gap All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low return gap –0.14 –0.20 –0.14 –0.12 –0.08 0.12** (–3.38) (–4.16) (–2.82) (–2.83) (–1.62) (2.47) 2 –0.08 –0.11 –0.11 –0.10 0.01 0.13** (–2.5) (–2.48) (–3.01) (–2.71) (0.26) (2.27) 3 –0.07 –0.08 –0.09 –0.05 –0.02 0.06 (–1.71) (–1.62) (–2.15) (–1.22) (–0.41) (1.35) High return gap –0.03 –0.09 –0.09 –0.08 0.08 0.17*** (–0.78) (–1.63) (–1.8) (–1.61) (1.58) (3.44) High-low 0.11*** 0.11** 0.06 0.04 0.17*** 0.06 (2.71) (2.22) (1.06) (0.67) (3.56) (1.09) A.3 RPI All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low RPI –0.06 –0.09 –0.09 –0.10 0.01 0.10 (–1.08) (–1.28) (–1.35) (–1.8) (0.08) (1.31) 2 –0.10 –0.19 –0.14 –0.12 0.03 0.21** (–1.78) (–2.33) (–2.6) (–2.1) (0.32) (2.36) 3 –0.10 –0.17 –0.08 –0.13 –0.01 0.16** (–1.76) (–2.31) (–1.31) (–2.18) (–0.1) (2.14) High RPI –0.06 –0.12 –0.10 –0.05 0.00 0.13* (–1.02) (–1.69) (–1.52) (–0.66) (0.06) (1.68) High-low 0.00 –0.03 –0.01 0.05 0.00 0.03 (–0.05) (–0.68) (–0.17) (0.97) (–0.02) (0.56) A.4 Active share All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Inactive –0.07 –0.01 –0.05 –0.07 –0.07 –0.06 (–2.91) (–0.35) (–2.01) (–2.24) (–1.43) (–0.85) 2 –0.08 –0.09 –0.14 –0.10 –0.03 0.07 (–2.27) (–1.72) (–2.88) (–2.48) (–0.6) (0.96) 3 –0.08 –0.21 –0.10 –0.12 0.03 0.24*** (–1.55) (–3.02) (–1.59) (–1.89) (0.40) (3.09) Active –0.03 –0.10 –0.18 –0.09 0.06 0.16** (–0.5) (–1.24) (–1.89) (–1.21) (0.78) (2.28) Active-inactive 0.04 –0.08 –0.13 –0.03 0.13 0.22** (0.55) (–1.01) (–1.35) (–0.36) (1.46) (2.25) B. Double sorts on trade-based AFP and fund characteristics B.1 CS All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low CS –0.02 –0.05 0.03 –0.02 –0.04 0.01 (–0.34) (–0.72) (0.44) (–0.3) (–0.59) (0.22) 2 –0.06 –0.05 –0.08 –0.09 –0.04 0.02 (–1.59) (–1.29) (–1.9) (–2.21) (–0.8) (0.55) 3 –0.10 –0.13 –0.12 –0.10 –0.05 0.08* (–3.08) (–3.22) (–3.19) (–2.88) (–1.2) (1.96) High CS –0.13 –0.14 –0.12 –0.14 –0.07 0.08** (–2.31) (–2.24) (–2.21) (–2.74) (–1.13) (2.05) High-low –0.10 –0.09 –0.15 –0.13 –0.03 0.07 (–1.21) (–1.06) (–1.63) (–1.61) (–0.3) (1.30) B.2 Return gap All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low return gap –0.14 –0.09 –0.15 –0.16 –0.14 –0.05 (–3.38) (–1.86) (–3.15) (–4.25) (–2.76) (–1.2) 2 –0.08 –0.12 –0.08 –0.09 –0.04 0.08** (–2.5) (–3.11) (–2.12) (–2.08) (–0.99) (2.12) 3 –0.07 –0.10 –0.05 –0.07 –0.06 0.04 (–1.71) (–2.24) (–1.15) (–1.73) (–1.38) (1.00) High return gap –0.03 –0.10 –0.04 –0.04 0.02 0.12** (–0.78) (–1.74) (–0.88) (–0.75) (0.42) (2.58) High-low 0.11*** 0.00 0.11* 0.12** 0.16*** 0.16*** (2.71) (–0.04) (1.81) (2.45) (3.70) (3.08) B.3 RPI All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low RPI –0.06 –0.07 –0.05 –0.08 –0.05 0.02 (–1.08) (–1.06) (–0.79) (–1.33) (–0.81) (0.47) 2 –0.10 –0.11 –0.10 –0.09 –0.07 0.04 (–1.78) (–1.49) (–1.77) (–1.42) (–0.89) (0.66) 3 –0.10 –0.09 –0.14 –0.07 –0.13 –0.03 (–1.76) (–1.26) (–2.25) (–1.16) (–1.7) (–0.5) High RPI –0.06 –0.13 –0.07 –0.06 0.00 0.13** (–1.02) (–1.64) (–1.06) (–0.95) (0.01) (1.99) High-low 0.00 –0.05 –0.02 0.02 0.05 0.10** (–0.05) (–1.01) (–0.48) (0.46) (1.17) (1.98) B.4 Active share All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Inactive –0.07 –0.06 –0.06 –0.07 –0.01 0.06 (–2.91) (–1.83) (–2) (–2.87) (–0.2) (1.47) 2 –0.08 –0.09 –0.09 –0.14 –0.07 0.02 (–2.27) (–1.68) (–1.55) (–3.24) (–1.69) (0.38) 3 –0.08 –0.07 –0.12 –0.10 –0.05 0.02 (–1.55) (–0.94) (–1.88) (–1.73) (–0.78) (0.40) Active –0.03 –0.11 –0.10 –0.04 –0.03 0.08 (–0.5) (–1.43) (–1.13) (–0.46) (–0.41) (1.62) Active-inactive 0.04 –0.04 –0.04 0.03 –0.02 0.02 (0.55) (–0.6) (–0.45) (0.43) (–0.28) (0.42) A. Double sorts on index-based AFP and fund characteristics AFP All Low 2 3 High High-low A.1. CS All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low CS –0.02 –0.05 –0.05 –0.01 0.00 0.04 (–0.34) (–0.68) (–0.82) (–0.25) (–0.04) (0.83) 2 –0.06 –0.10 –0.05 –0.08 –0.01 0.10** (–1.59) (–2.26) (–1.23) (–2.01) (–0.12) (2.28) 3 –0.10 –0.13 –0.15 –0.08 0.01 0.13*** (–3.08) (–3.12) (–3.83) (–2.2) (0.16) (2.87) High CS –0.13 –0.19 –0.14 –0.13 –0.02 0.18*** (–2.31) (–3.3) (–2.18) (–2.56) (–0.29) (3.05) High-low –0.10 –0.15* –0.09 –0.12 –0.02 0.13* (–1.21) (–1.72) (–0.88) (–1.43) (–0.17) (1.82) A.2 Return gap All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low return gap –0.14 –0.20 –0.14 –0.12 –0.08 0.12** (–3.38) (–4.16) (–2.82) (–2.83) (–1.62) (2.47) 2 –0.08 –0.11 –0.11 –0.10 0.01 0.13** (–2.5) (–2.48) (–3.01) (–2.71) (0.26) (2.27) 3 –0.07 –0.08 –0.09 –0.05 –0.02 0.06 (–1.71) (–1.62) (–2.15) (–1.22) (–0.41) (1.35) High return gap –0.03 –0.09 –0.09 –0.08 0.08 0.17*** (–0.78) (–1.63) (–1.8) (–1.61) (1.58) (3.44) High-low 0.11*** 0.11** 0.06 0.04 0.17*** 0.06 (2.71) (2.22) (1.06) (0.67) (3.56) (1.09) A.3 RPI All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low RPI –0.06 –0.09 –0.09 –0.10 0.01 0.10 (–1.08) (–1.28) (–1.35) (–1.8) (0.08) (1.31) 2 –0.10 –0.19 –0.14 –0.12 0.03 0.21** (–1.78) (–2.33) (–2.6) (–2.1) (0.32) (2.36) 3 –0.10 –0.17 –0.08 –0.13 –0.01 0.16** (–1.76) (–2.31) (–1.31) (–2.18) (–0.1) (2.14) High RPI –0.06 –0.12 –0.10 –0.05 0.00 0.13* (–1.02) (–1.69) (–1.52) (–0.66) (0.06) (1.68) High-low 0.00 –0.03 –0.01 0.05 0.00 0.03 (–0.05) (–0.68) (–0.17) (0.97) (–0.02) (0.56) A.4 Active share All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Inactive –0.07 –0.01 –0.05 –0.07 –0.07 –0.06 (–2.91) (–0.35) (–2.01) (–2.24) (–1.43) (–0.85) 2 –0.08 –0.09 –0.14 –0.10 –0.03 0.07 (–2.27) (–1.72) (–2.88) (–2.48) (–0.6) (0.96) 3 –0.08 –0.21 –0.10 –0.12 0.03 0.24*** (–1.55) (–3.02) (–1.59) (–1.89) (0.40) (3.09) Active –0.03 –0.10 –0.18 –0.09 0.06 0.16** (–0.5) (–1.24) (–1.89) (–1.21) (0.78) (2.28) Active-inactive 0.04 –0.08 –0.13 –0.03 0.13 0.22** (0.55) (–1.01) (–1.35) (–0.36) (1.46) (2.25) B. Double sorts on trade-based AFP and fund characteristics B.1 CS All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low CS –0.02 –0.05 0.03 –0.02 –0.04 0.01 (–0.34) (–0.72) (0.44) (–0.3) (–0.59) (0.22) 2 –0.06 –0.05 –0.08 –0.09 –0.04 0.02 (–1.59) (–1.29) (–1.9) (–2.21) (–0.8) (0.55) 3 –0.10 –0.13 –0.12 –0.10 –0.05 0.08* (–3.08) (–3.22) (–3.19) (–2.88) (–1.2) (1.96) High CS –0.13 –0.14 –0.12 –0.14 –0.07 0.08** (–2.31) (–2.24) (–2.21) (–2.74) (–1.13) (2.05) High-low –0.10 –0.09 –0.15 –0.13 –0.03 0.07 (–1.21) (–1.06) (–1.63) (–1.61) (–0.3) (1.30) B.2 Return gap All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low return gap –0.14 –0.09 –0.15 –0.16 –0.14 –0.05 (–3.38) (–1.86) (–3.15) (–4.25) (–2.76) (–1.2) 2 –0.08 –0.12 –0.08 –0.09 –0.04 0.08** (–2.5) (–3.11) (–2.12) (–2.08) (–0.99) (2.12) 3 –0.07 –0.10 –0.05 –0.07 –0.06 0.04 (–1.71) (–2.24) (–1.15) (–1.73) (–1.38) (1.00) High return gap –0.03 –0.10 –0.04 –0.04 0.02 0.12** (–0.78) (–1.74) (–0.88) (–0.75) (0.42) (2.58) High-low 0.11*** 0.00 0.11* 0.12** 0.16*** 0.16*** (2.71) (–0.04) (1.81) (2.45) (3.70) (3.08) B.3 RPI All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low RPI –0.06 –0.07 –0.05 –0.08 –0.05 0.02 (–1.08) (–1.06) (–0.79) (–1.33) (–0.81) (0.47) 2 –0.10 –0.11 –0.10 –0.09 –0.07 0.04 (–1.78) (–1.49) (–1.77) (–1.42) (–0.89) (0.66) 3 –0.10 –0.09 –0.14 –0.07 –0.13 –0.03 (–1.76) (–1.26) (–2.25) (–1.16) (–1.7) (–0.5) High RPI –0.06 –0.13 –0.07 –0.06 0.00 0.13** (–1.02) (–1.64) (–1.06) (–0.95) (0.01) (1.99) High-low 0.00 –0.05 –0.02 0.02 0.05 0.10** (–0.05) (–1.01) (–0.48) (0.46) (1.17) (1.98) B.4 Active share All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Inactive –0.07 –0.06 –0.06 –0.07 –0.01 0.06 (–2.91) (–1.83) (–2) (–2.87) (–0.2) (1.47) 2 –0.08 –0.09 –0.09 –0.14 –0.07 0.02 (–2.27) (–1.68) (–1.55) (–3.24) (–1.69) (0.38) 3 –0.08 –0.07 –0.12 –0.10 –0.05 0.02 (–1.55) (–0.94) (–1.88) (–1.73) (–0.78) (0.40) Active –0.03 –0.11 –0.10 –0.04 –0.03 0.08 (–0.5) (–1.43) (–1.13) (–0.46) (–0.41) (1.62) Active-inactive 0.04 –0.04 –0.04 0.03 –0.02 0.02 (0.55) (–0.6) (–0.45) (0.43) (–0.28) (0.42) This table presents the performance of portfolios formed using the AFP and fund characteristics that relate to fund skill. We sort funds independently into four groups based on AFP and into four groups based on the following fund characteristics: DGTW characteristic selectivity (CS, subpanel 1), return gap (subpanel 2), reliance on public information (subpanel 3), and active share (subpanel 4). The “All” row and “All” column represent portfolios based on a univariate sort of AFP and an alternative fund characteristic, respectively. We report the Carhart (1997) four-factor $$\alpha $$ as a monthly percentage, based on net returns for each of the 16 portfolios. $$t$$-statistics are shown in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 4 Performance predictive power of AFP: Double sorts A. Double sorts on index-based AFP and fund characteristics AFP All Low 2 3 High High-low A.1. CS All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low CS –0.02 –0.05 –0.05 –0.01 0.00 0.04 (–0.34) (–0.68) (–0.82) (–0.25) (–0.04) (0.83) 2 –0.06 –0.10 –0.05 –0.08 –0.01 0.10** (–1.59) (–2.26) (–1.23) (–2.01) (–0.12) (2.28) 3 –0.10 –0.13 –0.15 –0.08 0.01 0.13*** (–3.08) (–3.12) (–3.83) (–2.2) (0.16) (2.87) High CS –0.13 –0.19 –0.14 –0.13 –0.02 0.18*** (–2.31) (–3.3) (–2.18) (–2.56) (–0.29) (3.05) High-low –0.10 –0.15* –0.09 –0.12 –0.02 0.13* (–1.21) (–1.72) (–0.88) (–1.43) (–0.17) (1.82) A.2 Return gap All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low return gap –0.14 –0.20 –0.14 –0.12 –0.08 0.12** (–3.38) (–4.16) (–2.82) (–2.83) (–1.62) (2.47) 2 –0.08 –0.11 –0.11 –0.10 0.01 0.13** (–2.5) (–2.48) (–3.01) (–2.71) (0.26) (2.27) 3 –0.07 –0.08 –0.09 –0.05 –0.02 0.06 (–1.71) (–1.62) (–2.15) (–1.22) (–0.41) (1.35) High return gap –0.03 –0.09 –0.09 –0.08 0.08 0.17*** (–0.78) (–1.63) (–1.8) (–1.61) (1.58) (3.44) High-low 0.11*** 0.11** 0.06 0.04 0.17*** 0.06 (2.71) (2.22) (1.06) (0.67) (3.56) (1.09) A.3 RPI All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low RPI –0.06 –0.09 –0.09 –0.10 0.01 0.10 (–1.08) (–1.28) (–1.35) (–1.8) (0.08) (1.31) 2 –0.10 –0.19 –0.14 –0.12 0.03 0.21** (–1.78) (–2.33) (–2.6) (–2.1) (0.32) (2.36) 3 –0.10 –0.17 –0.08 –0.13 –0.01 0.16** (–1.76) (–2.31) (–1.31) (–2.18) (–0.1) (2.14) High RPI –0.06 –0.12 –0.10 –0.05 0.00 0.13* (–1.02) (–1.69) (–1.52) (–0.66) (0.06) (1.68) High-low 0.00 –0.03 –0.01 0.05 0.00 0.03 (–0.05) (–0.68) (–0.17) (0.97) (–0.02) (0.56) A.4 Active share All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Inactive –0.07 –0.01 –0.05 –0.07 –0.07 –0.06 (–2.91) (–0.35) (–2.01) (–2.24) (–1.43) (–0.85) 2 –0.08 –0.09 –0.14 –0.10 –0.03 0.07 (–2.27) (–1.72) (–2.88) (–2.48) (–0.6) (0.96) 3 –0.08 –0.21 –0.10 –0.12 0.03 0.24*** (–1.55) (–3.02) (–1.59) (–1.89) (0.40) (3.09) Active –0.03 –0.10 –0.18 –0.09 0.06 0.16** (–0.5) (–1.24) (–1.89) (–1.21) (0.78) (2.28) Active-inactive 0.04 –0.08 –0.13 –0.03 0.13 0.22** (0.55) (–1.01) (–1.35) (–0.36) (1.46) (2.25) B. Double sorts on trade-based AFP and fund characteristics B.1 CS All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low CS –0.02 –0.05 0.03 –0.02 –0.04 0.01 (–0.34) (–0.72) (0.44) (–0.3) (–0.59) (0.22) 2 –0.06 –0.05 –0.08 –0.09 –0.04 0.02 (–1.59) (–1.29) (–1.9) (–2.21) (–0.8) (0.55) 3 –0.10 –0.13 –0.12 –0.10 –0.05 0.08* (–3.08) (–3.22) (–3.19) (–2.88) (–1.2) (1.96) High CS –0.13 –0.14 –0.12 –0.14 –0.07 0.08** (–2.31) (–2.24) (–2.21) (–2.74) (–1.13) (2.05) High-low –0.10 –0.09 –0.15 –0.13 –0.03 0.07 (–1.21) (–1.06) (–1.63) (–1.61) (–0.3) (1.30) B.2 Return gap All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low return gap –0.14 –0.09 –0.15 –0.16 –0.14 –0.05 (–3.38) (–1.86) (–3.15) (–4.25) (–2.76) (–1.2) 2 –0.08 –0.12 –0.08 –0.09 –0.04 0.08** (–2.5) (–3.11) (–2.12) (–2.08) (–0.99) (2.12) 3 –0.07 –0.10 –0.05 –0.07 –0.06 0.04 (–1.71) (–2.24) (–1.15) (–1.73) (–1.38) (1.00) High return gap –0.03 –0.10 –0.04 –0.04 0.02 0.12** (–0.78) (–1.74) (–0.88) (–0.75) (0.42) (2.58) High-low 0.11*** 0.00 0.11* 0.12** 0.16*** 0.16*** (2.71) (–0.04) (1.81) (2.45) (3.70) (3.08) B.3 RPI All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low RPI –0.06 –0.07 –0.05 –0.08 –0.05 0.02 (–1.08) (–1.06) (–0.79) (–1.33) (–0.81) (0.47) 2 –0.10 –0.11 –0.10 –0.09 –0.07 0.04 (–1.78) (–1.49) (–1.77) (–1.42) (–0.89) (0.66) 3 –0.10 –0.09 –0.14 –0.07 –0.13 –0.03 (–1.76) (–1.26) (–2.25) (–1.16) (–1.7) (–0.5) High RPI –0.06 –0.13 –0.07 –0.06 0.00 0.13** (–1.02) (–1.64) (–1.06) (–0.95) (0.01) (1.99) High-low 0.00 –0.05 –0.02 0.02 0.05 0.10** (–0.05) (–1.01) (–0.48) (0.46) (1.17) (1.98) B.4 Active share All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Inactive –0.07 –0.06 –0.06 –0.07 –0.01 0.06 (–2.91) (–1.83) (–2) (–2.87) (–0.2) (1.47) 2 –0.08 –0.09 –0.09 –0.14 –0.07 0.02 (–2.27) (–1.68) (–1.55) (–3.24) (–1.69) (0.38) 3 –0.08 –0.07 –0.12 –0.10 –0.05 0.02 (–1.55) (–0.94) (–1.88) (–1.73) (–0.78) (0.40) Active –0.03 –0.11 –0.10 –0.04 –0.03 0.08 (–0.5) (–1.43) (–1.13) (–0.46) (–0.41) (1.62) Active-inactive 0.04 –0.04 –0.04 0.03 –0.02 0.02 (0.55) (–0.6) (–0.45) (0.43) (–0.28) (0.42) A. Double sorts on index-based AFP and fund characteristics AFP All Low 2 3 High High-low A.1. CS All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low CS –0.02 –0.05 –0.05 –0.01 0.00 0.04 (–0.34) (–0.68) (–0.82) (–0.25) (–0.04) (0.83) 2 –0.06 –0.10 –0.05 –0.08 –0.01 0.10** (–1.59) (–2.26) (–1.23) (–2.01) (–0.12) (2.28) 3 –0.10 –0.13 –0.15 –0.08 0.01 0.13*** (–3.08) (–3.12) (–3.83) (–2.2) (0.16) (2.87) High CS –0.13 –0.19 –0.14 –0.13 –0.02 0.18*** (–2.31) (–3.3) (–2.18) (–2.56) (–0.29) (3.05) High-low –0.10 –0.15* –0.09 –0.12 –0.02 0.13* (–1.21) (–1.72) (–0.88) (–1.43) (–0.17) (1.82) A.2 Return gap All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low return gap –0.14 –0.20 –0.14 –0.12 –0.08 0.12** (–3.38) (–4.16) (–2.82) (–2.83) (–1.62) (2.47) 2 –0.08 –0.11 –0.11 –0.10 0.01 0.13** (–2.5) (–2.48) (–3.01) (–2.71) (0.26) (2.27) 3 –0.07 –0.08 –0.09 –0.05 –0.02 0.06 (–1.71) (–1.62) (–2.15) (–1.22) (–0.41) (1.35) High return gap –0.03 –0.09 –0.09 –0.08 0.08 0.17*** (–0.78) (–1.63) (–1.8) (–1.61) (1.58) (3.44) High-low 0.11*** 0.11** 0.06 0.04 0.17*** 0.06 (2.71) (2.22) (1.06) (0.67) (3.56) (1.09) A.3 RPI All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Low RPI –0.06 –0.09 –0.09 –0.10 0.01 0.10 (–1.08) (–1.28) (–1.35) (–1.8) (0.08) (1.31) 2 –0.10 –0.19 –0.14 –0.12 0.03 0.21** (–1.78) (–2.33) (–2.6) (–2.1) (0.32) (2.36) 3 –0.10 –0.17 –0.08 –0.13 –0.01 0.16** (–1.76) (–2.31) (–1.31) (–2.18) (–0.1) (2.14) High RPI –0.06 –0.12 –0.10 –0.05 0.00 0.13* (–1.02) (–1.69) (–1.52) (–0.66) (0.06) (1.68) High-low 0.00 –0.03 –0.01 0.05 0.00 0.03 (–0.05) (–0.68) (–0.17) (0.97) (–0.02) (0.56) A.4 Active share All –0.12 –0.10 –0.08 0.00 0.12*** (–2.89) (–2.82) (–2.47) (–0.02) (2.90) Inactive –0.07 –0.01 –0.05 –0.07 –0.07 –0.06 (–2.91) (–0.35) (–2.01) (–2.24) (–1.43) (–0.85) 2 –0.08 –0.09 –0.14 –0.10 –0.03 0.07 (–2.27) (–1.72) (–2.88) (–2.48) (–0.6) (0.96) 3 –0.08 –0.21 –0.10 –0.12 0.03 0.24*** (–1.55) (–3.02) (–1.59) (–1.89) (0.40) (3.09) Active –0.03 –0.10 –0.18 –0.09 0.06 0.16** (–0.5) (–1.24) (–1.89) (–1.21) (0.78) (2.28) Active-inactive 0.04 –0.08 –0.13 –0.03 0.13 0.22** (0.55) (–1.01) (–1.35) (–0.36) (1.46) (2.25) B. Double sorts on trade-based AFP and fund characteristics B.1 CS All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low CS –0.02 –0.05 0.03 –0.02 –0.04 0.01 (–0.34) (–0.72) (0.44) (–0.3) (–0.59) (0.22) 2 –0.06 –0.05 –0.08 –0.09 –0.04 0.02 (–1.59) (–1.29) (–1.9) (–2.21) (–0.8) (0.55) 3 –0.10 –0.13 –0.12 –0.10 –0.05 0.08* (–3.08) (–3.22) (–3.19) (–2.88) (–1.2) (1.96) High CS –0.13 –0.14 –0.12 –0.14 –0.07 0.08** (–2.31) (–2.24) (–2.21) (–2.74) (–1.13) (2.05) High-low –0.10 –0.09 –0.15 –0.13 –0.03 0.07 (–1.21) (–1.06) (–1.63) (–1.61) (–0.3) (1.30) B.2 Return gap All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low return gap –0.14 –0.09 –0.15 –0.16 –0.14 –0.05 (–3.38) (–1.86) (–3.15) (–4.25) (–2.76) (–1.2) 2 –0.08 –0.12 –0.08 –0.09 –0.04 0.08** (–2.5) (–3.11) (–2.12) (–2.08) (–0.99) (2.12) 3 –0.07 –0.10 –0.05 –0.07 –0.06 0.04 (–1.71) (–2.24) (–1.15) (–1.73) (–1.38) (1.00) High return gap –0.03 –0.10 –0.04 –0.04 0.02 0.12** (–0.78) (–1.74) (–0.88) (–0.75) (0.42) (2.58) High-low 0.11*** 0.00 0.11* 0.12** 0.16*** 0.16*** (2.71) (–0.04) (1.81) (2.45) (3.70) (3.08) B.3 RPI All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Low RPI –0.06 –0.07 –0.05 –0.08 –0.05 0.02 (–1.08) (–1.06) (–0.79) (–1.33) (–0.81) (0.47) 2 –0.10 –0.11 –0.10 –0.09 –0.07 0.04 (–1.78) (–1.49) (–1.77) (–1.42) (–0.89) (0.66) 3 –0.10 –0.09 –0.14 –0.07 –0.13 –0.03 (–1.76) (–1.26) (–2.25) (–1.16) (–1.7) (–0.5) High RPI –0.06 –0.13 –0.07 –0.06 0.00 0.13** (–1.02) (–1.64) (–1.06) (–0.95) (0.01) (1.99) High-low 0.00 –0.05 –0.02 0.02 0.05 0.10** (–0.05) (–1.01) (–0.48) (0.46) (1.17) (1.98) B.4 Active share All –0.10 –0.07 –0.08 –0.05 0.05** (–2.57) (–2.05) (–2.51) (–1.34) (2.01) Inactive –0.07 –0.06 –0.06 –0.07 –0.01 0.06 (–2.91) (–1.83) (–2) (–2.87) (–0.2) (1.47) 2 –0.08 –0.09 –0.09 –0.14 –0.07 0.02 (–2.27) (–1.68) (–1.55) (–3.24) (–1.69) (0.38) 3 –0.08 –0.07 –0.12 –0.10 –0.05 0.02 (–1.55) (–0.94) (–1.88) (–1.73) (–0.78) (0.40) Active –0.03 –0.11 –0.10 –0.04 –0.03 0.08 (–0.5) (–1.43) (–1.13) (–0.46) (–0.41) (1.62) Active-inactive 0.04 –0.04 –0.04 0.03 –0.02 0.02 (0.55) (–0.6) (–0.45) (0.43) (–0.28) (0.42) This table presents the performance of portfolios formed using the AFP and fund characteristics that relate to fund skill. We sort funds independently into four groups based on AFP and into four groups based on the following fund characteristics: DGTW characteristic selectivity (CS, subpanel 1), return gap (subpanel 2), reliance on public information (subpanel 3), and active share (subpanel 4). The “All” row and “All” column represent portfolios based on a univariate sort of AFP and an alternative fund characteristic, respectively. We report the Carhart (1997) four-factor $$\alpha $$ as a monthly percentage, based on net returns for each of the 16 portfolios. $$t$$-statistics are shown in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. In these double sorts, the return gap stands out as a strong predictor of future fund performance. For instance, in Table 4, panel A.2, funds with a high return gap outperform their peers with a low return gap for two quartiles of funds sorted by AFP. Despite this strong performance, AFP predicts future fund performance among three quartiles of funds sorted by return gap, even though AFP is limited to quarterly reporting and does not capture intraquarter trades, whereas return gaps do. Therefore, it is useful to combine the information signals from both the return gap and AFP for fund manager selection. As Table 4, panel A.2, shows, mutual funds in the top quartiles of the AFP and return gap yield a net four-factor alpha of 0.96% per year, whereas funds in the bottom quartiles of the AFP and return gap generate a net four-factor alpha of $$-$$2.40% per year. Mutual funds investors who switch from the unskilled funds to the skilled funds in our sample period would have achieved an increase in the four-factor alpha of 3.36% per year. 3.3 Predictive panel regressions We also use multivariate regressions to examine the incremental predictive power of AFP for mutual fund performance after we control for other performance predictors and fund characteristics. Our measure of mutual fund performance is the four-factor alpha measured as the difference between the realized fund return in excess of the expected return from a four-factor model, which includes the market, size, value, and momentum factors. To estimate the factor loadings, we use rolling-window time-series regressions of fund returns in the previous 3 years. In addition to CS, return gap, RPI, and active share, the fund characteristics we consider include fund size, measured as the natural log of fund assets under management, the natural log of fund age in years, the expense ratio, fund turnover, percentage flows in the past quarter, and fund alpha estimated in the past 3 years. Table 5 presents the results from the predictive panel regressions. We start with a univariate regression with each of the skill measure, and then run a multivariate regression controlling for fund characteristics. Finally, we specify a regression which includes all the measures of skill and fund characteristics. To control for aggregate movements in fund returns over time, we include time fixed effects in all the regressions. Furthermore, because the residuals might correlate within funds, we cluster standard errors by fund (Petersen 2009).8 Table 5 AFP and mutual fund returns: Predictive panel regressions A. Index-based AFP (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) AFP_Index 2.955*** 2.467*** 3.063*** (4.38) (3.58) (3.46) CS 0.420** –0.186 –0.185 (1.98) (–0.89) (–0.73) Return gap 10.68*** 7.077** 3.661 (3.99) (2.52) (0.94) RPI –0.230* –0.245** –0.306** (–1.95) (–2.04) (–2.42) Active share 0.141*** 0.185*** 0.174*** (3.75) (4.93) (4.45) log(TNA) –0.00759*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00894** (–2.93) (–2.92) (–2.61) (–3.45) (–2.05) (–2.45) log(Age) –0.00376 –0.00364 –0.00448 0.00129 –0.00285 0.00143 (–0.65) (–0.62) (–0.76) (0.17) (–0.42) (0.19) Expense –8.630*** –8.589*** –8.763*** –7.488*** –8.835*** –8.815*** (–7.23) (–7.17) (–7.17) (–4.81) (–6.33) (–5.70) Turnover –0.0206*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0171** (–3.16) (–3.15) (–3.03) (–2.18) (–2.83) (–2.16) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.73) (2.82) (2.63) (2.13) (2.72) (2.47) Past alpha 6.020*** 6.225*** 6.432*** 2.468 3.952** 3.586** (4.08) (4.14) (4.35) (1.42) (2.27) (1.96) Adj. R$$^{\mathrm{2}}$$ 0.0928 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0990 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 B. Trade-based AFP AFP_Trade 1.472** 1.242* 2.067** (2.33) (1.90) (2.40) CS 0.420** –0.186 –0.177 (1.98) (–0.89) (–0.70) Return gap 10.68*** 7.077** 3.696 (3.99) (2.52) (0.95) RPI –0.230* –0.245** –0.308** (–1.95) (–2.04) (–2.43) Active share 0.141*** 0.185*** 0.173*** (3.75) (4.93) (4.44) log(TNA) –0.00748*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00880** (–2.88) (–2.92) (–2.61) (–3.45) (–2.05) (–2.41) log(Age) –0.00374 –0.00364 –0.00448 0.00129 –0.00285 0.00175 (–0.64) (–0.62) (–0.76) (0.17) (–0.42) (0.24) Expense –8.615*** –8.589*** –8.763*** –7.488*** –8.835*** –8.795*** (–7.20) (–7.17) (–7.17) (–4.81) (–6.33) (–5.68) Turnover –0.0207*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0173** (–3.17) (–3.15) (–3.03) (–2.18) (–2.83) (–2.19) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.72) (2.82) (2.63) (2.13) (2.72) (2.46) Past alpha 6.001*** 6.225*** 6.432*** 2.468 3.952** 3.538* (4.05) (4.14) (4.35) (1.42) (2.27) (1.94) Adj. R$$^{\mathrm{2}}$$ 0.0927 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0989 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 A. Index-based AFP (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) AFP_Index 2.955*** 2.467*** 3.063*** (4.38) (3.58) (3.46) CS 0.420** –0.186 –0.185 (1.98) (–0.89) (–0.73) Return gap 10.68*** 7.077** 3.661 (3.99) (2.52) (0.94) RPI –0.230* –0.245** –0.306** (–1.95) (–2.04) (–2.42) Active share 0.141*** 0.185*** 0.174*** (3.75) (4.93) (4.45) log(TNA) –0.00759*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00894** (–2.93) (–2.92) (–2.61) (–3.45) (–2.05) (–2.45) log(Age) –0.00376 –0.00364 –0.00448 0.00129 –0.00285 0.00143 (–0.65) (–0.62) (–0.76) (0.17) (–0.42) (0.19) Expense –8.630*** –8.589*** –8.763*** –7.488*** –8.835*** –8.815*** (–7.23) (–7.17) (–7.17) (–4.81) (–6.33) (–5.70) Turnover –0.0206*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0171** (–3.16) (–3.15) (–3.03) (–2.18) (–2.83) (–2.16) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.73) (2.82) (2.63) (2.13) (2.72) (2.47) Past alpha 6.020*** 6.225*** 6.432*** 2.468 3.952** 3.586** (4.08) (4.14) (4.35) (1.42) (2.27) (1.96) Adj. R$$^{\mathrm{2}}$$ 0.0928 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0990 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 B. Trade-based AFP AFP_Trade 1.472** 1.242* 2.067** (2.33) (1.90) (2.40) CS 0.420** –0.186 –0.177 (1.98) (–0.89) (–0.70) Return gap 10.68*** 7.077** 3.696 (3.99) (2.52) (0.95) RPI –0.230* –0.245** –0.308** (–1.95) (–2.04) (–2.43) Active share 0.141*** 0.185*** 0.173*** (3.75) (4.93) (4.44) log(TNA) –0.00748*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00880** (–2.88) (–2.92) (–2.61) (–3.45) (–2.05) (–2.41) log(Age) –0.00374 –0.00364 –0.00448 0.00129 –0.00285 0.00175 (–0.64) (–0.62) (–0.76) (0.17) (–0.42) (0.24) Expense –8.615*** –8.589*** –8.763*** –7.488*** –8.835*** –8.795*** (–7.20) (–7.17) (–7.17) (–4.81) (–6.33) (–5.68) Turnover –0.0207*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0173** (–3.17) (–3.15) (–3.03) (–2.18) (–2.83) (–2.19) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.72) (2.82) (2.63) (2.13) (2.72) (2.46) Past alpha 6.001*** 6.225*** 6.432*** 2.468 3.952** 3.538* (4.05) (4.14) (4.35) (1.42) (2.27) (1.94) Adj. R$$^{\mathrm{2}}$$ 0.0927 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0989 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 This table presents coefficient estimates from predictive panel regressions estimating the association between the AFP and future fund performance. Like in Equation (3), AFP is the sum of the product of active portfolio weights (changes in portfolio weights) and subsequent 3-day abnormal returns surrounding earnings announcements for the index-based (trade-based) measure. We also include other holdings-based measures of skill such as CS, return gap, RPI, and active share. All measures of skill are winsorized at the top and the bottom 5%. We measure future mutual fund performance using Carhart’s (1997) four-factor alpha (as a percentage), where fund betas are estimated using rolling-window regressions in the past 3 years. The panel regressions control for fund size, fund age, expense ratio, fund turnover, fund percentage flow in the past quarter, and fund alpha in the past 3 years. The regressions include time fixed effects, and the standard errors are clustered by fund. $$t$$-statistics are presented in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 5 AFP and mutual fund returns: Predictive panel regressions A. Index-based AFP (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) AFP_Index 2.955*** 2.467*** 3.063*** (4.38) (3.58) (3.46) CS 0.420** –0.186 –0.185 (1.98) (–0.89) (–0.73) Return gap 10.68*** 7.077** 3.661 (3.99) (2.52) (0.94) RPI –0.230* –0.245** –0.306** (–1.95) (–2.04) (–2.42) Active share 0.141*** 0.185*** 0.174*** (3.75) (4.93) (4.45) log(TNA) –0.00759*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00894** (–2.93) (–2.92) (–2.61) (–3.45) (–2.05) (–2.45) log(Age) –0.00376 –0.00364 –0.00448 0.00129 –0.00285 0.00143 (–0.65) (–0.62) (–0.76) (0.17) (–0.42) (0.19) Expense –8.630*** –8.589*** –8.763*** –7.488*** –8.835*** –8.815*** (–7.23) (–7.17) (–7.17) (–4.81) (–6.33) (–5.70) Turnover –0.0206*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0171** (–3.16) (–3.15) (–3.03) (–2.18) (–2.83) (–2.16) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.73) (2.82) (2.63) (2.13) (2.72) (2.47) Past alpha 6.020*** 6.225*** 6.432*** 2.468 3.952** 3.586** (4.08) (4.14) (4.35) (1.42) (2.27) (1.96) Adj. R$$^{\mathrm{2}}$$ 0.0928 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0990 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 B. Trade-based AFP AFP_Trade 1.472** 1.242* 2.067** (2.33) (1.90) (2.40) CS 0.420** –0.186 –0.177 (1.98) (–0.89) (–0.70) Return gap 10.68*** 7.077** 3.696 (3.99) (2.52) (0.95) RPI –0.230* –0.245** –0.308** (–1.95) (–2.04) (–2.43) Active share 0.141*** 0.185*** 0.173*** (3.75) (4.93) (4.44) log(TNA) –0.00748*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00880** (–2.88) (–2.92) (–2.61) (–3.45) (–2.05) (–2.41) log(Age) –0.00374 –0.00364 –0.00448 0.00129 –0.00285 0.00175 (–0.64) (–0.62) (–0.76) (0.17) (–0.42) (0.24) Expense –8.615*** –8.589*** –8.763*** –7.488*** –8.835*** –8.795*** (–7.20) (–7.17) (–7.17) (–4.81) (–6.33) (–5.68) Turnover –0.0207*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0173** (–3.17) (–3.15) (–3.03) (–2.18) (–2.83) (–2.19) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.72) (2.82) (2.63) (2.13) (2.72) (2.46) Past alpha 6.001*** 6.225*** 6.432*** 2.468 3.952** 3.538* (4.05) (4.14) (4.35) (1.42) (2.27) (1.94) Adj. R$$^{\mathrm{2}}$$ 0.0927 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0989 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 A. Index-based AFP (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) AFP_Index 2.955*** 2.467*** 3.063*** (4.38) (3.58) (3.46) CS 0.420** –0.186 –0.185 (1.98) (–0.89) (–0.73) Return gap 10.68*** 7.077** 3.661 (3.99) (2.52) (0.94) RPI –0.230* –0.245** –0.306** (–1.95) (–2.04) (–2.42) Active share 0.141*** 0.185*** 0.174*** (3.75) (4.93) (4.45) log(TNA) –0.00759*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00894** (–2.93) (–2.92) (–2.61) (–3.45) (–2.05) (–2.45) log(Age) –0.00376 –0.00364 –0.00448 0.00129 –0.00285 0.00143 (–0.65) (–0.62) (–0.76) (0.17) (–0.42) (0.19) Expense –8.630*** –8.589*** –8.763*** –7.488*** –8.835*** –8.815*** (–7.23) (–7.17) (–7.17) (–4.81) (–6.33) (–5.70) Turnover –0.0206*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0171** (–3.16) (–3.15) (–3.03) (–2.18) (–2.83) (–2.16) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.73) (2.82) (2.63) (2.13) (2.72) (2.47) Past alpha 6.020*** 6.225*** 6.432*** 2.468 3.952** 3.586** (4.08) (4.14) (4.35) (1.42) (2.27) (1.96) Adj. R$$^{\mathrm{2}}$$ 0.0928 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0990 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 B. Trade-based AFP AFP_Trade 1.472** 1.242* 2.067** (2.33) (1.90) (2.40) CS 0.420** –0.186 –0.177 (1.98) (–0.89) (–0.70) Return gap 10.68*** 7.077** 3.696 (3.99) (2.52) (0.95) RPI –0.230* –0.245** –0.308** (–1.95) (–2.04) (–2.43) Active share 0.141*** 0.185*** 0.173*** (3.75) (4.93) (4.44) log(TNA) –0.00748*** –0.00758*** –0.00680*** –0.0123*** –0.00700** –0.00880** (–2.88) (–2.92) (–2.61) (–3.45) (–2.05) (–2.41) log(Age) –0.00374 –0.00364 –0.00448 0.00129 –0.00285 0.00175 (–0.64) (–0.62) (–0.76) (0.17) (–0.42) (0.24) Expense –8.615*** –8.589*** –8.763*** –7.488*** –8.835*** –8.795*** (–7.20) (–7.17) (–7.17) (–4.81) (–6.33) (–5.68) Turnover –0.0207*** –0.0206*** –0.0197*** –0.0155** –0.0248*** –0.0173** (–3.17) (–3.15) (–3.03) (–2.18) (–2.83) (–2.19) Past flow 0.107*** 0.110*** 0.103*** 0.107** 0.130*** 0.124** (2.72) (2.82) (2.63) (2.13) (2.72) (2.46) Past alpha 6.001*** 6.225*** 6.432*** 2.468 3.952** 3.538* (4.05) (4.14) (4.35) (1.42) (2.27) (1.94) Adj. R$$^{\mathrm{2}}$$ 0.0927 0.0911 0.0927 0.0910 0.0925 0.0912 0.101 0.0984 0.0962 0.0944 0.0989 N 239,966 222,398 239,919 222,354 230,597 214,372 150,351 138,771 167,343 153,011 123,731 This table presents coefficient estimates from predictive panel regressions estimating the association between the AFP and future fund performance. Like in Equation (3), AFP is the sum of the product of active portfolio weights (changes in portfolio weights) and subsequent 3-day abnormal returns surrounding earnings announcements for the index-based (trade-based) measure. We also include other holdings-based measures of skill such as CS, return gap, RPI, and active share. All measures of skill are winsorized at the top and the bottom 5%. We measure future mutual fund performance using Carhart’s (1997) four-factor alpha (as a percentage), where fund betas are estimated using rolling-window regressions in the past 3 years. The panel regressions control for fund size, fund age, expense ratio, fund turnover, fund percentage flow in the past quarter, and fund alpha in the past 3 years. The regressions include time fixed effects, and the standard errors are clustered by fund. $$t$$-statistics are presented in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The results show that AFP reliably predicts future fund performance after we control for other measures of fund skill and fund characteristics. For instance, in the case of the index-based AFP using the four-factor net alpha, the slope coefficient for AFP is 2.18, with a $$t$$-statistic of 2.65. When we measure fund performance using the gross alpha, we obtain qualitatively and quantitatively similar results. The fund characteristics included in the regression relate to future fund performance in ways consistent with the previous findings. For example, fund size is negatively related to future performance, consistent with large funds underperforming relative to small funds, as Chen et al. (2004) document. Past flows have a positive relation with future performance, consistent with the smart-money effect documented by Gruber (1996) and Zheng (1999). Over our sample period, we observe a certain degree of performance persistence among active funds. All the four alternative measures of skill are significantly related to future fund alpha in the univariate regressions. RPI, active share, and return gap are also significantly related to future alphas when we include common controls for fund characteristics. CS measure loses its performance forecast power when we control for fund characteristics, although, as we discussed earlier, it is not the goal of Daniel et al. (1997) to predict fund performance. When we include all performance measures and fund characteristics in the same regression, AFP, RPI and active share are again significantly related to future performance. Return gap loses its predictive power. This could be driven by the fact that the sample size is reduced substantially when we require a fund to have no missing observations on all the skill measures. It is also possible that part of predictive power of return gap is captured by AFP in the regression setting. Overall, the predictive power of AFP seems robust as AFP is significantly related to future fund alpha in all three regression specifications as well as in portfolio sorts. Why do some performance measures show significant performance predictive power in panel regressions, but exhibit no such power in portfolio sorts, and vice versa? We believe that these findings illustrate a key difference between panel regressions and portfolio sorts. In a portfolio sort, the spread in returns captured by the predicting variables is driven by cross-fund variation. In a panel regression, the variations in alpha come from both cross-fund and time-series dimensions. Therefore, some performance measures are useful to fund investors in telling the difference in a given fund’s performance over time, while other measures are helpful to rank mutual funds at a given point in time. 4. Understanding AFP 4.1 Drivers of AFP To better understand the drivers of AFP, we resort to a stock-level analysis of how asymmetric information affects the return predictability of active weights and trades. An important skill in active management is the manager’s ability to invest in mispriced securities with gaps between the fundamental value and market price. This ability manifests itself in the form of a positive return when the market price converges toward the fundamental value. But where do opportunities lie for skilled managers to identify mispriced securities? We hypothesize that skilled fund managers are more likely to exploit mispricing among firms with more information asymmetry. Our proxies for information asymmetry include analyst coverage, idiosyncratic volatilities, and analyst disagreement. An important role of sell-side analysts in equity markets is to provide market participants with timely and accurate earnings forecasts, which may reduce the information advantages of a particular group of investors. Therefore, firms with less analyst coverage are likely to have more information asymmetry. Another commonly used measure is idiosyncratic volatilities, which could capture the amount of private information for a given firm: firms with higher idiosyncratic volatilities are likely to have more information asymmetry. A similar idea could apply to the disagreement among financial analysts. If AFP is an effective measure of fund investment skill, we expect its key founding components, the active weights or fund trades, to have more forecasting power among stocks with lower analyst coverage, higher idiosyncratic volatilities, and higher dispersion in analyst earnings forecasts. Table 6 shows the results, using the interaction between active weights (or fund trades) and proxies for information asymmetry to predict future abnormal stock returns around earnings announcements. The unit of observations for Table 6 is at the fund-stock-quarter level. For statistical inference, we follow a two-step procedure, like in Grinblatt et al. (2011). First, in each quarter we perform cross-sectional regressions at the fund-stock level. Then, similar to Fama and Macbeth (1973) with Newey and West (1987) adjustments, we conduct statistical inference based on the time-series of the coefficients. For both active weights and fund trades, we observe a negative coefficient for the interaction with analyst coverage, a positive coefficient for the interaction with idiosyncratic volatilities, and a positive coefficient for the interaction with analyst disagreement. These results suggest that active fund managers tend to exhibit more skill when investing in stocks with higher information asymmetry. This ability to estimate the fundamental values of firms with greater information asymmetry contributes to the forecasting power of AFP. Table 6 Understanding AFP: Asymmetric information Index-based Trade-based (1) (2) (3) (4) (5) (6) Active weights 8.403*** –0.094 4.262*** 5.187*** –0.293 2.337*** (6.38) (–0.05) (5.46) (6.07) (–0.38) (5.50) Active weights –1.954*** –1.194*** $$\times$$ Coverage (–2.79) (–3.28) Coverage 0.137*** 0.141*** (4.74) (4.89) Active weights $$\times$$ IV 3.219*** 1.475*** (3.49) (2.74) IV –0.087*** –0.079*** (–2.74) (–2.69) Active weights 4.365** 3.735*** $$\times$$ Disagreement (2.07) (2.69) Disagreement –0.101* –0.097* (–1.75) (–1.72) Size –0.024 0.000 0.012 –0.030* –0.002 0.009 (–1.40) (0.02) (0.69) (–1.75) (–0.17) (0.56) Book-to-market –0.099*** –0.104*** –0.097** –0.098*** –0.098*** –0.099** (–2.96) (–3.19) (–2.00) (–2.97) (–3.06) (–2.06) Momentum 0.200*** 0.188*** 0.200** 0.220*** 0.203*** 0.221*** (2.90) (2.83) (2.40) (3.30) (3.15) (2.75) Average adj. R$$^{\mathrm{2}}$$ 0.006 0.006 0.007 0.006 0.006 0.007 Index-based Trade-based (1) (2) (3) (4) (5) (6) Active weights 8.403*** –0.094 4.262*** 5.187*** –0.293 2.337*** (6.38) (–0.05) (5.46) (6.07) (–0.38) (5.50) Active weights –1.954*** –1.194*** $$\times$$ Coverage (–2.79) (–3.28) Coverage 0.137*** 0.141*** (4.74) (4.89) Active weights $$\times$$ IV 3.219*** 1.475*** (3.49) (2.74) IV –0.087*** –0.079*** (–2.74) (–2.69) Active weights 4.365** 3.735*** $$\times$$ Disagreement (2.07) (2.69) Disagreement –0.101* –0.097* (–1.75) (–1.72) Size –0.024 0.000 0.012 –0.030* –0.002 0.009 (–1.40) (0.02) (0.69) (–1.75) (–0.17) (0.56) Book-to-market –0.099*** –0.104*** –0.097** –0.098*** –0.098*** –0.099** (–2.96) (–3.19) (–2.00) (–2.97) (–3.06) (–2.06) Momentum 0.200*** 0.188*** 0.200** 0.220*** 0.203*** 0.221*** (2.90) (2.83) (2.40) (3.30) (3.15) (2.75) Average adj. R$$^{\mathrm{2}}$$ 0.006 0.006 0.007 0.006 0.006 0.007 This table presents the regression results on how information asymmetry influences the predictive power of Active Weights for subsequent earnings announcement returns. We measure earnings announcement returns using the 3-day abnormal returns in percentage terms surrounding earnings announcements. We measure active weights using two approaches: an index-based measure is the weight of a stock in a fund’s portfolio in excess of the stock’s weight in the fund’s benchmark index and a trade-based measure is the change in the weight of a stock in a fund’s portfolio in the past year, controlling for stock price movements. Analyst coverage is the natural log of one plus the number of analysts covering the firm. Idiosyncratic volatility (IV) of daily stock returns is the standard deviation of residual returns from the Fama-French model in the past quarter. Disagreement (analyst forecast dispersion) is the standard deviation of analyst earnings forecasts, times 10. The control variables include firm size (the natural log of market cap), the book-to-market ratio, and momentum (past 1-year return skipping the most recent month). We performed these regressions using stock-fund pooled data for each quarter from 1984Q1 to 2014Q2. The statistical inference is based on the Fama and Macbeth (1973) procedure with Newey and West (1987) adjustments. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 6 Understanding AFP: Asymmetric information Index-based Trade-based (1) (2) (3) (4) (5) (6) Active weights 8.403*** –0.094 4.262*** 5.187*** –0.293 2.337*** (6.38) (–0.05) (5.46) (6.07) (–0.38) (5.50) Active weights –1.954*** –1.194*** $$\times$$ Coverage (–2.79) (–3.28) Coverage 0.137*** 0.141*** (4.74) (4.89) Active weights $$\times$$ IV 3.219*** 1.475*** (3.49) (2.74) IV –0.087*** –0.079*** (–2.74) (–2.69) Active weights 4.365** 3.735*** $$\times$$ Disagreement (2.07) (2.69) Disagreement –0.101* –0.097* (–1.75) (–1.72) Size –0.024 0.000 0.012 –0.030* –0.002 0.009 (–1.40) (0.02) (0.69) (–1.75) (–0.17) (0.56) Book-to-market –0.099*** –0.104*** –0.097** –0.098*** –0.098*** –0.099** (–2.96) (–3.19) (–2.00) (–2.97) (–3.06) (–2.06) Momentum 0.200*** 0.188*** 0.200** 0.220*** 0.203*** 0.221*** (2.90) (2.83) (2.40) (3.30) (3.15) (2.75) Average adj. R$$^{\mathrm{2}}$$ 0.006 0.006 0.007 0.006 0.006 0.007 Index-based Trade-based (1) (2) (3) (4) (5) (6) Active weights 8.403*** –0.094 4.262*** 5.187*** –0.293 2.337*** (6.38) (–0.05) (5.46) (6.07) (–0.38) (5.50) Active weights –1.954*** –1.194*** $$\times$$ Coverage (–2.79) (–3.28) Coverage 0.137*** 0.141*** (4.74) (4.89) Active weights $$\times$$ IV 3.219*** 1.475*** (3.49) (2.74) IV –0.087*** –0.079*** (–2.74) (–2.69) Active weights 4.365** 3.735*** $$\times$$ Disagreement (2.07) (2.69) Disagreement –0.101* –0.097* (–1.75) (–1.72) Size –0.024 0.000 0.012 –0.030* –0.002 0.009 (–1.40) (0.02) (0.69) (–1.75) (–0.17) (0.56) Book-to-market –0.099*** –0.104*** –0.097** –0.098*** –0.098*** –0.099** (–2.96) (–3.19) (–2.00) (–2.97) (–3.06) (–2.06) Momentum 0.200*** 0.188*** 0.200** 0.220*** 0.203*** 0.221*** (2.90) (2.83) (2.40) (3.30) (3.15) (2.75) Average adj. R$$^{\mathrm{2}}$$ 0.006 0.006 0.007 0.006 0.006 0.007 This table presents the regression results on how information asymmetry influences the predictive power of Active Weights for subsequent earnings announcement returns. We measure earnings announcement returns using the 3-day abnormal returns in percentage terms surrounding earnings announcements. We measure active weights using two approaches: an index-based measure is the weight of a stock in a fund’s portfolio in excess of the stock’s weight in the fund’s benchmark index and a trade-based measure is the change in the weight of a stock in a fund’s portfolio in the past year, controlling for stock price movements. Analyst coverage is the natural log of one plus the number of analysts covering the firm. Idiosyncratic volatility (IV) of daily stock returns is the standard deviation of residual returns from the Fama-French model in the past quarter. Disagreement (analyst forecast dispersion) is the standard deviation of analyst earnings forecasts, times 10. The control variables include firm size (the natural log of market cap), the book-to-market ratio, and momentum (past 1-year return skipping the most recent month). We performed these regressions using stock-fund pooled data for each quarter from 1984Q1 to 2014Q2. The statistical inference is based on the Fama and Macbeth (1973) procedure with Newey and West (1987) adjustments. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. 4.2 When do active fundamental managers outperform their peers? Along another dimension, we exploit the variation in the predictive power of AFP for performance over time. Specifically, we examine how the association between AFP and future fund performance varies with aggregate disagreement among financial analysts. When analyst disagreement is higher, fund managers’ ability to estimate the fundamental values should be more valuable, and the association between AFP and future fund performance therefore should be stronger. To examine this conjecture, we construct a time-series variable that aggregates stock-level analyst disagreement (i.e., cross-sectional average of the standard deviation of analyst earnings forecasts, deflated by the absolute value of the consensus forecast for individual stocks). We test how the relative performance of mutual funds with high AFP varies with aggregate analyst disagreement. Table 7 presents the results. For ease of interpretation, we standardize the aggregate analyst disagreement to have a mean of 0 and standard deviation of 1. Consistent with our conjecture, the results indicate that mutual funds with higher AFP tend to deliver higher performance when analyst disagreement is high. The time variation in the performance of active fundamental managers is not only statistically significant but also economically large. In multivariate regressions, a one-standard-deviation increase in analyst disagreement more than doubles the performance forecasting power of AFP. These results again suggest that AFP is driven by a manager’s ability to estimate firm fundamental values. Table 7 Time-varying performance predictive power of AFP Index-based AFP Trade-based AFP Net Net Gross Gross Net Net Gross Gross (1) (2) (3) (4) (5) (6) (7) (8) AFP 2.373*** 2.047*** 2.445*** 2.033*** 1.226* 1.240* 1.348** 1.244* (3.41) (2.89) (3.51) (2.87) (1.80) (1.79) (1.98) (1.80) AFP $$\times$$ Disagreement 2.081*** 1.871** 2.091*** 1.866** 3.103*** 2.790*** 3.117*** 2.787*** (2.85) (2.49) (2.85) (2.48) (4.23) (3.68) (4.23) (3.67) Disagreement 0.0493*** 0.0494*** 0.0460*** 0.0485*** 0.0498*** 0.0498*** 0.0466*** 0.0489*** (11.13) (10.76) (10.43) (10.55) (11.22) (10.80) (10.52) (10.59) log(TNA) –0.00294 –0.00470* –0.00276 –0.00451* (–1.14) (–1.82) (–1.06) (–1.74) log(Age) –0.0125** –0.0114* –0.0128** –0.0117** (–2.13) (–1.95) (–2.17) (–1.99) Expense –7.798*** –0.949 –7.797*** –0.949 (–6.60) (–0.80) (–6.59) (–0.80) Turnover –0.0221*** –0.0211*** –0.0221*** –0.0211*** (–3.22) (–3.11) (–3.22) (–3.10) Past flow 0.0902** 0.0830** 0.0893** 0.0821** (2.20) (2.02) (2.17) (2.00) Past alpha 1.759 1.584 1.723 1.547 (1.22) (1.10) (1.19) (1.07) Adj. R$$^{\mathrm{2}}$$ 0.00104 0.00163 0.000940 0.00119 0.001 0.002 0.001 0.001 N 239,966 222,398 239,966 222,398 239,966 222,398 239,966 222,398 Index-based AFP Trade-based AFP Net Net Gross Gross Net Net Gross Gross (1) (2) (3) (4) (5) (6) (7) (8) AFP 2.373*** 2.047*** 2.445*** 2.033*** 1.226* 1.240* 1.348** 1.244* (3.41) (2.89) (3.51) (2.87) (1.80) (1.79) (1.98) (1.80) AFP $$\times$$ Disagreement 2.081*** 1.871** 2.091*** 1.866** 3.103*** 2.790*** 3.117*** 2.787*** (2.85) (2.49) (2.85) (2.48) (4.23) (3.68) (4.23) (3.67) Disagreement 0.0493*** 0.0494*** 0.0460*** 0.0485*** 0.0498*** 0.0498*** 0.0466*** 0.0489*** (11.13) (10.76) (10.43) (10.55) (11.22) (10.80) (10.52) (10.59) log(TNA) –0.00294 –0.00470* –0.00276 –0.00451* (–1.14) (–1.82) (–1.06) (–1.74) log(Age) –0.0125** –0.0114* –0.0128** –0.0117** (–2.13) (–1.95) (–2.17) (–1.99) Expense –7.798*** –0.949 –7.797*** –0.949 (–6.60) (–0.80) (–6.59) (–0.80) Turnover –0.0221*** –0.0211*** –0.0221*** –0.0211*** (–3.22) (–3.11) (–3.22) (–3.10) Past flow 0.0902** 0.0830** 0.0893** 0.0821** (2.20) (2.02) (2.17) (2.00) Past alpha 1.759 1.584 1.723 1.547 (1.22) (1.10) (1.19) (1.07) Adj. R$$^{\mathrm{2}}$$ 0.00104 0.00163 0.000940 0.00119 0.001 0.002 0.001 0.001 N 239,966 222,398 239,966 222,398 239,966 222,398 239,966 222,398 This table shows the variation in the association between the AFP and future fund performance across time. We include both index-based and trade-based AFP, like in Equation (3). Disagreement (analyst forecast dispersion) is the cross-sectional average of the standard deviation of analyst earnings forecasts. For ease of interpretation, it is standardized to have a mean of zero and standard deviation of one. We measure future mutual fund performance using Carhart’s (1997) four-factor alpha (as a percentage), where fund betas are estimated using rolling-window regressions in the past 3 years. The panel regressions control for fund size, fund age, expense ratio, fund turnover, fund percentage flow in the past quarter, and fund alpha in the past 3 years. The regressions include time fixed effects, and the standard errors are clustered by fund. $$t$$-statistics are presented in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 7 Time-varying performance predictive power of AFP Index-based AFP Trade-based AFP Net Net Gross Gross Net Net Gross Gross (1) (2) (3) (4) (5) (6) (7) (8) AFP 2.373*** 2.047*** 2.445*** 2.033*** 1.226* 1.240* 1.348** 1.244* (3.41) (2.89) (3.51) (2.87) (1.80) (1.79) (1.98) (1.80) AFP $$\times$$ Disagreement 2.081*** 1.871** 2.091*** 1.866** 3.103*** 2.790*** 3.117*** 2.787*** (2.85) (2.49) (2.85) (2.48) (4.23) (3.68) (4.23) (3.67) Disagreement 0.0493*** 0.0494*** 0.0460*** 0.0485*** 0.0498*** 0.0498*** 0.0466*** 0.0489*** (11.13) (10.76) (10.43) (10.55) (11.22) (10.80) (10.52) (10.59) log(TNA) –0.00294 –0.00470* –0.00276 –0.00451* (–1.14) (–1.82) (–1.06) (–1.74) log(Age) –0.0125** –0.0114* –0.0128** –0.0117** (–2.13) (–1.95) (–2.17) (–1.99) Expense –7.798*** –0.949 –7.797*** –0.949 (–6.60) (–0.80) (–6.59) (–0.80) Turnover –0.0221*** –0.0211*** –0.0221*** –0.0211*** (–3.22) (–3.11) (–3.22) (–3.10) Past flow 0.0902** 0.0830** 0.0893** 0.0821** (2.20) (2.02) (2.17) (2.00) Past alpha 1.759 1.584 1.723 1.547 (1.22) (1.10) (1.19) (1.07) Adj. R$$^{\mathrm{2}}$$ 0.00104 0.00163 0.000940 0.00119 0.001 0.002 0.001 0.001 N 239,966 222,398 239,966 222,398 239,966 222,398 239,966 222,398 Index-based AFP Trade-based AFP Net Net Gross Gross Net Net Gross Gross (1) (2) (3) (4) (5) (6) (7) (8) AFP 2.373*** 2.047*** 2.445*** 2.033*** 1.226* 1.240* 1.348** 1.244* (3.41) (2.89) (3.51) (2.87) (1.80) (1.79) (1.98) (1.80) AFP $$\times$$ Disagreement 2.081*** 1.871** 2.091*** 1.866** 3.103*** 2.790*** 3.117*** 2.787*** (2.85) (2.49) (2.85) (2.48) (4.23) (3.68) (4.23) (3.67) Disagreement 0.0493*** 0.0494*** 0.0460*** 0.0485*** 0.0498*** 0.0498*** 0.0466*** 0.0489*** (11.13) (10.76) (10.43) (10.55) (11.22) (10.80) (10.52) (10.59) log(TNA) –0.00294 –0.00470* –0.00276 –0.00451* (–1.14) (–1.82) (–1.06) (–1.74) log(Age) –0.0125** –0.0114* –0.0128** –0.0117** (–2.13) (–1.95) (–2.17) (–1.99) Expense –7.798*** –0.949 –7.797*** –0.949 (–6.60) (–0.80) (–6.59) (–0.80) Turnover –0.0221*** –0.0211*** –0.0221*** –0.0211*** (–3.22) (–3.11) (–3.22) (–3.10) Past flow 0.0902** 0.0830** 0.0893** 0.0821** (2.20) (2.02) (2.17) (2.00) Past alpha 1.759 1.584 1.723 1.547 (1.22) (1.10) (1.19) (1.07) Adj. R$$^{\mathrm{2}}$$ 0.00104 0.00163 0.000940 0.00119 0.001 0.002 0.001 0.001 N 239,966 222,398 239,966 222,398 239,966 222,398 239,966 222,398 This table shows the variation in the association between the AFP and future fund performance across time. We include both index-based and trade-based AFP, like in Equation (3). Disagreement (analyst forecast dispersion) is the cross-sectional average of the standard deviation of analyst earnings forecasts. For ease of interpretation, it is standardized to have a mean of zero and standard deviation of one. We measure future mutual fund performance using Carhart’s (1997) four-factor alpha (as a percentage), where fund betas are estimated using rolling-window regressions in the past 3 years. The panel regressions control for fund size, fund age, expense ratio, fund turnover, fund percentage flow in the past quarter, and fund alpha in the past 3 years. The regressions include time fixed effects, and the standard errors are clustered by fund. $$t$$-statistics are presented in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. 4.3 Persistence of AFP Previous studies indicate some persistence in fund returns but also note that persistence can be largely explained away by the momentum factor, except for worst performers (e.g., Brown and Goetzmann 1995; Elton, Gruber, and Blake 1996; Carhart 1997). Table 8 summarizes the findings on the persistence of individual managers’ AFP. For each quarter between 1984Q1 and 2012Q2, we sort mutual funds into decile portfolios by their AFP and compute the average AFP for the subsequent 2 years. Panel A shows the results for index-based AFP, and panel B shows the results for trade-based AFP. Table 8 Persistence of AFP AFP$$_{\mathrm{t+1}}$$ $$t$$-statistic AFP$$_{\mathrm{t+2}}$$ $$t$$-statistic AFP$$_{\mathrm{t+3}}$$ $$t$$-statistic AFP$$_{\mathrm{t+4}}$$ $$t$$-statistic AFP$$_{\mathrm{t+8}}$$ $$t$$-statistic A. Index-based AFP Low 2.36 0.97 4.05 1.54 3.02 1.19 6.84 2.78 7.47 2.91 2 4.22 2.15 3.45 1.70 2.90 1.33 5.19 2.51 4.07 2.15 3 4.21 2.18 5.22 2.74 3.68 2.14 6.51 3.72 5.05 2.30 4 3.66 2.03 5.03 2.69 3.30 1.74 6.15 3.66 6.13 3.42 5 5.70 3.03 6.32 3.43 3.81 1.97 4.30 2.36 5.36 2.65 6 5.31 3.10 5.15 2.84 7.86 4.47 6.87 3.76 7.12 3.99 7 6.17 3.21 7.38 4.37 5.87 3.00 6.45 3.52 5.44 2.73 8 8.14 3.94 8.61 5.73 8.87 4.33 6.40 3.19 8.49 4.34 9 9.64 4.44 10.96 5.30 9.00 4.68 9.43 4.77 8.10 3.28 High 12.15 4.38 15.22 5.69 13.50 5.41 11.83 4.48 13.26 5.01 High-low 9.79 3.94 11.17 4.33 10.48 4.16 4.99 2.03 5.79 2.34 B. Trade-based AFP Low 3.70 2.35 2.92 1.78 6.28 3.53 4.20 2.63 1.44 0.98 2 2.26 1.84 1.89 1.61 1.65 1.21 3.37 2.92 2.90 2.38 3 2.73 2.57 0.34 0.31 1.70 1.32 2.79 2.29 1.34 0.91 4 0.55 0.43 2.08 1.89 1.97 1.63 2.77 2.33 1.21 0.93 5 1.02 1.05 1.85 1.75 0.48 0.39 2.86 2.03 3.43 2.59 6 1.18 1.02 2.66 2.89 2.74 2.37 1.86 1.46 3.63 2.42 7 2.48 2.14 3.00 2.62 1.92 1.53 –0.50 –0.44 2.09 1.69 8 4.04 2.99 3.66 2.58 1.52 1.12 2.29 1.42 2.98 2.32 9 4.97 3.69 5.80 4.51 2.90 1.90 2.36 1.62 4.38 2.97 High 7.68 4.22 8.97 5.34 6.04 3.50 2.40 1.27 2.89 1.70 High-low 3.98 1.94 6.04 3.71 –0.23 –0.12 –1.80 –0.93 1.45 0.85 AFP$$_{\mathrm{t+1}}$$ $$t$$-statistic AFP$$_{\mathrm{t+2}}$$ $$t$$-statistic AFP$$_{\mathrm{t+3}}$$ $$t$$-statistic AFP$$_{\mathrm{t+4}}$$ $$t$$-statistic AFP$$_{\mathrm{t+8}}$$ $$t$$-statistic A. Index-based AFP Low 2.36 0.97 4.05 1.54 3.02 1.19 6.84 2.78 7.47 2.91 2 4.22 2.15 3.45 1.70 2.90 1.33 5.19 2.51 4.07 2.15 3 4.21 2.18 5.22 2.74 3.68 2.14 6.51 3.72 5.05 2.30 4 3.66 2.03 5.03 2.69 3.30 1.74 6.15 3.66 6.13 3.42 5 5.70 3.03 6.32 3.43 3.81 1.97 4.30 2.36 5.36 2.65 6 5.31 3.10 5.15 2.84 7.86 4.47 6.87 3.76 7.12 3.99 7 6.17 3.21 7.38 4.37 5.87 3.00 6.45 3.52 5.44 2.73 8 8.14 3.94 8.61 5.73 8.87 4.33 6.40 3.19 8.49 4.34 9 9.64 4.44 10.96 5.30 9.00 4.68 9.43 4.77 8.10 3.28 High 12.15 4.38 15.22 5.69 13.50 5.41 11.83 4.48 13.26 5.01 High-low 9.79 3.94 11.17 4.33 10.48 4.16 4.99 2.03 5.79 2.34 B. Trade-based AFP Low 3.70 2.35 2.92 1.78 6.28 3.53 4.20 2.63 1.44 0.98 2 2.26 1.84 1.89 1.61 1.65 1.21 3.37 2.92 2.90 2.38 3 2.73 2.57 0.34 0.31 1.70 1.32 2.79 2.29 1.34 0.91 4 0.55 0.43 2.08 1.89 1.97 1.63 2.77 2.33 1.21 0.93 5 1.02 1.05 1.85 1.75 0.48 0.39 2.86 2.03 3.43 2.59 6 1.18 1.02 2.66 2.89 2.74 2.37 1.86 1.46 3.63 2.42 7 2.48 2.14 3.00 2.62 1.92 1.53 –0.50 –0.44 2.09 1.69 8 4.04 2.99 3.66 2.58 1.52 1.12 2.29 1.42 2.98 2.32 9 4.97 3.69 5.80 4.51 2.90 1.90 2.36 1.62 4.38 2.97 High 7.68 4.22 8.97 5.34 6.04 3.50 2.40 1.27 2.89 1.70 High-low 3.98 1.94 6.04 3.71 –0.23 –0.12 –1.80 –0.93 1.45 0.85 This table shows the persistence of the AFP for both index-based (panel A) and trade-based (panel B) measures. Like in Equation (3), AFP, in basis points, is the sum of the product of active fund weights (or trades) and subsequent 3-day abnormal returns surrounding earnings announcements. For each quarter between 1984 and 2014, we sort funds into decile portfolios using their AFP and compute the average AFP for the subsequent eight quarters. Table 8 Persistence of AFP AFP$$_{\mathrm{t+1}}$$ $$t$$-statistic AFP$$_{\mathrm{t+2}}$$ $$t$$-statistic AFP$$_{\mathrm{t+3}}$$ $$t$$-statistic AFP$$_{\mathrm{t+4}}$$ $$t$$-statistic AFP$$_{\mathrm{t+8}}$$ $$t$$-statistic A. Index-based AFP Low 2.36 0.97 4.05 1.54 3.02 1.19 6.84 2.78 7.47 2.91 2 4.22 2.15 3.45 1.70 2.90 1.33 5.19 2.51 4.07 2.15 3 4.21 2.18 5.22 2.74 3.68 2.14 6.51 3.72 5.05 2.30 4 3.66 2.03 5.03 2.69 3.30 1.74 6.15 3.66 6.13 3.42 5 5.70 3.03 6.32 3.43 3.81 1.97 4.30 2.36 5.36 2.65 6 5.31 3.10 5.15 2.84 7.86 4.47 6.87 3.76 7.12 3.99 7 6.17 3.21 7.38 4.37 5.87 3.00 6.45 3.52 5.44 2.73 8 8.14 3.94 8.61 5.73 8.87 4.33 6.40 3.19 8.49 4.34 9 9.64 4.44 10.96 5.30 9.00 4.68 9.43 4.77 8.10 3.28 High 12.15 4.38 15.22 5.69 13.50 5.41 11.83 4.48 13.26 5.01 High-low 9.79 3.94 11.17 4.33 10.48 4.16 4.99 2.03 5.79 2.34 B. Trade-based AFP Low 3.70 2.35 2.92 1.78 6.28 3.53 4.20 2.63 1.44 0.98 2 2.26 1.84 1.89 1.61 1.65 1.21 3.37 2.92 2.90 2.38 3 2.73 2.57 0.34 0.31 1.70 1.32 2.79 2.29 1.34 0.91 4 0.55 0.43 2.08 1.89 1.97 1.63 2.77 2.33 1.21 0.93 5 1.02 1.05 1.85 1.75 0.48 0.39 2.86 2.03 3.43 2.59 6 1.18 1.02 2.66 2.89 2.74 2.37 1.86 1.46 3.63 2.42 7 2.48 2.14 3.00 2.62 1.92 1.53 –0.50 –0.44 2.09 1.69 8 4.04 2.99 3.66 2.58 1.52 1.12 2.29 1.42 2.98 2.32 9 4.97 3.69 5.80 4.51 2.90 1.90 2.36 1.62 4.38 2.97 High 7.68 4.22 8.97 5.34 6.04 3.50 2.40 1.27 2.89 1.70 High-low 3.98 1.94 6.04 3.71 –0.23 –0.12 –1.80 –0.93 1.45 0.85 AFP$$_{\mathrm{t+1}}$$ $$t$$-statistic AFP$$_{\mathrm{t+2}}$$ $$t$$-statistic AFP$$_{\mathrm{t+3}}$$ $$t$$-statistic AFP$$_{\mathrm{t+4}}$$ $$t$$-statistic AFP$$_{\mathrm{t+8}}$$ $$t$$-statistic A. Index-based AFP Low 2.36 0.97 4.05 1.54 3.02 1.19 6.84 2.78 7.47 2.91 2 4.22 2.15 3.45 1.70 2.90 1.33 5.19 2.51 4.07 2.15 3 4.21 2.18 5.22 2.74 3.68 2.14 6.51 3.72 5.05 2.30 4 3.66 2.03 5.03 2.69 3.30 1.74 6.15 3.66 6.13 3.42 5 5.70 3.03 6.32 3.43 3.81 1.97 4.30 2.36 5.36 2.65 6 5.31 3.10 5.15 2.84 7.86 4.47 6.87 3.76 7.12 3.99 7 6.17 3.21 7.38 4.37 5.87 3.00 6.45 3.52 5.44 2.73 8 8.14 3.94 8.61 5.73 8.87 4.33 6.40 3.19 8.49 4.34 9 9.64 4.44 10.96 5.30 9.00 4.68 9.43 4.77 8.10 3.28 High 12.15 4.38 15.22 5.69 13.50 5.41 11.83 4.48 13.26 5.01 High-low 9.79 3.94 11.17 4.33 10.48 4.16 4.99 2.03 5.79 2.34 B. Trade-based AFP Low 3.70 2.35 2.92 1.78 6.28 3.53 4.20 2.63 1.44 0.98 2 2.26 1.84 1.89 1.61 1.65 1.21 3.37 2.92 2.90 2.38 3 2.73 2.57 0.34 0.31 1.70 1.32 2.79 2.29 1.34 0.91 4 0.55 0.43 2.08 1.89 1.97 1.63 2.77 2.33 1.21 0.93 5 1.02 1.05 1.85 1.75 0.48 0.39 2.86 2.03 3.43 2.59 6 1.18 1.02 2.66 2.89 2.74 2.37 1.86 1.46 3.63 2.42 7 2.48 2.14 3.00 2.62 1.92 1.53 –0.50 –0.44 2.09 1.69 8 4.04 2.99 3.66 2.58 1.52 1.12 2.29 1.42 2.98 2.32 9 4.97 3.69 5.80 4.51 2.90 1.90 2.36 1.62 4.38 2.97 High 7.68 4.22 8.97 5.34 6.04 3.50 2.40 1.27 2.89 1.70 High-low 3.98 1.94 6.04 3.71 –0.23 –0.12 –1.80 –0.93 1.45 0.85 This table shows the persistence of the AFP for both index-based (panel A) and trade-based (panel B) measures. Like in Equation (3), AFP, in basis points, is the sum of the product of active fund weights (or trades) and subsequent 3-day abnormal returns surrounding earnings announcements. For each quarter between 1984 and 2014, we sort funds into decile portfolios using their AFP and compute the average AFP for the subsequent eight quarters. The results indicate that AFP is a persistent fund attribute. For index-based AFP, the divergence in AFP between mutual funds in the top decile of high AFP and those in the bottom decile of low AFP remains economically meaningful and statistically significant for the 2 years after portfolio formation. Notably, this persistence is particularly pronounced for funds with superior skills in decile 10. Although the trade-based AFP shows a lower degree of persistence, funds with high AFP in decile 10 still have high and statistically significant AFP three quarters after the portfolio formation. 5. Robustness Tests Several studies show that certain stock characteristics are associated with abnormal returns around firm earnings announcements. For example, Bernard and Thomas (1989, 1990) find that firms’ earnings surprises tend to be persistent. Therefore, as a robustness test, we determine whether our results are driven by mutual fund managers who tilt their portfolios toward stock characteristics that associate with future earnings announcement returns. Specifically, for each quarter, we run cross-sectional regressions of the 3-day abnormal returns during earnings announcements on stock characteristics, such as firm size, the book-to-market ratio, past stock returns, and past earnings announcement returns. We then use the regression residuals as inputs to compute mutual funds’ AFP. We sort mutual funds into ten portfolios, in accordance with this modified measure of AFP, and rebalance the portfolios quarterly. Table 9 reports the returns on these fund portfolios, net of and before expenses. The results indicate that even after we orthogonalize abnormal returns surrounding earnings announcements to relevant stock characteristics, AFP remains a strong predictor of future fund performance. For example, mutual funds with high index-based AFP outperform their peers with low AFP by 20 bps per month, which cannot be explained by their differential exposures to risk or style factors. Table 9 AFP and mutual fund returns: Robust AFP using residual earnings announcement returns Low 2 3 4 5 6 7 8 9 High High-low A. Index-based AFP Average return 0.86 0.86 0.88 0.87 0.85 0.88 0.88 0.92 0.94 1.06 0.20** (3.05) (3.20) (3.37) (3.36) (3.24) (3.40) (3.35) (3.39) (3.42) (3.57) (2.31) CAPM $$\alpha $$ –0.12 –0.10 –0.07 –0.07 –0.10 –0.06 –0.07 –0.05 –0.03 0.06 0.18** (–1.65) (–1.78) (–1.28) (–1.55) (–2.18) (–1.35) (–1.56) (–0.88) (–0.5) (0.62) (2.13) Fama-French $$\alpha $$ –0.14 –0.12 –0.09 –0.10 –0.13 –0.09 –0.08 –0.06 –0.03 0.10 0.24*** (–2.28) (–2.53) (–1.83) (–2.5) (–3.18) (–2.22) (–2.19) (–1.33) (–0.53) (1.62) (2.78) Carhart $$\alpha $$ –0.10 –0.10 –0.07 –0.09 –0.12 –0.09 –0.08 –0.07 –0.05 0.06 0.17** (–1.69) (–2.08) (–1.47) (–2.18) (–2.87) (–2.04) (–2.22) (–1.75) (–0.99) (1.02) (2.29) Pastor-Stambaugh $$\alpha $$ –0.11 –0.11 –0.08 –0.10 –0.13 –0.09 –0.09 –0.08 –0.05 0.08 0.19** (–1.82) (–2.24) (–1.54) (–2.41) (–3) (–2.09) (–2.3) (–1.76) (–0.94) (1.22) (2.59) Six-factor: PEAD1 –0.28 –0.24 –0.24 –0.23 –0.29 –0.23 –0.19 –0.15 –0.14 0.03 0.30*** (–3.3) (–3.64) (–4.31) (–4.17) (–5.36) (–4.38) (–3.63) (–2.34) (–1.96) (0.27) (2.92) Six-factor: PEAD 2 –0.30 –0.26 –0.26 –0.24 –0.30 –0.25 –0.20 –0.16 –0.15 0.02 0.32*** (–3.63) (–4.04) (–4.82) (–4.56) (–5.9) (–4.92) (–4.04) (–2.49) (–2.12) (0.18) (2.92) Six-factor: PEAD3 –0.21 –0.17 –0.14 –0.17 –0.19 –0.14 –0.13 –0.12 –0.11 0.01 0.23*** (–3.51) (–3.37) (–2.92) (–3.92) (–4.22) (–3.29) (–3.55) (–2.8) (–2.25) (0.20) (3.09) B. Trade-based AFP Average return 0.88 0.91 0.90 0.94 0.89 0.89 0.90 0.91 0.95 0.95 0.07** (3.22) (3.58) (3.69) (3.92) (3.66) (3.73) (3.70) (3.69) (3.75) (3.49) (1.98) CAPM $$\alpha $$ –0.14 –0.07 –0.06 0.00 –0.07 –0.05 –0.05 –0.05 –0.03 –0.06 0.08** (–1.97) (–1.52) (–1.26) (–0.08) (–1.46) (–1.28) (–1.25) (–1.21) (–0.6) (–0.86) (2.09) Fama-French $$\alpha $$ –0.12 –0.08 –0.09 –0.04 –0.10 –0.08 –0.08 –0.07 –0.03 –0.02 0.10*** (–2.65) (–2.22) (–2.51) (–1.08) (–2.71) (–2.42) (–2.19) (–1.96) (–0.76) (–0.52) (2.76) Carhart $$\alpha $$ –0.14 –0.08 –0.08 –0.02 –0.09 –0.07 –0.07 –0.07 –0.04 –0.06 0.08** (–2.9) (–1.9) (–2.13) (–0.44) (–2.3) (–2.05) (–2.03) (–1.95) (–1.1) (–1.29) (2.25) Pastor-Stambaugh $$\alpha $$ –0.14 –0.09 –0.09 –0.03 –0.10 –0.09 –0.08 –0.08 –0.05 –0.06 0.08** (–2.76) (–2.07) (–2.28) (–0.7) (–2.46) (–2.44) (–2.29) (–2.33) (–1.28) (–1.19) (2.27) Six-factor: PEAD1 –0.24 –0.20 –0.22 –0.18 –0.25 –0.20 –0.19 –0.19 –0.13 –0.10 0.14*** (–2.92) (–2.95) (–3.81) (–3.37) (–4.44) (–3.43) (–3.36) (–3.64) (–2.53) (–1.25) (3.37) Six-factor: PEAD 2 –0.26 –0.22 –0.24 –0.20 –0.27 –0.21 –0.21 –0.20 –0.14 –0.11 0.15*** (–3.17) (–3.35) (–4.28) (–3.98) (–4.98) (–3.85) (–3.74) (–3.93) (–2.75) (–1.36) (3.47) Six-factor: PEAD3 –0.23 –0.14 –0.15 –0.09 –0.16 –0.15 –0.14 –0.15 –0.09 –0.11 0.12*** (–3.88) (–3.05) (–3.51) (–1.88) (–3.44) (–3.55) (–3.33) (–3.7) (–1.94) (–1.96) (2.73) Low 2 3 4 5 6 7 8 9 High High-low A. Index-based AFP Average return 0.86 0.86 0.88 0.87 0.85 0.88 0.88 0.92 0.94 1.06 0.20** (3.05) (3.20) (3.37) (3.36) (3.24) (3.40) (3.35) (3.39) (3.42) (3.57) (2.31) CAPM $$\alpha $$ –0.12 –0.10 –0.07 –0.07 –0.10 –0.06 –0.07 –0.05 –0.03 0.06 0.18** (–1.65) (–1.78) (–1.28) (–1.55) (–2.18) (–1.35) (–1.56) (–0.88) (–0.5) (0.62) (2.13) Fama-French $$\alpha $$ –0.14 –0.12 –0.09 –0.10 –0.13 –0.09 –0.08 –0.06 –0.03 0.10 0.24*** (–2.28) (–2.53) (–1.83) (–2.5) (–3.18) (–2.22) (–2.19) (–1.33) (–0.53) (1.62) (2.78) Carhart $$\alpha $$ –0.10 –0.10 –0.07 –0.09 –0.12 –0.09 –0.08 –0.07 –0.05 0.06 0.17** (–1.69) (–2.08) (–1.47) (–2.18) (–2.87) (–2.04) (–2.22) (–1.75) (–0.99) (1.02) (2.29) Pastor-Stambaugh $$\alpha $$ –0.11 –0.11 –0.08 –0.10 –0.13 –0.09 –0.09 –0.08 –0.05 0.08 0.19** (–1.82) (–2.24) (–1.54) (–2.41) (–3) (–2.09) (–2.3) (–1.76) (–0.94) (1.22) (2.59) Six-factor: PEAD1 –0.28 –0.24 –0.24 –0.23 –0.29 –0.23 –0.19 –0.15 –0.14 0.03 0.30*** (–3.3) (–3.64) (–4.31) (–4.17) (–5.36) (–4.38) (–3.63) (–2.34) (–1.96) (0.27) (2.92) Six-factor: PEAD 2 –0.30 –0.26 –0.26 –0.24 –0.30 –0.25 –0.20 –0.16 –0.15 0.02 0.32*** (–3.63) (–4.04) (–4.82) (–4.56) (–5.9) (–4.92) (–4.04) (–2.49) (–2.12) (0.18) (2.92) Six-factor: PEAD3 –0.21 –0.17 –0.14 –0.17 –0.19 –0.14 –0.13 –0.12 –0.11 0.01 0.23*** (–3.51) (–3.37) (–2.92) (–3.92) (–4.22) (–3.29) (–3.55) (–2.8) (–2.25) (0.20) (3.09) B. Trade-based AFP Average return 0.88 0.91 0.90 0.94 0.89 0.89 0.90 0.91 0.95 0.95 0.07** (3.22) (3.58) (3.69) (3.92) (3.66) (3.73) (3.70) (3.69) (3.75) (3.49) (1.98) CAPM $$\alpha $$ –0.14 –0.07 –0.06 0.00 –0.07 –0.05 –0.05 –0.05 –0.03 –0.06 0.08** (–1.97) (–1.52) (–1.26) (–0.08) (–1.46) (–1.28) (–1.25) (–1.21) (–0.6) (–0.86) (2.09) Fama-French $$\alpha $$ –0.12 –0.08 –0.09 –0.04 –0.10 –0.08 –0.08 –0.07 –0.03 –0.02 0.10*** (–2.65) (–2.22) (–2.51) (–1.08) (–2.71) (–2.42) (–2.19) (–1.96) (–0.76) (–0.52) (2.76) Carhart $$\alpha $$ –0.14 –0.08 –0.08 –0.02 –0.09 –0.07 –0.07 –0.07 –0.04 –0.06 0.08** (–2.9) (–1.9) (–2.13) (–0.44) (–2.3) (–2.05) (–2.03) (–1.95) (–1.1) (–1.29) (2.25) Pastor-Stambaugh $$\alpha $$ –0.14 –0.09 –0.09 –0.03 –0.10 –0.09 –0.08 –0.08 –0.05 –0.06 0.08** (–2.76) (–2.07) (–2.28) (–0.7) (–2.46) (–2.44) (–2.29) (–2.33) (–1.28) (–1.19) (2.27) Six-factor: PEAD1 –0.24 –0.20 –0.22 –0.18 –0.25 –0.20 –0.19 –0.19 –0.13 –0.10 0.14*** (–2.92) (–2.95) (–3.81) (–3.37) (–4.44) (–3.43) (–3.36) (–3.64) (–2.53) (–1.25) (3.37) Six-factor: PEAD 2 –0.26 –0.22 –0.24 –0.20 –0.27 –0.21 –0.21 –0.20 –0.14 –0.11 0.15*** (–3.17) (–3.35) (–4.28) (–3.98) (–4.98) (–3.85) (–3.74) (–3.93) (–2.75) (–1.36) (3.47) Six-factor: PEAD3 –0.23 –0.14 –0.15 –0.09 –0.16 –0.15 –0.14 –0.15 –0.09 –0.11 0.12*** (–3.88) (–3.05) (–3.51) (–1.88) (–3.44) (–3.55) (–3.33) (–3.7) (–1.94) (–1.96) (2.73) This table presents the performance of decile fund portfolios formed according to their index-based (panel A) and trade-based (panel B) AFP. Like in Equation (3), we compute values using subsequent residual earnings announcement returns defined as residuals from cross-sectional regressions of 3-day abnormal returns during earnings announcements on firm size, book-to-market, past 12-month return, and the 3-day abnormal returns on earnings announcement days in the previous quarter. We form and rebalance the decile portfolios at the end of 2 months after each quarter from 1984Q1 to 2014Q2, and the return series ranges from June 1984 to November 2014. Decile 10 is the portfolio of funds with the highest AFP value. We compute equally weighted net fund returns on the portfolios, as well as risk-adjusted returns based on the CAPM, the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model, the Pastor and Stambaugh (2003) five-factor model, and three versions of the six-factor models that include the return to a strategy that exploits the post-earnings announcement drift. We report the alphas as a monthly percentage. The $$t$$-statistics are shown in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 9 AFP and mutual fund returns: Robust AFP using residual earnings announcement returns Low 2 3 4 5 6 7 8 9 High High-low A. Index-based AFP Average return 0.86 0.86 0.88 0.87 0.85 0.88 0.88 0.92 0.94 1.06 0.20** (3.05) (3.20) (3.37) (3.36) (3.24) (3.40) (3.35) (3.39) (3.42) (3.57) (2.31) CAPM $$\alpha $$ –0.12 –0.10 –0.07 –0.07 –0.10 –0.06 –0.07 –0.05 –0.03 0.06 0.18** (–1.65) (–1.78) (–1.28) (–1.55) (–2.18) (–1.35) (–1.56) (–0.88) (–0.5) (0.62) (2.13) Fama-French $$\alpha $$ –0.14 –0.12 –0.09 –0.10 –0.13 –0.09 –0.08 –0.06 –0.03 0.10 0.24*** (–2.28) (–2.53) (–1.83) (–2.5) (–3.18) (–2.22) (–2.19) (–1.33) (–0.53) (1.62) (2.78) Carhart $$\alpha $$ –0.10 –0.10 –0.07 –0.09 –0.12 –0.09 –0.08 –0.07 –0.05 0.06 0.17** (–1.69) (–2.08) (–1.47) (–2.18) (–2.87) (–2.04) (–2.22) (–1.75) (–0.99) (1.02) (2.29) Pastor-Stambaugh $$\alpha $$ –0.11 –0.11 –0.08 –0.10 –0.13 –0.09 –0.09 –0.08 –0.05 0.08 0.19** (–1.82) (–2.24) (–1.54) (–2.41) (–3) (–2.09) (–2.3) (–1.76) (–0.94) (1.22) (2.59) Six-factor: PEAD1 –0.28 –0.24 –0.24 –0.23 –0.29 –0.23 –0.19 –0.15 –0.14 0.03 0.30*** (–3.3) (–3.64) (–4.31) (–4.17) (–5.36) (–4.38) (–3.63) (–2.34) (–1.96) (0.27) (2.92) Six-factor: PEAD 2 –0.30 –0.26 –0.26 –0.24 –0.30 –0.25 –0.20 –0.16 –0.15 0.02 0.32*** (–3.63) (–4.04) (–4.82) (–4.56) (–5.9) (–4.92) (–4.04) (–2.49) (–2.12) (0.18) (2.92) Six-factor: PEAD3 –0.21 –0.17 –0.14 –0.17 –0.19 –0.14 –0.13 –0.12 –0.11 0.01 0.23*** (–3.51) (–3.37) (–2.92) (–3.92) (–4.22) (–3.29) (–3.55) (–2.8) (–2.25) (0.20) (3.09) B. Trade-based AFP Average return 0.88 0.91 0.90 0.94 0.89 0.89 0.90 0.91 0.95 0.95 0.07** (3.22) (3.58) (3.69) (3.92) (3.66) (3.73) (3.70) (3.69) (3.75) (3.49) (1.98) CAPM $$\alpha $$ –0.14 –0.07 –0.06 0.00 –0.07 –0.05 –0.05 –0.05 –0.03 –0.06 0.08** (–1.97) (–1.52) (–1.26) (–0.08) (–1.46) (–1.28) (–1.25) (–1.21) (–0.6) (–0.86) (2.09) Fama-French $$\alpha $$ –0.12 –0.08 –0.09 –0.04 –0.10 –0.08 –0.08 –0.07 –0.03 –0.02 0.10*** (–2.65) (–2.22) (–2.51) (–1.08) (–2.71) (–2.42) (–2.19) (–1.96) (–0.76) (–0.52) (2.76) Carhart $$\alpha $$ –0.14 –0.08 –0.08 –0.02 –0.09 –0.07 –0.07 –0.07 –0.04 –0.06 0.08** (–2.9) (–1.9) (–2.13) (–0.44) (–2.3) (–2.05) (–2.03) (–1.95) (–1.1) (–1.29) (2.25) Pastor-Stambaugh $$\alpha $$ –0.14 –0.09 –0.09 –0.03 –0.10 –0.09 –0.08 –0.08 –0.05 –0.06 0.08** (–2.76) (–2.07) (–2.28) (–0.7) (–2.46) (–2.44) (–2.29) (–2.33) (–1.28) (–1.19) (2.27) Six-factor: PEAD1 –0.24 –0.20 –0.22 –0.18 –0.25 –0.20 –0.19 –0.19 –0.13 –0.10 0.14*** (–2.92) (–2.95) (–3.81) (–3.37) (–4.44) (–3.43) (–3.36) (–3.64) (–2.53) (–1.25) (3.37) Six-factor: PEAD 2 –0.26 –0.22 –0.24 –0.20 –0.27 –0.21 –0.21 –0.20 –0.14 –0.11 0.15*** (–3.17) (–3.35) (–4.28) (–3.98) (–4.98) (–3.85) (–3.74) (–3.93) (–2.75) (–1.36) (3.47) Six-factor: PEAD3 –0.23 –0.14 –0.15 –0.09 –0.16 –0.15 –0.14 –0.15 –0.09 –0.11 0.12*** (–3.88) (–3.05) (–3.51) (–1.88) (–3.44) (–3.55) (–3.33) (–3.7) (–1.94) (–1.96) (2.73) Low 2 3 4 5 6 7 8 9 High High-low A. Index-based AFP Average return 0.86 0.86 0.88 0.87 0.85 0.88 0.88 0.92 0.94 1.06 0.20** (3.05) (3.20) (3.37) (3.36) (3.24) (3.40) (3.35) (3.39) (3.42) (3.57) (2.31) CAPM $$\alpha $$ –0.12 –0.10 –0.07 –0.07 –0.10 –0.06 –0.07 –0.05 –0.03 0.06 0.18** (–1.65) (–1.78) (–1.28) (–1.55) (–2.18) (–1.35) (–1.56) (–0.88) (–0.5) (0.62) (2.13) Fama-French $$\alpha $$ –0.14 –0.12 –0.09 –0.10 –0.13 –0.09 –0.08 –0.06 –0.03 0.10 0.24*** (–2.28) (–2.53) (–1.83) (–2.5) (–3.18) (–2.22) (–2.19) (–1.33) (–0.53) (1.62) (2.78) Carhart $$\alpha $$ –0.10 –0.10 –0.07 –0.09 –0.12 –0.09 –0.08 –0.07 –0.05 0.06 0.17** (–1.69) (–2.08) (–1.47) (–2.18) (–2.87) (–2.04) (–2.22) (–1.75) (–0.99) (1.02) (2.29) Pastor-Stambaugh $$\alpha $$ –0.11 –0.11 –0.08 –0.10 –0.13 –0.09 –0.09 –0.08 –0.05 0.08 0.19** (–1.82) (–2.24) (–1.54) (–2.41) (–3) (–2.09) (–2.3) (–1.76) (–0.94) (1.22) (2.59) Six-factor: PEAD1 –0.28 –0.24 –0.24 –0.23 –0.29 –0.23 –0.19 –0.15 –0.14 0.03 0.30*** (–3.3) (–3.64) (–4.31) (–4.17) (–5.36) (–4.38) (–3.63) (–2.34) (–1.96) (0.27) (2.92) Six-factor: PEAD 2 –0.30 –0.26 –0.26 –0.24 –0.30 –0.25 –0.20 –0.16 –0.15 0.02 0.32*** (–3.63) (–4.04) (–4.82) (–4.56) (–5.9) (–4.92) (–4.04) (–2.49) (–2.12) (0.18) (2.92) Six-factor: PEAD3 –0.21 –0.17 –0.14 –0.17 –0.19 –0.14 –0.13 –0.12 –0.11 0.01 0.23*** (–3.51) (–3.37) (–2.92) (–3.92) (–4.22) (–3.29) (–3.55) (–2.8) (–2.25) (0.20) (3.09) B. Trade-based AFP Average return 0.88 0.91 0.90 0.94 0.89 0.89 0.90 0.91 0.95 0.95 0.07** (3.22) (3.58) (3.69) (3.92) (3.66) (3.73) (3.70) (3.69) (3.75) (3.49) (1.98) CAPM $$\alpha $$ –0.14 –0.07 –0.06 0.00 –0.07 –0.05 –0.05 –0.05 –0.03 –0.06 0.08** (–1.97) (–1.52) (–1.26) (–0.08) (–1.46) (–1.28) (–1.25) (–1.21) (–0.6) (–0.86) (2.09) Fama-French $$\alpha $$ –0.12 –0.08 –0.09 –0.04 –0.10 –0.08 –0.08 –0.07 –0.03 –0.02 0.10*** (–2.65) (–2.22) (–2.51) (–1.08) (–2.71) (–2.42) (–2.19) (–1.96) (–0.76) (–0.52) (2.76) Carhart $$\alpha $$ –0.14 –0.08 –0.08 –0.02 –0.09 –0.07 –0.07 –0.07 –0.04 –0.06 0.08** (–2.9) (–1.9) (–2.13) (–0.44) (–2.3) (–2.05) (–2.03) (–1.95) (–1.1) (–1.29) (2.25) Pastor-Stambaugh $$\alpha $$ –0.14 –0.09 –0.09 –0.03 –0.10 –0.09 –0.08 –0.08 –0.05 –0.06 0.08** (–2.76) (–2.07) (–2.28) (–0.7) (–2.46) (–2.44) (–2.29) (–2.33) (–1.28) (–1.19) (2.27) Six-factor: PEAD1 –0.24 –0.20 –0.22 –0.18 –0.25 –0.20 –0.19 –0.19 –0.13 –0.10 0.14*** (–2.92) (–2.95) (–3.81) (–3.37) (–4.44) (–3.43) (–3.36) (–3.64) (–2.53) (–1.25) (3.37) Six-factor: PEAD 2 –0.26 –0.22 –0.24 –0.20 –0.27 –0.21 –0.21 –0.20 –0.14 –0.11 0.15*** (–3.17) (–3.35) (–4.28) (–3.98) (–4.98) (–3.85) (–3.74) (–3.93) (–2.75) (–1.36) (3.47) Six-factor: PEAD3 –0.23 –0.14 –0.15 –0.09 –0.16 –0.15 –0.14 –0.15 –0.09 –0.11 0.12*** (–3.88) (–3.05) (–3.51) (–1.88) (–3.44) (–3.55) (–3.33) (–3.7) (–1.94) (–1.96) (2.73) This table presents the performance of decile fund portfolios formed according to their index-based (panel A) and trade-based (panel B) AFP. Like in Equation (3), we compute values using subsequent residual earnings announcement returns defined as residuals from cross-sectional regressions of 3-day abnormal returns during earnings announcements on firm size, book-to-market, past 12-month return, and the 3-day abnormal returns on earnings announcement days in the previous quarter. We form and rebalance the decile portfolios at the end of 2 months after each quarter from 1984Q1 to 2014Q2, and the return series ranges from June 1984 to November 2014. Decile 10 is the portfolio of funds with the highest AFP value. We compute equally weighted net fund returns on the portfolios, as well as risk-adjusted returns based on the CAPM, the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model, the Pastor and Stambaugh (2003) five-factor model, and three versions of the six-factor models that include the return to a strategy that exploits the post-earnings announcement drift. We report the alphas as a monthly percentage. The $$t$$-statistics are shown in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. 6. Conclusion Here, we have proposed a new measure, AFP, to identify skilled mutual fund managers. This measure evaluates fund investment skills conditioned on the release of firms’ fundamental information. Specifically, for each fund, we examined the covariance between a fund’s portfolio weight deviation from a passive benchmark (or from previous holdings) and the underlying stock performance on days when firms release fundamental information to the public Because asset prices on information days are more likely to reflect firm fundamentals, AFP can have more power in identifying investment skills. We estimate AFP using stock performance during a short window around earnings announcements. Analyzing 2,538 actively managed U.S. equity funds over the period 1984–2014, we find a positive skill (AFP) for an average mutual fund manager. Perhaps most important, we find that AFP predicts future fund performance. For instance, funds in the top decile with high index-based AFP outperform those with low AFP by 2.76% in terms of raw returns and 2.40% in terms of Carhart’s four-factor alpha. The performance difference cannot be explained by risk adjustments, controls for liquidity, post-earnings announcement drifts, or other fund characteristics. We also consider the performance predictability of AFP in relation to a number of skill measures proposed in the mutual fund literature including CS, the return gap, RPI, and active share. We find that the performance predictability of AFP is distinct, and not subsumed by the existing measures. We also find that whereas these alternative measures of skill tend to excel in forecasting future fund performance along either cross-sectional or time-series dimensions, AFP, at least in our sample period, appears to excel along both dimensions. Thus, our study illustrates the promise of performance evaluation conditional on information release. Recent studies in asset pricing have made substantial progress by focusing the patterns of asset prices on news release days. Further exploring the implications of these studies for performance evaluation would be a worthwhile avenue for continued research. We thank Andrew Karolyi (editor) and two anonymous referees for helpful comments and suggestions. We also thank Massimo Massa, Pedro Matos, Rick Sias, and Laura Starks and participants at the 2015 American Finance Association Conference, Hong Kong University of Science and Technology Finance Symposium, China International Conference in Finance, Professional Asset Management Conference (Erasmus University), Chinese University of Hong Kong, Michigan State University, Purdue University, Shanghai Advanced Institute of Finance, and University of Las Vegas. This paper supersedes our previous manuscript titled “Identifying skilled mutual fund managers by their ability to forecast earnings.” Supplementary data can be found on The Review of Financial Studies Web site. Footnotes 1 For example, Fama and French (2010) argue that if the cross-section of mutual fund alphas has a normal distribution with mean zero, then a cross-sectional standard deviation of 1.25% per year, or 0.10% per month, captures the tails of the cross-section of alpha estimates. For our sample of funds from 1984 to 2014, the time-series standard deviation of alpha is 1.96% (1.87%) per month for the Fama and French three-factor (Carhart four-factor) model. Therefore, to observe a statistically significant alpha with a $$t$$-statistic of at least 1.96 for a truly skilled fund manager endowed with an alpha one standard deviation above average, the performance evaluator would need more than 100 years of return history. 2 One strand in this literature estimates alpha values using fund returns and documents that mutual funds, on average, underperform passive benchmarks (e.g., Jensen 1968; Malkiel 1995; Gruber 1996; Carhart 1997; Fama and French 2010). Another strand examines the portfolio holdings of mutual funds to study managers’ investment abilities (e.g., Grinblatt and Titman 1989, 1993; Daniel et al. 1997; Wermers 2000). More recent literature also suggests that some active managers consistently deliver positive returns, despite the average underperformance (e.g., Chevalier and Ellison 1999; Cohen, Coval, and Pastor 2005; Kacperczyk, Sialm and Zheng 2005, 2008; Kacperczyk and Seru 2007; Cremers and Petajisto 2009; Barras, Scaillet and Wermers 2010; Amihud and Goyenko 2013). 3 For the trade-based AFP, we used changes in portfolio weights over both one and four quarters. The two measures generate similar results. To be consistent with the previous literature (e.g., Grinblatt and Titman 1993; Daniel et al. 1997), we report the results based on fund trades in the past four quarters. 4 On an annualized basis (multiplied by 250/3), the mean index-based (trade-based) AFP translates into an alpha of 8.57% (4.41%). 5Ali et al. (2014) find that mutual funds as a group tend to trade on the PEAD; however, because of competition, mutual funds aggressively pursuing this strategy on average fail to outperform, consistent with our findings. 6 We also considered the industry concentration index (ICI) measure as proposed by Kacperczyk, Sialm and Zheng (2005). Their measure shares a conceptual similarity with active share in emphasizing how active mutual fund managers are in their investment decisions. Our results on AFP remain similar when we use the ICI to replace active share or when we use the ICI in combination with active share. 7 The insignificant forecasting power of CS, RPI, and active share does not contradict the original findings. The results on active share are consistent with those in Table 8 of Cremers and Petajisto (2009, p. 3356), who show that the spread in returns between high and low active share funds is statistically indistinguishable from zero. A positive spread is documented in Table 4 of their paper, where the authors use benchmark-adjusted returns to measure fund performance, and they show in a subsequent paper (Cremers, Petajisto, and Zitzewitz 2012) that benchmarks contain alpha. Regarding RPI, Kacperczyk and Seru (2007) show in a panel regression setting that RPI is negatively related to fund performance, a result that we confirm in Table 5 and will discuss later. However, Kacperczyk and Seru did not report results using portfolio sorts. The study of Daniel et al. (1997) is primarily concerned with attributing realized fund performance to different components of skill; they intended to explain fund returns, but did not predict fund performance. 8 The performance forecasting power of AFP remains statistically significant when we cluster the standard errors by both fund and time. References Ali, A., Chen, X. Yao, T. and Yu. T. 2014 . Mutual fund competition and profiting from the post earnings announcement drift. Unpublished Working Paper . Ali, A., Durtschi, C. Lev, B. and Trombley. M. 2004 . Changes in institutional ownership and subsequent earnings announcement abnormal returns. Journal of Accounting, Auditing, and Finance 3 : 221 – 48 . Google Scholar CrossRef Search ADS Amihud, Y., and Goyenko. R. 2013 . Mutual fund’s $$R^{\mathrm{2}}$$ as predictor of performance. Review of Financial Studies 26 : 667 – 94 . Google Scholar CrossRef Search ADS Baker, M., Litov, L. Wachter, J. A. and Wurgler. J. 2010 . Can mutual fund managers pick stocks? Evidence from their trades prior to earnings announcements. Journal of Financial and Quantitative Analysis 45 : 1111 – 31 . Google Scholar CrossRef Search ADS Ball, R., and Brown. P. 1968 . An empirical evaluation of accounting income numbers. Journal of Accounting Research 6 : 159 – 78 . Google Scholar CrossRef Search ADS Barras, L., Scaillet, O. and Wermers. R. 2010 . False discoveries in mutual fund performance: Measuring luck in estimated alphas. Journal of Finance 65 : 179 – 216 . Google Scholar CrossRef Search ADS Bernard, V. L., and Thomas. J. K. 1989 . Post-earnings-announcement drift: Delayed price response or risk premium? Journal of Accounting Research 27 : 1 – 36 . Google Scholar CrossRef Search ADS Bernard, V. L., and Thomas. J. K. 1990 . Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics 13 : 305 – 40 . Google Scholar CrossRef Search ADS Brown, S. J., and Goetzmann. W. N. 1995 . Performance persistence. Journal of Finance 50 : 679 – 98 . Google Scholar CrossRef Search ADS Carhart, M. M. 1997 . On persistence in mutual fund performance. Journal of Finance 52 : 57 – 82 . Google Scholar CrossRef Search ADS Chen, J., Hong, H. Huang, M. and Kubik. J. 2004 . Does fund size erode performance? Liquidity, organizational diseconomies and active money management. American Economic Review 94 : 1276 – 302 . Google Scholar CrossRef Search ADS Chevalier, J., and Ellison. G. 1999 . Are some mutual fund managers better than others? Cross-sectional patterns in behavior and performance. Journal of Finance 54 : 875 – 99 . Google Scholar CrossRef Search ADS Cohen, L., Frazzini, A. and Malloy. C. 2008 . The small world of investing: Board connections and mutual fund returns. Journal of Political Economy 116 : 951 – 79 . Google Scholar CrossRef Search ADS Cohen, R. B., Coval, J. D. and Pastor. L. 2005 . Judging fund managers by the company that they keep. Journal of Finance 60 : 1057 – 96 . Google Scholar CrossRef Search ADS Cremers, M., Ferreira, M. Matos, P. and Starks. L. 2013 . The mutual fund industry worldwide: Explicit and closet indexing, fees, and performance. Working Paper. Cremers, M., and Petajisto. A. 2009 . How active is your fund manager? A new measure that predicts performance. Review of Financial Studies 22 : 3329 – 65 . Google Scholar CrossRef Search ADS Cremers, M., Petajisto, A. and Zitzewitz. E. 2012 . Should benchmark indices have alpha? Revisiting performance evaluation. Critical Finance Review 2 : 1 – 48 . Google Scholar CrossRef Search ADS Daniel, K., Grinblatt, M. Titman, S. and Wermers. R. 1997 . Measuring mutual fund performance with characteristic-based benchmarks. Journal of Finance 52 : 1035 – 58 . Google Scholar CrossRef Search ADS Engelberg, J., McLean, D. and Pontiff. J. 2015 . Anomalies and news. Working Paper. Engelberg, J., Reed, A. and Ringgenberg. M. 2012 . How are shorts informed? Short sellers, news, and information processing. Journal of Financial Economics 105 : 260 – 78 . Google Scholar CrossRef Search ADS Elton, E. J., Gruber, M. J. and Blake. C. R. 1996 . The persistence of risk-adjusted mutual fund performance. Journal of Business 69 : 133 – 57 . Google Scholar CrossRef Search ADS Fama, E. F., and French. K. R. 1993 . Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33 : 3 – 56 . Google Scholar CrossRef Search ADS Fama, E. F., and French. K. R. 2010 . Luck versus skill in the cross section of mutual fund returns. Journal of Finance 65 : 1915 – 47 . Google Scholar CrossRef Search ADS Fama, E. F., and Macbeth. J. D. 1973 . Risk and return: Empirical tests. Journal of Political Economy 81 : 607 – 36 . Google Scholar CrossRef Search ADS Fang, L. H., Peress, J. and Zheng. L. 2014 . Does media coverage of stocks affect mutual funds’ trading and performance? Review of Financial Studies 27 : 3441 – 66 . Google Scholar CrossRef Search ADS Grinblatt, M., Keloharju, M. and Linnainmaa. J. 2011 . IQ, trading behavior, and performance. Journal of Financial Economics 104 : 339 – 62 . Google Scholar CrossRef Search ADS Grinblatt, M., and Titman. S. 1989 . Mutual fund performance: An analysis of quarterly portfolio holdings. Journal of Business 62 : 393 – 416 . Google Scholar CrossRef Search ADS Grinblatt, M., and Titman. S. 1993 . Performance measurement without benchmarks: An examination of mutual fund returns. Journal of Business 66 : 47 – 68 . Google Scholar CrossRef Search ADS Gruber, M. J. 1996 . Another puzzle: The growth in actively managed mutual funds. Journal of Finance 51 : 783 – 810 . Google Scholar CrossRef Search ADS Jegadeesh, N., and Titman. S. 1993 . Returns to buying winners and selling losers: Implications for stock market efficiency. Journal of Finance 48 : 65 – 91 . Google Scholar CrossRef Search ADS Jensen, M. C. 1968 . The performance of mutual funds in the period 1945–1964. Journal of Finance 23 : 389 – 416 . Google Scholar CrossRef Search ADS Kacperczyk, M., and Seru. A. 2007 . Fund manager use of public information: New evidence on managerial skills. Journal of Finance 62 : 485 – 528 . Google Scholar CrossRef Search ADS Kacperczyk, M., Sialm, C. and Zheng. L. 2005 . On the industry concentration of actively managed equity mutual funds. Journal of Finance 60 : 1983 – 2011 . Google Scholar CrossRef Search ADS Kacperczyk, M., Sialm, C. and Zheng. L. 2008 . Unobserved actions of mutual funds. Review of Financial Studies 21 : 2379 – 416 . Google Scholar CrossRef Search ADS Kacperczyk, M., Van Nieuwerburgh, S. and Veldkamp. L. 2014 . Time-varying fund manager skill. Journal of Finance 69 : 1455 – 84 . Google Scholar CrossRef Search ADS La Porta, R.l, Lakonishok, J. Shleifer, A. and Vishny. R. 1997 . Good news for value stocks: Further evidence on market efficiency. Journal of Finance 52 : 859 – 74 . Google Scholar CrossRef Search ADS Lakonishok, J., Shleifer, A. and Vishny. R. W. 1994 . Contrarian investment, extrapolation, and risk. Journal of Finance 49 : 1541 – 78 . Google Scholar CrossRef Search ADS Livnat, J., and Mendenhall. R. R. 2006 . Comparing the post-earnings-announcement drift for surprises calculated from analyst and time series forecasts. Journal of Accounting Research 44 : 177 – 205 . Google Scholar CrossRef Search ADS Lo, A. 2008 . Where do alphas come from? A measure of the value of active investment management. Journal of Investment Management 6 – 29 . Lou, D. 2012 . A flow-based explanation for return predictability. Review of Financial Studies 25 : 3457 – 89 . Google Scholar CrossRef Search ADS Lucca, D. O., and Moench. E. 2015 . The pre-FOMC announcement drift. Journal of Finance 70 : 329 – 37 . Google Scholar CrossRef Search ADS Malkiel, B. G. 1995 . Returns from investing in equity mutual funds 1971–1991. Journal of Finance 50 : 549 – 72 . Google Scholar CrossRef Search ADS Newey, W. K., and West. K. D. 1987 . A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55 : 703 – 8 . Google Scholar CrossRef Search ADS Pastor, L., and Stambaugh. R. F. 2003 . Liquidity risk and expected stock returns. Journal of Political Economy 113 : 642 – 85 . Google Scholar CrossRef Search ADS Petajisto, A. 2013 . Active share and mutual fund performance. Financial Analysts Journal 69 : 73 – 93 . Google Scholar CrossRef Search ADS Petersen, M. 2009 . Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies 22 : 435 – 80 . Google Scholar CrossRef Search ADS Savor, P., and Wilson. M. 2013 . How much do investors care about macroeconomic risk? Evidence from scheduled economic announcements. Journal of Financial and Quantitative Analysis 48 : 343 – 75 . Google Scholar CrossRef Search ADS Savor, P., and Wilson. M. 2014 . Asset pricing: A tale of two days. Journal of Financial Economics 113 : 171 – 201 . Google Scholar CrossRef Search ADS Savor, P., and Wilson. M. 2016 . Earnings announcements and systematic risk. Journal of Finance 71 : 83 – 138 . Google Scholar CrossRef Search ADS Sloan, R. 1996 . Do stock prices fully reflect information in accruals and cash flows about future earnings? Accounting Review 71 : 289 – 315 . Stambaugh, R. F. 2014 . Presidential address: Investment noise and trends. Journal of Finance 69 : 1415 – 53 . Google Scholar CrossRef Search ADS Vayanos, D., and Woolley. P. 2013 . An institutional theory of momentum and reversal. Review of Financial Studies 26 : 1087 – 145 . Google Scholar CrossRef Search ADS Wermers, R. 2000 . Mutual fund performance: An empirical decomposition into stock-picking talent, style, transaction costs, and expenses. Journal of Finance 55 : 1655 – 95 . Google Scholar CrossRef Search ADS Zheng, L. 1999 . Is money smart? A study of mutual fund investors’ fund selection ability. Journal of Finance 54 : 901 – 33 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Review of Financial Studies Oxford University Press

Active Fundamental Performance

Loading next page...
 
/lp/ou_press/active-fundamental-performance-18vQ0K0uAw
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.
ISSN
0893-9454
eISSN
1465-7368
D.O.I.
10.1093/rfs/hhy020
Publisher site
See Article on Publisher Site

Abstract

Abstract We propose a new measure, active fundamental performance (AFP), to identify skilled mutual fund managers. AFP evaluates fund investment skills conditioned on the release of firms’ fundamental information. For each fund, we examine the covariance between deviations of its portfolio weights from a benchmark portfolio and the underlying stock performance on days when firms publicize fundamental information. Because asset prices on these information days better reflect firm fundamentals, AFP can more effectively identify investment skills. From 1984 to 2014, funds in the top decile of high AFP subsequently outperformed those in the bottom decile by 2% to 3% per year. Received August 25, 2016; editorial decision December 17, 2017 by Editor Andrew Karolyi. Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online. Despite a large body of literature examining performance evaluation, identifying skilled mutual fund managers remains challenging. Noise and random shocks to asset returns make it difficult to differentiate skills from luck. Because observed mutual fund alphas typically are small and noisy, an evaluator would need an unfeasibly long return series to identify a skilled manager reliably.1 Recent studies show that asset prices reflect fundamental information to a greater extent on news release days: the risk-return relationship tends to manifest more clearly, and stock-level mispricing tends to be corrected with the release of important fundamental information (e.g., Savor and Wilson 2013, 2014, 2016; Lucca and Moench 2015; Engelberg, McLean, and Pontiff 2015). Such findings have important implications for performance evaluation. When stock prices are more informative of their fundamental values, fund performance is more indicative of the manager’s investment skill. However, no existing performance measures have exploited this window of opportunity. With the current study, we intend to fill this gap. We propose a measure of fund investment skills conditioned on the release of firm fundamental information. We focus on price movements driven by fundamental information, and our measure uses available fund performance and stock return data more efficiently to increase the power of performance evaluation. Specifically, the proposed performance measure, which we refer to as “active fundamental performance” (AFP), captures the covariance between a fund’s portfolio weight deviation from a passive benchmark (or from previous holdings) and the stocks’ performance during a 3-day window surrounding subsequent earnings announcements. We choose this short window as a proxy for the release of fundamental information because earnings announcements, as major company information events, are associated with a substantial “correction” in stock prices. For instance, Sloan (1996) documents that approximately 40% of the profits from accrual strategies cluster in a 3-day earnings announcement window. La Porta et al. (1997) report that between 25% and 30% of the returns to various value strategies considered by Lakonishok, Shleifer, and Vishny (1994) are concentrated on the 3 days around earnings announcements. Jegadeesh and Titman (1993) estimate that approximately 25% of momentum profits are concentrated on the 3 days surrounding earnings announcements. More recently, Engelberg, McLean, and Pontiff (2015) investigate 97 stock return anomalies and find that anomaly returns are 7 times higher on earnings announcement days. This evidence suggests that the short window around earnings announcements is a period in which stock prices converge to fundamental values. Therefore, we posit that a fund’s underlying stock performance around earnings announcements can be particularly revealing in identifying active managers skilled in selecting mispriced securities and successfully predicting the correction in their prices. We analyze quarterly holdings data for 2,538 unique, actively managed U.S. equity funds over the period 1984–2014. For each fund in each quarter, we compute the index-based AFP (the covariance between a fund’s active weights, or portfolio weight deviations from its benchmark index, and the underlying stocks’ performance during subsequent earnings announcements) and the trade-based AFP (the covariance between changes in a fund’s portfolio weights and the underlying stocks’ performance during subsequent earnings announcements). We find that the index-based (trade-based) AFP measure is positive on average, with a cross-fund mean of 9.95 (5.40) basis points (bps) per 3-day window and a standard deviation of 32.46 (22.09) bps. These results suggest substantial cross-sectional heterogeneity in mutual funds’ performance; at least some managers exhibit skill. In addition, the measure shows a positive, albeit only moderate, correlation with other commonly used performance measures. For example, the index-based (trade-based) AFP has average cross-sectional correlations of 25%, 24%, 32%, and 17% (6%, 6%, 8%, and 23%) with raw fund returns, the four-factor alpha, Daniel et al.’s (1997) characteristic selectivity (CS) measure, and the Grinblatt and Titman (1993) measure, respectively. Thus, compared with other performance measures, AFP captures unique fund characteristics and substantial incremental information. We find that AFP strongly predicts subsequent fund performance. In univariate sorts by index-based (trade-based) AFP, mutual funds in the top decile with the highest AFP outperform those in the bottom decile with the lowest AFP by 2.76% (1.44%) per year. This outperformance cannot be accounted for by their different exposures to risk or style factors. For example, after adjusting for their differential loadings on the market, size, value, and momentum factors, mutual funds in the top decile with the highest index-based (trade-based) AFP continue to outperform those in the bottom decile by 2.40% (1.44%) per year. In other tests, we control for the effects of liquidity and post-earnings announcement drifts, and AFP remains a strong predictor of future fund performance. Note that our fund portfolio strategy is based on information about fund holdings lagged by at least 2 months. The U.S. Securities and Exchange Commission (SEC) requires mutual funds to disclose their portfolio composition within 2 months, making this strategy implementable for mutual fund investors or funds of mutual funds aiming to improve their selection performance. In double sorts, we find that the strong predictive power of AFP for future fund returns is incremental beyond that of various return- and holding-based performance measures, such as CS (Daniel et al. 1997), return gap (Kacperczyk, Sialm and Zheng 2008), reliance on public information (RPI; Kacperczyk and Seru 2007), and active share (Cremers and Petajisto 2009). We also find that mutual fund investors can further improve their returns if they combine the information contained in AFP with signals from return gap. For example, mutual funds in the top quartiles of the index-based AFP and return gap outperform those in the bottom quartiles by 3.36% per year based on the four-factor model. We also consider multiple regressions that jointly control for alternative measures of fund investment skill such as CS, return gap, RPI, and active share, together with other fund characteristics, including fund age, size, expense ratios, turnover, past flow, and past performance. We find that AFP remains statistically significant in predicting future fund performance in these regressions. In both the portfolio and the regression analyses, AFP significantly predicts future fund performance more consistently than the alternative performance measures. We conduct further analyses to shed light on the information sources for AFP. Using stock-level analyses, we find that both active weights and changes in portfolio weights can predict abnormal return around future earnings announcements and exhibit stronger predictability for companies with more information asymmetry (i.e., those with lower analyst coverage, higher idiosyncratic volatilities, and higher analyst disagreement). Moreover, high-AFP funds’ superior performance is more pronounced when financial analysts’ forecasts diverge. These results support the notion that the ability to form a superior estimate of firms’ fundamental values drives AFP and its performance forecasting power. The AFP measures also exhibit time-series persistence; for instance, mutual funds in the top decile with the highest index-based AFP continue to exhibit significantly higher AFP than those in the bottom decile in the subsequent 2 years. This persistence is largely due to the superior AFP of skilled funds in the top decile. The current study contributes to the broad literature on mutual fund performance and market efficiency by introducing a new performance measure as well as a new and more efficient way of using available data.2 We show that focusing on important data points (information days) and integrating information on the extent and quality of active management into a single measure yield substantially more power in identifying skilled managers. The AFP measure is related to Grinblatt and Titman’s (1993) benchmark-free performance measure and Daniel et al.’s (1997) characteristic-based performance measure, which are based on fund holdings and subsequent stock returns. Our innovation of conditioning on a fundamental information release goes further and exploits the high information-to-noise ratio during the time period in which prices tend to converge to their fundamental values. The conditional approach sharpens signals from subsequent stock performance and substantially improves the power to evaluate stock-picking skills. In unreported tests, we find that when we replace earnings announcement returns with stock returns in the entire subsequent quarter, AFP’s forecasting power for future fund performance disappears. Our study has implications for the growing literature that quantifies the level of activeness in a managed portfolio (e.g., Cremers and Petajisto 2009; Cremers et al. 2013; Petajisto 2013; Stambaugh 2014). We contribute to this literature stream by nesting the degree of activeness into a tight framework for performance evaluation. Our approach focuses on signed active portfolio weights (or changes in weights) and examines the covariance between the weights and the underlying stock returns on information days. Empirically, we show that the AFP measure helps separate skilled versus unskilled active funds with high active shares, which makes it particularly useful for mutual fund investors. Ali et al. (2004) and Baker et al. (2010) document evidence that trades by aggregate institutional investors and aggregate mutual funds forecast subsequent earnings surprises, respectively. Cohen, Frazzini, and Malloy (2008) track the shared educational network between mutual fund and corporate managers and find that connected stock holdings achieve high average returns, particularly during earnings announcements. In combination with studies documenting earnings announcements as the short period when stock prices quickly converge to the fundamental values, these papers provide a useful foundation for us to create a powerful fund-level performance measure that identifies skilled fund managers. Although the current study is closely related to Baker et al. (2010), it differs in several key aspects. First, whereas Baker et al. (2010) focus on whether mutual fund managers on average can pick stocks, we investigate how to identify individual managers who can select stocks. Second, in terms of empirical tests, Baker et al. (2010) focus on stock-level analyses, assessing earnings announcement returns of all fund trades or trades of a certain group of funds. In contrast, the current study focuses on fund-level analyses by proposing a new fund performance measure. Third, the emphasis in Baker et al. (2010) is on the ability of fund managers to forecast a firm’s future earnings number. An important contribution of our study is to introduce the notion of performance evaluation, conditional on the release of fundamental information. We use earnings announcements as a proxy and starting point, but our measure has broad implications and can include other fundamental information events. Another literature stream examines how investors use publicly available information and explores the potential performance implications. For instance, Kacperczyk and Seru (2007) examine mutual fund managers’ reliance on public information as reflected in analyst stock recommendations, and Fang, Peress, and Zheng (2014) investigate their reliance on mass media. Both studies show that relying on public information has negative consequences for mutual fund managers’ performance. In a different context, Engelberg, Reed, and Ringgenberg (2012) investigate the source of short sellers’ skills and show that short sellers have a superior ability to interpret and process publicly available information, which translates into investment skills. In contrast with these studies focusing on how fund managers make use of public information, which is backward looking, we evaluate fund performance conditional on the release of fundamental information with a forward-looking focus. The investment ability of fund managers includes but is not limited to their ability to process public information. The key to our identification is that the realization of superior fundamental performance tends to be catalyzed when important firm-level information such as earnings news is released to the public domain. 1. Methodology: AFP In this section, we develop our new measure of performance evaluation, AFP. The starting point is the covariance between portfolio weights and future asset returns, Cov($$w_{i,t}$$,$$ R_{i,t})$$, as shown in Equation (1). As Grinblatt and Titman (1993) and Lo (2008) point out, aggregating the covariance over all investments is an intuitive way to examine managers’ skill to forecast asset returns. The portfolio weights reflect managers’ strategic decisions to buy, sell, or avoid an asset in anticipation of its future returns. A skilled manager whose percentage holdings of assets increase in future asset returns will on average exhibit a positive covariance. The covariance measure is theoretically appealing, in that it captures fund performance due to active portfolio management (Grinblatt and Titman 1993; Lo 2008): \begin{align} &\underbrace {\sum\limits_{i=1}^N {Cov(w_{i,t} ,R_{i,t} )} }_{Active}=\sum\limits_{i=1}^N {E(w_{i,t} \times R_{i,t} )} -\sum\limits_{i=1}^N {E(w_{i,t} )\times E(R_{i,t} )}\notag \\ &\quad=\underbrace {E(R_{p,t} )}_{Total}-\underbrace {\sum\limits_{i=1}^N {E(w_{i,t} )\times E(R_{i,t} )} }_{Passive}, \end{align} (1) where $$w_{i,t}$$ is the weight of security $$i$$ in the fund’s portfolio at the beginning of period $$t$$, $$R_{i,t}$$ is security $$i$$’s return during period $$t$$, and $$N$$ is the number of securities in the fund’s portfolio. Equation (1) shows that the sum of the covariances across securities captures the difference between total portfolio returns $$R_{p,t}$$ and the passive return to the portfolio. We propose two innovations to the covariance measure. Motivated by recent studies on information releases and asset prices (see, e.g., Savor and Wilson 2013, 2014, 2016; Lucca and Moench 2015; Engelberg, McLean, and Pontiff 2015), we measure future asset returns within a short window during which price movements are driven by firm fundamental information. This approach helps mitigate the influence of noise in asset returns and increases the power of the covariance measure to identify investment skills. Specifically, we examine stock returns during 3-day windows around earnings announcements to focus on time periods during which stock prices tend to reflect this fundamental information. The advantage of focusing on earnings announcement returns lies in the high information-to-noise ratio on earnings announcement days and the large cross-section of earnings announcement events, both of which increase the power of performance evaluation. A disadvantage arises because firms release earnings numbers relatively infrequently, typically once per quarter. Although we choose a 3-day event window—a standard practice in event studies of earnings announcements—our results also are robust to using a 5-day event window. Our second innovation is to use benchmark-adjusted portfolio weights, instead of raw portfolio weights, to capture a manager’s active investment decisions. Cremers and Petajisto (2009) pioneered the idea of using benchmark-adjusted portfolio weights to measure a fund manager’s active investment decisions. We provide a theoretical illustration of this approach in Internet Appendix A. Because an actively managed mutual fund’s performance is typically benchmarked against a passive index, the passive index is a natural choice for the fund’s benchmark portfolio. Alternatively, we could use the fund’s own past holdings as a benchmark, thereby using changes in portfolio weights (i.e., fund trades) to compute AFP. Active weights, or deviations in portfolio weights from the benchmark, contain the fund manager’s information revealed through both recent and longer-term holdings. Changes in portfolio weights reflect recent portfolio adjustments and thus recent information about security returns. Ex ante, both active weights and changes in portfolio weights can be useful for assessing the manager’s skill, though the former may be more comprehensive in capturing the manager’s information set. Herein, we compute both index-based and trade-based AFP, and we report both sets of results. As Equation (2) indicates, our covariance measure reflects the expected abnormal fund returns during earnings announcements, attributable to the fund’s active portfolio decisions: \begin{align} \textit{Cov}(w_{i,t} - w_{i,t}^b, \textit{CAR}_{i,t}) &= E[(w_{i,t} - w_t^b) \times \textit{CAR}_{i,t} ] - E(w_{i,t} - w_{i,t}^b) \times E(\textit{CAR}_{i,t})\notag\\ &= E[(w_{i,t} - w_{i,t}^b) \times \textit{CAR}_{i,t} ] - 0 \times E(\textit{CAR}_{i,t})\notag\\ &= E[(w_{i,t} - w_{i,t}^b) \times \textit{CAR}_{i,t} ]. \end{align} (2) We develop an empirical analog of the sum of this covariance measure across securities, termed AFP, in Equation (3): \begin{align} \textit{AFP}_{j,t} = \sum\limits_{i=1}^{N_j} (w_{i,t}^j - w_{i,t}^{b_j})\textit{CAR}_{i,t}, \end{align} (3) where $$\textit{AFP}_{j,t}$$ is mutual fund $$j$$’s AFP based on its portfolio selection in quarter $$t$$, $${\boldsymbol{w}_{\boldsymbol{i,t}}^{\boldsymbol{j}}}$$ is the weight of stock $$i$$ in fund $$j$$’s portfolio at the start of quarter $$t$$, $${\boldsymbol{w}_{\boldsymbol{i,t}}^{\boldsymbol{b}_{\boldsymbol{j}}}}$$ is the weight of stock $$i$$ in fund $$j$$’s benchmark portfolio at the start of quarter $$t$$, $$\textit{CAR}_{i,t}$$ is stock $$i$$’s 3-day cumulative abnormal return surrounding its quarterly earnings announcement during quarter $$t$$, and $$N_{j}$$ is the number of stocks in fund $$j$$’s investment universe (i.e., the union of stocks held by the fund and those in the fund’s benchmark portfolio). The daily abnormal returns refer to the difference in daily returns between a stock and its size and book-to-market matched portfolio. We sum the daily abnormal returns from 1 day before to 1 day after earnings announcements to obtain the 3-day abnormal return. 2. Computing AFP 2.1 Sample construction We obtain the portfolio holdings for actively managed equity mutual funds from Thomson Reuters’s CDA/Spectrum Mutual Fund Holdings Database and returns for the individual mutual funds and other fund characteristics from the Center for Research in Security Prices (CRSP) Survivor-Bias-Free U.S. Mutual Fund Database. We then use the MFLINKS data set to merge the two databases, excluding balanced, bond, money market, international, index, and sector funds, as well as funds not invested primarily in equity securities. After applying this filter, the sample consists of 2,538 unique funds, ranging in time from the first quarter of 1984 to the second quarter of 2014. To select the benchmark index for fund managers, we follow Cremers and Petajisto’s (2009) procedure. The universe of benchmark indexes includes 19 widely used benchmark indexes: S&P 500, S&P 400, S&P 600, S&P 500/Barra Value, S&P 500/Barra Growth, Russell 1000, Russell 2000, Russell 3000, Russell Midcap, value and growth variants of the four Russell indexes, Wilshire 5000, and Wilshire 4500. For each fund in each quarter, we select an index that minimizes the average distance between the fund portfolio weights and the benchmark index weights. Data on the index holdings of the 12 Russell indexes since their inception come from the Frank Russell Company, and data on the S&P 500, S&P 400, and S&P 600 index holdings since December 1994 are provided by COMPUSTAT. For the remaining indexes and periods, we use the index funds holdings to approximate index holdings. The information on daily stock prices and returns for common stocks traded on the NYSE, AMEX, and NASDAQ comes from the CRSP daily stock files. We obtain firms’ announcement dates for quarterly earnings from COMPUSTAT and analysts’ consensus earnings forecasts from I/B/E/S. Panel A of Table 1 shows the summary statistics for mutual funds in our sample. An average fund in our sample manages $\$$ 1.39 billion of assets and is 15.70 years old. Their mutual fund investors achieve an average return of 2.39% per quarter. The net percentage fund flow is skewed to the right; the quarterly fund flow has a mean of 1.35% but a median of only $$-$$1.05%. On average, mutual funds in our sample incur an annual expense ratio of 1.23% and turn over their portfolios by 86.08% per year. These numbers are in line with the previous literature. Table 1 Descriptive statistics A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 This table presents descriptive statistics for our sample of mutual funds. The sample consists of 2,538 distinct mutual funds from 1984Q1 to 2014Q2. Panel A presents the summary statistics for fund characteristics. TNA is the quarter-end total net fund assets in millions of dollars; age is the fund age in years; quarterly return is the quarterly net fund return as a percentage; flow is the quarterly growth rate of assets under management as a percentage after adjusting for the appreciation of the fund’s assets; expense is the fund expense ratio as a percentage, and turnover is the turnover ratio of the fund as a percentage. Panel B shows the time-series average of the cross-sectional Spearman correlation coefficients for the variables of interest. Table 1 Descriptive statistics A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 A. Summary statistics of fund characteristics Mean SD 25th pctl Median 75th pctl Total number of funds 2,538 TNA ( $\$$ million) 1,391.33 5,391.59 72.30 246.08 883.23 Age (years) 15.70 14.55 6.00 11.00 19.00 Quarterly return (%) 2.39 10.24 –2.60 3.16 8.52 Flow (%) 1.35 13.79 –4.11 –1.05 3.37 Expense (%) 1.23 0.46 0.95 1.19 1.48 Turnover (%) 86.08 101.58 34.00 63.00 109.00 B. Average Spearman cross-sectional correlation coefficients TNA Age Quarterly return Flow Expense Age 0.29 Quarterly return 0.01 –0.01 Flow –0.01 –0.13 0.15 Expense –0.22 –0.21 –0.01 0.03 Turnover –0.09 –0.11 0.02 0.05 0.19 This table presents descriptive statistics for our sample of mutual funds. The sample consists of 2,538 distinct mutual funds from 1984Q1 to 2014Q2. Panel A presents the summary statistics for fund characteristics. TNA is the quarter-end total net fund assets in millions of dollars; age is the fund age in years; quarterly return is the quarterly net fund return as a percentage; flow is the quarterly growth rate of assets under management as a percentage after adjusting for the appreciation of the fund’s assets; expense is the fund expense ratio as a percentage, and turnover is the turnover ratio of the fund as a percentage. Panel B shows the time-series average of the cross-sectional Spearman correlation coefficients for the variables of interest. Panel B of Table 1 shows the average Spearman cross-sectional correlation coefficients among fund characteristics. The results confirm our intuition: the average 29% correlation coefficient between fund size and age indicates that large funds tend to have a longer track record, and the correlation between fund size and expense ratio of $$-$$22% indicates that large funds tend to incur lower expense ratios. We also find a negative correlation of $$-$$13% between fund age and fund flow, consistent with the idea that established mutual funds with longer life spans tend to be stable, with a smaller percentage of inflows. In the next subsection, we analyze AFP. 2.2 Computing AFP For each fund in each quarter, we compute both index-based and trade-based AFP. For the trade-based AFP, we measure changes in portfolio weights and account for mechanical changes in portfolio weights due to stock price changes.3 Most earnings announcements (more than 89% in our sample) occur in the first 2 months after the quarter ends, so we use the time line in Figure 1: the active weights (changes in portfolio holdings) for fund $$j$$ are measured at the end of month $$t$$ (e.g., end of March), and the earnings announcements are observed in the first 2 months in the following quarter (e.g., April and May). We use Equation (3) to compute the AFP for fund $$j$$, or AFP$$_{j,t}$$. In the analysis of fund performance in the next section, we track the performance of fund $$j$$ for the subsequent 3 months, including months $$t+$$3, $$t+$$4, and $$t+$$5 (e.g., June–August). Figure 1 View largeDownload slide Time line for AFP This figure illustrates the measurement of AFP. In this illustration, AFP is measured at the end of May. The portfolio compositions of mutual funds and their benchmarks (lagged portfolio weights adjusted for price changes for trade-based AFP) are measured in March and observable by the end of May. The earnings announcements arrive in April and in May. Mutual fund portfolios are formed using AFP (available in real time) at the beginning of June and held until the end of August. We then recompute AFP using new information; doing so initiates portfolio rebalancing. Figure 1 View largeDownload slide Time line for AFP This figure illustrates the measurement of AFP. In this illustration, AFP is measured at the end of May. The portfolio compositions of mutual funds and their benchmarks (lagged portfolio weights adjusted for price changes for trade-based AFP) are measured in March and observable by the end of May. The earnings announcements arrive in April and in May. Mutual fund portfolios are formed using AFP (available in real time) at the beginning of June and held until the end of August. We then recompute AFP using new information; doing so initiates portfolio rebalancing. We choose this time line to balance two key considerations. On the one hand, the closer earnings announcement dates are to the reported holdings dates, the more accurate are the AFP measures due to unobserved trades by mutual funds within the quarter. On the other hand, a longer time span between the announcement dates and the reported holdings dates allows us to include more firms in the analysis. Using the first 2 months of a quarter balances these two considerations: most firms report their quarterly performance in the first 2 months of a quarter, and the time span between the reported portfolio composition and earnings announcement performance is reasonably close. Moreover, from a practical perspective, because mutual funds are required to report their holdings within 2 months, the information required to compute AFP is available to investors in real time. Thus, the trading strategy is implementable. To provide further justification for our time line, in Figure 2 we plot the average AFP value for a median fund over the 13 weeks following a typical quarter end. It indicates that for an average fund, the value of AFP stabilizes during the eighth or ninth week after the quarter ends, when we observe the fund’s portfolio composition. Thus, we conclude that incorporating earnings events that occur after the first 2 months do not improve the information content of a fund’s AFP. Figure 2 View largeDownload slide Cumulative AFP over the weeks following quarter ends This figure plots the AFP for a median mutual fund in our sample, cumulative over the weeks following quarter ends, when the active fund weights or changes in portfolio weights are measured. Like in Equation (3), AFP is the sum of the product of the fund’s active portfolio weights (panel A) or changes in portfolio weights (panel B) and subsequent 3-day abnormal returns surrounding earnings announcements. The value of cumulative AFP at the end of week 13 is scaled to equal 1. Figure 2 View largeDownload slide Cumulative AFP over the weeks following quarter ends This figure plots the AFP for a median mutual fund in our sample, cumulative over the weeks following quarter ends, when the active fund weights or changes in portfolio weights are measured. Like in Equation (3), AFP is the sum of the product of the fund’s active portfolio weights (panel A) or changes in portfolio weights (panel B) and subsequent 3-day abnormal returns surrounding earnings announcements. The value of cumulative AFP at the end of week 13 is scaled to equal 1. For an average fund in a typical quarter, the index-based (trade-based) AFP is equal to 10.28 (5.30) bps, with a standard deviation of 84.72 (97.10) bps.4 A substantial proportion of the high variability of AFP comes from cross-fund dispersion. For each fund, we compute the average AFP over its entire life. The cross-fund standard deviation for index-based (trade-based) AFP is 32.46 (22.09) bps, which is 3.26 (4.09) times the cross-section mean of 9.95 (5.40) bps. This high cross-fund dispersion in AFP is the main focus of the current research. 3. Predicting Mutual Fund Performance by AFP 3.1 Portfolio sorts Using a portfolio-based analysis, we examine the profitability of a strategy that invests in mutual funds according to their AFP. Specifically, at the end of each May, August, November, and February, we sort mutual funds into ten portfolios by their AFP and rebalance them quarterly. We compute equally weighted returns for each decile portfolio over the following quarter, both after and before expenses. In addition, we estimate the risk-adjusted returns on the portfolios as intercepts from time-series regressions, according to the Capital Asset Pricing Model (CAPM); Fama and French’s (1993) three-factor model (market, size, and value); Carhart’s (1997) four-factor model, which augments the Fama-French factors with Jegadeesh and Titman’s (1993) momentum factor; and the five-factor model, which adds Pastor and Stambaugh’s (2003) liquidity risk factor to the four-factor model. For instance, the Carhart four-factor alpha is the intercept from the following time-series regression: \begin{align} R_{p,t} -R_{f,t}&= \alpha_p + \beta_m (R_{m,t}-R_{f,t}) + \beta_{\textit{smb}}\textit{SMB}_t +\beta_{\textit{hml}}\textit{HML}_t +\beta_{\textit{umd}}\textit{UMD}_t +\varepsilon_{p,t},\notag\\ \end{align} (4) where $$R_{p,t}$$ is the return in month $$t$$ for fund portfolio $$p$$, $$R_{f,t}$$ is the 1-month Treasury-bill rate in month $$t$$, $$R_{m,t}$$ is the value-weighted stock market return in month $$t$$, SMB$$_{t}$$ is the difference in returns between small and large capitalization stocks in month $$t$$, HML$$_{t}$$ is the return difference between high and low book-to-market stocks in month $$t$$, and UMD$$_{t}$$ is the return difference between stocks with high and low past returns in month $$t$$. Table 2 presents the portfolio results, which indicate that AFP predicts future fund performance. Panel A shows the net returns for portfolios of funds sorted by their AFP. In the quarter following portfolio formation, mutual funds with high index-based (trade-based) AFP in decile 10 outperform the funds with the lowest AFP in decile 1 by 23 (12) bps per month, or 2.76% (1.44%) per year. This superior performance cannot be attributed to their high propensity to take risks or different investment styles; the differences in the alphas obtained from the CAPM, Fama-French three-factor, Carhart four-factor, and Pastor-Stambaugh five-factor models are 21, 27, 20, and 22 (12, 15, 12, and 12) bps per month for index-based (trade-based) AFP, respectively, and all of these differences are statistically significant. Table 2 AFP and mutual fund returns: Decile portfolios A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net returns Average return 0.87 0.90 0.93 0.91 0.93 0.92 0.96 1.02 1.00 1.10 0.23*** (3.42) (3.74) (3.95) (3.89) (3.96) (3.90) (4.01) (4.18) (4.06) (4.16) (3.14) CAPM $$\alpha $$ –0.18 –0.12 –0.08 –0.10 –0.08 –0.09 –0.06 –0.01 –0.03 0.03 0.21*** (–2.79) (–2.49) (–1.79) (–2.26) (–2.1) (–2.3) (–1.52) (–0.3) (–0.58) (0.37) (2.97) Fama-French $$\alpha$$ –0.19 –0.14 –0.09 –0.11 –0.09 –0.11 –0.07 –0.01 –0.02 0.07 0.27*** (–3.52) (–3.28) (–2.52) (–3.27) (–2.87) (–3.28) (–2.13) (–0.26) (–0.39) (1.38) (3.59) Carhart $$\alpha $$ –0.16 –0.12 –0.08 –0.10 –0.09 –0.11 –0.08 –0.03 –0.03 0.04 0.20*** (–2.95) (–2.84) (–2.29) (–2.8) (–2.39) (–3.16) (–2.24) (–0.72) (–0.76) (0.82) (3.23) Pastor-Stambaugh $$\alpha $$ –0.17 –0.13 –0.09 –0.11 –0.09 –0.12 –0.08 –0.03 –0.03 0.05 0.22*** (–3.17) (–3.06) (–2.49) (–3.08) (–2.64) (–3.43) (–2.4) (–0.79) (–0.79) (0.93) (3.49) Gross returns Average return 0.97 1.00 1.03 1.00 1.02 1.01 1.05 1.11 1.10 1.20 0.23*** (3.83) (4.14) (4.35) (4.29) (4.35) (4.29) (4.41) (4.57) (4.46) (4.56) (3.14) CAPM $$\alpha $$ –0.08 –0.02 0.02 0.00 0.01 0.00 0.03 0.08 0.07 0.13 0.21*** (–1.2) (–0.47) (0.38) (–0.03) (0.35) (0.02) (0.79) (1.74) (1.18) (1.74) (2.97) Fama-French $$\alpha $$ –0.09 –0.04 0.00 –0.02 0.00 –0.01 0.02 0.09 0.08 0.18 0.27*** (–1.6) (–0.92) (0.11) (–0.57) (–0.04) (–0.42) (0.72) (2.44) (2.00) (3.33) (3.59) Carhart $$\alpha $$ –0.05 –0.02 0.01 –0.01 0.01 –0.02 0.02 0.07 0.07 0.15 0.20*** (–0.98) (–0.54) (0.29) (–0.24) (0.23) (–0.47) (0.46) (1.88) (1.69) (2.84) (3.23) Pastor-Stambaugh $$\alpha $$ –0.06 –0.03 0.00 –0.02 0.00 –0.02 0.01 0.07 0.07 0.15 0.22*** (–1.19) (–0.74) (0.05) (–0.51) (–0.01) (–0.7) (0.28) (1.79) (1.62) (2.89) (3.49) B. Decile portfolios formed according to the trade-based AFP Net returns Average return 0.87 0.92 0.92 0.90 0.93 0.88 0.88 0.93 0.95 0.99 0.12*** (3.23) (3.73) (3.80) (3.79) (3.88) (3.72) (3.67) (3.77) (3.82) (3.63) (2.98) CAPM $$\alpha $$ –0.16 –0.06 –0.05 –0.05 –0.03 –0.07 –0.08 –0.05 –0.03 –0.04 0.12*** (–2.29) (–1.15) (–1.03) (–1.22) (–0.76) (–1.83) (–2.01) (–0.99) (–0.63) (–0.55) (2.94) Fama-French $$\alpha$$ –0.14 –0.07 –0.08 –0.09 –0.05 –0.10 –0.11 –0.06 –0.03 0.01 0.15*** (–3.18) (–1.99) (–2.07) (–2.35) (–1.57) (–2.87) (–3.04) (–1.45) (–0.76) (0.11) (3.93) Carhart $$\alpha $$ –0.15 –0.07 –0.07 –0.07 –0.04 –0.08 –0.10 –0.07 –0.04 –0.03 0.12*** (–3.28) (–1.63) (–1.89) (–1.92) (–1.15) (–2.32) (–2.62) (–1.85) (–0.96) (–0.7) (3.25) Pastor-Stambaugh $$\alpha $$ –0.15 –0.07 –0.08 –0.08 –0.05 –0.09 –0.11 –0.08 –0.04 –0.03 0.12*** (–3.23) (–1.76) (–2.06) (–2.18) (–1.33) (–2.62) (–2.93) (–2.03) (–1.14) (–0.62) (3.34) Gross returns Average return 0.98 1.02 1.01 1.00 1.02 0.97 0.97 1.03 1.05 1.09 0.12*** (3.62) (4.13) (4.20) (4.18) (4.27) (4.12) (4.06) (4.16) (4.21) (4.03) (3.02) CAPM $$\alpha $$ –0.05 0.04 0.05 0.04 0.06 0.02 0.01 0.05 0.07 0.06 0.12*** (–0.76) (0.85) (1.05) (0.93) (1.45) (0.52) (0.27) (1.00) (1.29) (0.84) (2.98) Fama-French $$\alpha $$ –0.03 0.03 0.02 0.01 0.04 –0.01 –0.01 0.04 0.07 0.11 0.15*** (–0.79) (0.68) (0.54) (0.21) (1.14) (–0.15) (–0.36) (1.08) (1.92) (2.47) (3.98) Carhart $$\alpha $$ –0.05 0.03 0.02 0.02 0.05 0.01 0.00 0.02 0.06 0.07 0.12*** (–1.03) (0.78) (0.57) (0.57) (1.37) (0.32) (–0.02) (0.63) (1.63) (1.53) (3.29) Pastor-Stambaugh $$\alpha $$ –0.05 0.03 0.01 0.01 0.04 0.00 –0.01 0.02 0.06 0.08 0.12*** (–1.01) (0.62) (0.38) (0.29) (1.13) (0.01) (–0.33) (0.46) (1.47) (1.59) (3.39) A. Decile portfolios formed according to the index-based AFP Low 2 3 4 5 6 7 8 9 High High-low Net returns Average return 0.87 0.90 0.93 0.91 0.93 0.92 0.96 1.02 1.00 1.10 0.23*** (3.42) (3.74) (3.95) (3.89) (3.96) (3.90) (4.01) (4.18) (4.06) (4.16) (3.14) CAPM $$\alpha $$ –0.18 –0.12 –0.08 –0.10 –0.08 –0.09 –0.06 –0.01 –0.03 0.03 0.21*** (–2.79) (–2.49) (–1.79) (–2.26) (–2.1) (–2.3) (–1.52) (–0.3) (–0.58) (0.37) (2.97) Fama-French $$\alpha$$ –0.19 –0.14 –0.09 –0.11 –0.09 –0.11 –0.07 –0.01 –0.02 0.07 0.27*** (–3.52) (–3.28) (–2.52) (–3.27) (–2.87) (–3.28) (–2.13) (–0.26) (–0.39) (1.38) (3.59) Carhart $$\alpha $$ –0.16 –0.12 –0.08 –0.10 –0.09 –0.11 –0.08 –0.03 –0.03 0.04 0.20*** (–2.95) (–2.84) (–2.29) (–2.8) (–2.39) (–3.16) (–2.24) (–0.72) (–0.76) (0.82) (3.23) Pastor-Stambaugh $$\alpha $$ –0.17 –0.13 –0.09 –0.11 –0.09 –0.12 –0.08 –0.03 –0.03 0.05 0.22*** (–3.17) (–3.06) (–2.49) (&nda