When is Stock Picking by Mutual Funds Successful? Paul Schultz*

April 19, 2007

Preliminary. Do not quote.

*

University of Notre Dame. I am grateful to Rick Mendenhall, Hayong Yun, and seminar participants at the University of Notre Dame and the University of Western Ontario for comments.

In this paper, I provide evidence that mutual funds are particularly adept at finding underpriced stocks among small growth companies. On average, small growth stocks with high returns over the prior year that are held by funds outperform stocks with similar characteristics by 1.06% per month. While I present evidence that mutual funds are able to find underpriced stocks among larger firms and value stocks as well, the abnormal returns on their investments in these stocks are much smaller. This paper adds to a large literature on mutual fund performance that has arisen over the last forty years. The performance of professionally managed portfolios is interesting in and of itself, but it is also interesting for its implications about market efficiency. The returns earned by mutual funds are real -in contrast many academic studies examine market efficiency using strategies that may or may not be implementable in practice. Also, if anybody can find undervalued securities it should be the professionals employed by mutual funds. They have the training, access to information, and incentives to produce superior returns. The returns that we would expect mutual funds to earn in an efficient market has been a subject of debate. An aggressive interpretation of the efficient market hypothesis might suggest that funds should not produce abnormal returns either before or after expenses. Grossman and Stiglitz (1980), however, point out that there is no incentive to undertake security analysis unless there is enough mispricing to cover the costs of analysis. This suggests that mutual funds should be able to produce abnormal returns before expenses and trading costs. This does not mean that active mutual funds should outperform passive funds after trading costs. For one thing, successful funds will attract additional capital until diseconomies of scale make it impossible for them to produce abnormal returns (see Berk and Green (2004)). In addition, individuals with scarce investment skills should be able to extract most of the gains from their abilities in the form of higher fees. Empirical evidence generally supports the hypotheses that mutual funds underperform after expenses. In the classic early study of mutual funds performance, Jensen (1969) examines the returns after expenses of 115 mutual funds over the 1955-1964 period. Despite a survivorship bias in his data, he estimates that on average, mutual funds underperformed by about 1.1% per year. This conclusion held even after adding back all expenses except brokerage commissions. In other words, funds did not do well enough on their trading to pay the commissions. In a 1

comprehensive study of the returns to 1,892 diversified equity funds over 1962-2003, Carhart (1997) finds that on average, the funds slightly underperformed the market. After adjustment for characteristics of fund holdings though, underperformance was more striking. Funds tended to overweight small firms, and firms with low book-to-market ratios in their portfolios. After adjusting for these factors, funds on average underperformed by just under 2% per year. Wermers (2000) compares mutual fund returns before and after expenses. Over the 20 year period from 1975 through 1994, the S&P 500 earned annual returns of 15.4%, and the CRSP value-weighted index earned returns of 15.6%. When mutual funds are weighted by total net assets, the portfolios they held returned 16.9% per year, or 1.3% per year more than the market. When the funds are equally-weighted, they returned 17.7% per year, outperforming the market by 2.1% per year. The net returns, after expenses and transactions costs averaged 14.6% for a value-weighted average of funds, and 14.9% for an equal-weighted average. Hence after expenses, funds underperformed the market by 0.7% to 1.0% per year. Other studies show directly that funds are able to pick undervalued stocks even if funds don’t beat benchmarks when expenses are incorporated. Daniel, Grinblatt, Titman, and Wermers (1997) examine holdings of mutual funds at the end of each quarter from 1974 through 1994. They then compare returns of each stock in each fund’s portfolio with the returns of one of 125 portfolios based on quintiles of size, book-to-market, and momentum. Funds as a whole outperform by about 79 basis points per year before expenses. Aggressive growth funds outperform by about 1.49% per year. The Daniel, Grinblatt, Titman, and Wermers (1997) paper, like other studies of mutual fund stock-picking that test market efficiency, are plagued by the joint hypothesis problem. Stocks owned by funds may perform well relative to benchmarks either because the funds pick undervalued stocks, or because they invest in stocks that are riskier than the benchmarks. Baker, Litov, Wachter, and Wurgler (2005) finesse this issue by examining the returns of stocks that funds have bought and sold around their next earnings announcement. Baker, Litov, Wachter, and Wurgler find that the difference in returns between purchased and sold stocks is around 12 basis points at the next earnings announcement. Differences in risk should have little impact on returns over a short three-day period. Hence this evidence that fund managers can pick stocks is unambiguous and statistically significant even if the measured economic consequences of their 2

stock-picking is small. Of course, not all funds are equally adept at picking stocks, and recent research shows several ways to identify mutual funds that are able to find undervalued stocks. Funds that believe they can identify undervalued securities invest heavily in the most promising stocks rather than holding more diversified portfolios. Funds run by less skillful managers, on the other hand, are closet indexers. Cremers and Petajisto (2006) use CDA Spectrum data from 1980 through 2003 to identify fund holdings. They define the active share of fund’s holdings as the proportion of its portfolio invested in shares that exceed what would be held if the fund simply held its benchmark. About 1/3 of all funds have active shares of 90% or more. Cremers and Petajisto show that funds with larger active shares have greater expenses and higher turnover, but also produce greater risk-adjusted net returns. Similar results are obtained by Baks, Busse, and Green (2006). Using fund holdings from CDA Spectrum, they examine whether fund managers who take big bets earn larger returns than fund managers who diversify their holdings more broadly. Baks, Busse, and Green (2006) show that managers who concentrate their holdings among a small number of stocks outperform those that are more diversified. This is true both for raw returns and characteristic-adjusted returns and is true both before and after expenses. Focused managers outperform better diversified managers by 30 basis points a month, or almost 4% per year. The results are statistically weak though. The difference in returns between concentrated and diversified funds is especially large for largest bets that the funds make. Annualized return differences between the most heavily held stocks of concentrated versus diversified funds runs to about 10% per year. Funds that are able to pick stocks are also more likely to concentrate their holdings in specific industries. Kacperczyk, Sialm, and Zheng (2004) examine mutual fund holding concentration by industry. They classify all stocks into 10 broadly defined industries. They then calculate a measure of industry concentration based on the squared differences between each fund’s holdings in an industry, and the industry’s weight in the entire stock market. Using the four-factor Carhart model, they find that funds with above median industry concentration produce abnormal gross returns of 1.58% per year, and abnormal net returns of 0.33%. Funds with below median concentration produce gross abnormal returns of 0.36% and net abnormal returns of -0.77%. 3

Fund size is also a determinant of the fund’s ability to pick stocks. Chen, Hong, Huang, and Kubik (2004) (Henceforth CHHK) run cross-sectional regressions of fund returns on the log of net assets and other variables for each month over 1963 - 1999. They then estimate FamaMacBeth time series averages and t-statistics for the coefficients. They estimate risk-adjusted fund returns in several ways - a simple market return adjustment, the capital asset pricing model, and the three and four factor models. Regardless of the risk adjustment and whether or not returns are measured after expenses, the regressions show that smaller funds outperform larger funds. A two standard-deviation shock to fund size alters fund returns by about 96 basis points per year. CHKK demonstrate that it is the size of the fund itself, and not the fund family that is a drag on returns. In fact, all else equal, a fund that belongs to a large fund family will have higher returns perhaps reflecting lower commissions or higher fees for lending stock. Fund families, unlike funds, allow managers to have autonomy in their investment decisions and to retain their capital. CHKK suggest that if funds were organized in a way that allowed co-managers to retain control over assets, rather than making investment decisions by committee, size would not hamper funds returns. Kacperczyk and Seru (2007) provide an alternative way of identifying skillful stock selection. They use changes in analysts’ consensus recommendations over the prior four quarters as a measure of public information. R2's from regressions of changes in fund holdings on changes in consensus recommendations indicate that on average, 29% of changes in fund holdings can be explained with this measure of public information. There is considerable cross-sectional variation however, with R2's of 1.86% or less for 10% of the funds, and R2's greater than 76% for 10% of the funds. Abnormal returns earned by funds are significantly negatively related to the proportion of changes in holdings that are explained by public information. These results appear to be quite robust to changes in the measurement of abnormal returns. This paper also examines funds’ ability to identify underpriced stocks, but focuses on the characteristics of the stocks that make this possible rather than the characteristics of the funds. The work presented here is similar in spirit to Coval and Moskowitz (2001), who show that funds earn abnormally large returns on investments in nearby firms. Local stocks are investments where access to informal information gives funds a comparative advantage in security analysis. 4

This paper similarly examines funds’ ability to pick stocks among those where the funds’ analytical skills are likely to be especially valuable. Growth stocks, defined as those with low book-to-market ratios, are likely to be more difficult to analyze than value stocks, defined as stocks with high book-to-market ratios. Growth stock prices reflect cash flows to be generated in the distant future, which are more difficult to predict than near-term cash flows. Growth stocks’ cash flows may depend on successful research and development or technological innovation. Under these circumstances, sophisticated securities analysis is especially likely to be valuable. Security analysis is also likely to be valuable for stocks that are not well known. Small stocks, or stocks with low market capitalizations, are likely to be followed by few investors. Lack of analyst coverage is one reason why a small stock would be unknown to most investors. In addition, few investors would be familiar with a small company through its products, its advertisements, or through interactions with its employees. This suggests that security analysis is more likely to uncover mispricings among small companies. This intuition is supported by the results of this paper. There is some weak evidence that funds can pick stocks in general. Over the entire 25 year sample period, the stock holdings of funds beat returns of stocks in the same size, book-to-market, and momentum quintiles by 12 basis points per month when the funds are equal-weighted, and by 11 basis points per month when the funds are value-weighted. There is some difference in the performance of fund holdings for stocks in different size, book-to-market and momentum quintiles. Fund holdings of small stocks, growth stocks, and winners, outperform similar stocks by more than fund holdings of large stocks, value stocks and losers outperform comparable stocks. When returns are examined separately for stocks held by funds from each of the 125 size, book-to-market and momentum characteristic portfolios though, a more dramatic picture emerges. Fund holdings of stocks that are simultaneously small, growth stocks, and previous year winners outperform similar stocks by 1.06% per month. In general, fund holdings of stocks in the characteristic portfolios that tilt toward small size and growth outperform similar stocks by wide and highly significant margins. Chen, Hong, Huang, and Kubik (2004) document that small mutual funds offer highe returns than larger funds. The results of this paper could suggest that the superior performance of small mutual funds could be a result of small funds investing more in small growth stocks. If ind 5

that smaller funds do indeed earn larger abnormal returns on their stock holdings than larger funds, but they earn larger returns on their investments in both large value stocks and small growth stocks. The rest of the paper is organized as follows. The data used in this paper is described in Section I. Empirical results are described in Section II. Section III summarizes the paper and draws conclusions.

I. Data

I examine the stock holdings of funds that are classified as aggressive growth, growth, or growth and income for 1980 through 2005. I am only concerned with funds that actively attempt to find undervalued securities. Hence funds that include the word “index” in their name are excluded. Fund holdings are obtained from the CDA spectrum data. I exclude funds for a particular quarter if they have fewer than 20 stocks in their portfolio. Information on individual stocks for each month is obtained from CRSP. Following Daniel, Grinblatt, Titman, and Wermers (1997), I categorize stocks by size, book-to-market, and momentum, and then measure abnormal returns of individual stocks by subtracting out the value-weighted average return of stocks with similar characteristics. First, at the end of each month from the end of 1979 through 2005, I calculate the size (capitalization) of all CRSP stocks by multiplying the closing price by the number of outstanding shares. All stocks are sorted by size each quarter, and place into five quintiles based on the distribution of sizes of NYSE stocks. Nasdaq stocks tend to be smaller than NYSE stocks, hence the smaller quintiles have a significantly larger number of stocks. Stocks in each of the size quintiles are next sorted into five quintiles based on book-tomarket value. Book values are obtained from the Compustat annual file by adding stockholder’s equity and deferred tax and investment tax credit, and subtracting out the value of preferred stock. I use the redemption value of the preferred if available, then the liquidation value, and finally the carrying value if that is Compustat’s only value for preferred stock. Book-to-market values are calculated for each quarter by dividing the most recent book value by the current size. Most studies that use book-to-market as a factor require book values that are a minimum 6

of several months old. There are good reasons for this requirement. First, if a researcher is examining an investment strategy, data like book values that is used in the strategy must be available for the strategy to be implemented. Second, announcements of higher than expected earnings usually imply higher than expected book values. So, stock that are selected to have high book values that are not publicly known can be expected to have spuriously high returns. Neither of these considerations apply in this paper. I am not testing any strategy that relies on knowing non-public information, but only seeing whether the stocks that mutual funds own outperform similar stocks. Spuriously high returns from selecting high book-to-market stocks based on nonpublic information is also not an issue here. Fund holdings are compared with stocks of similar size and book-to-market and thus both the fund holdings and their benchmark are subject to the same biases from using book values that are not public. The advantage of not requiring a minimum time between the date of the book value and the formation of characteristic portfolios is that fewer securities need to be discarded. This is particularly important if funds hold recent IPOs. Finally, each of the 25 size and book-to-market portfolios is split into five portfolios based on returns over the previous 12 months ending one month before the formation month. I do not require a stock to have traded for any minimum period for inclusion in a characteristic portfolio. Stocks with shorter trading histories are unlikely to be in the extreme momentum portfolios. After all stocks are sorted into 125 characteristic portfolios, a value-weighted average return of all stocks in each portfolio is calculated for the following calendar quarter. There is no requirement that stocks trade for the entire quarter to be included in the characteristic portfolio, and hence no survivorship bias. The characteristic portfolios are described in Table I. For each of the 125 characteristic portfolios an average size, book-to-market ratio, and return over the prior 12 months is calculated across the 313 months from December 1979 through December 2005. Grand averages are then calculated for market capitalization, book-to-market, and momentum quintiles by taking averages across each of the 25 characteristic portfolios in a quintile. Panel A of Table I shows results. The mean capitalization of firms in the small firm quintile is $50.6 million while the mean capitalization of firms in the large firm quintile is just over $13 billion. Average book-to7

market ratios range from 0.03 for growth stocks to 1.59 for value stocks. Average returns over the previous 12 months range from -27.9% for losers to 100.8% for winners. Average numbers of stocks in characteristic portfolios are shown at the bottom of Panel A. Recall that the size quintile breakpoints are based on NYSE listed stocks. The smaller Nasdaq and Amex stocks will tend to crowd into the smaller stock portfolios, and hence these characteristic portfolios will have more stocks. On average, over the 25 year sample period, there are 13 stocks in each of the 25 large stock characteristic portfolios. The minimum number for any month is 10. The number of stocks per characteristic portfolio is slightly larger for the second, third and fourth size quintiles, and much larger for characteristic portfolios of small firms. On average, there were 129 stocks in each of the 25 characteristic portfolios of small firms. In the remainder of the paper, abnormal returns of individual stocks are always calculated relative to the value-weighted return of all other stocks in their characteristic portfolios. Many of the results, however, will be presented for 25 size and book-to-market portfolios that are obtained by merging each of the momentum portfolios within a size and book-to-market classification. Panels B, and C of Table I show average size and book-to-market ratios for the 25 size and book-to-market classifications. Sizes tend to be smaller for value firms than growth firms. Of particular interest is that the smallest stocks are small growth firms, with an average capitalization of $61.1 million. As we will see, these are the stocks where mutual funds have the most stock picking success. As shown in Panel C, the mean book-to-market ratio for these small value stocks is 3.245, much higher than that of other value stocks.

II. Results

Figure 1a shows the number of aggressive growth, growth, and growth and income funds reporting holdings each month from March 1980 through June 2006. Figure 1b shows the total value of U.S. shares held by these funds each month. Over the past 25 years, the number of funds has increased by about 150%, while the value of the stocks that they hold has increased ten-fold. It can also be seen in each graph that the number of funds that report during months at the end of calendar quarters, that is March, June, September, and December, far exceed the number who 8

report holdings during other months. The spectrum data provides a snapshot of each fund’s holdings of stock at the end of the quarter or six month period. In calculating returns to fund holdings, I implicitly assume that the fund’s holdings are maintained unchanged for the next three months. Since funds turn over their portfolios about once a year, this is a reasonable assumption. If funds have some ability to find underpriced securities however, this assumption will result in an underestimation of their stock picking ability. New positions in undervalued shares could have been held for several weeks before they show up in the snapshot of a fund’s holdings. Similarly, if a fund decides to sell shares they believe to be overpriced, the methodology of this paper may assume that the shares are held for several more weeks. It is worth emphasizing that the returns examined here are the returns of the stocks held by the mutual funds, and not the returns of the funds themselves. Since the goal of the paper is to find out if funds are better able to find mispricing in certain types of stocks than in others, the use of stock returns is appropriate. Returns to fund investors also reflect expenses, fees, and the necessity of holding some cash.

A. Which Stocks do Funds Hold? Table II provides information on how funds invest across different size and book-tomarket quintiles. Panel A shows the time-series average of the proportion of sample funds that hold stocks in each category. To obtain this, I calculate the proportion of sample funds that hold a given stock each month, using all fund positions reported in the last three months. I then average across stocks in the size and book-to-market quintile, and then calculate a grand average across all quarters. For example, on average, across all quarters from 1980 - 2005, a large value stocks could be expected to be held by 7.5% of all funds in the sample that quarter. Panel A demonstrates that the proportion of funds holding a stock drops off steadily as firm size decreases. Stocks in the smallest size quintile would be expected to be held by about one in five hundred funds. Panel A also reveals that, for the larger size quintiles, growth stocks are likely to be held by a larger proportion of funds than value stocks. For stocks in the small size quintile the results are less clear. Small growth stocks are likely to be held by a smaller proportion of funds than stocks in book-to-market categories two through hour, but are held by a larger proportion of 9

funds than small value funds. Panel B reports the number rather than the proportion of funds holding stocks in each category. Stocks in the largest quintile are held by, on average, more than 60 funds regardless of their book-to-market category. Large growth stocks, on average, are held by 96.4 funds on average. Clearly, large stocks are familiar to large numbers of mutual fund managers and competition to find undervalued securities is fierce among large stocks. Small growth stocks, on the other hand, are held by 1.9 funds on average. Small stocks with higher book-to-market ratios are held by between 1.4 and 2.2 funds on average. Clearly, fewer funds are competing in this sector of the stock market. This point is also made in Panel C, which reports the proportion of stocks in size and book-to-market quintiles that are not held by any funds. Most small stocks are not held by any sample mutual funds, regardless of their book-to-market category. More than 58% of small growth stocks are not held by any funds. I do not report fund holdings by momentum quintiles in Table II because most of the variation in fund holdings can be attributed to size or book-to-market. In general though, funds are more likely to hold recent winners than recent losers.

B. The Returns of Stocks Held by Funds Panel A of Table III reports the average, across all months, of the difference between the equal-weighted average of fund returns and the CRSP value-weighted index. Average returns of stocks that are held by each fund are a weighted average return of the fund’s holdings of stocks , where the weights are the dollar value of the positions. An equal-weighted average of returns of fund is then calculated. Over the entire period, an equal-weighted average of fund holdings outperformed the market by 20 basis per quarter, or about 2.4% per year. The t-statistic is 3.35, indicating the difference in monthly returns is significant at the 1% level. When the difference in returns between fund holdings and the market is calculated over 1980 through 1992, the difference increases slightly to 19 basis points per month. The t-statistic remains significant at 2.18. Return differences over 1993 - 2005 average 21 basis points per month with a t-statistic of 2.55. The last row of Panel A reports results when the fourth quarter of 1999, an outlier, is omitted. Returns of stocks held by funds still exceeds the return of the market by 17 basis points 10

per quarter. The t-statistic for the difference is 3.05. Panel B of Table III repeats the analysis but uses a value-weighted average of the returns across funds, where the weights are the total value of stock positions held by the funds. In effect, this panel compares the portfolio of holdings of all mutual funds with the return on the market. Now, when all quarters are used, the average difference between the returns of stocks held by funds and the market return falls to 17 basis points per month. The statistical significance remains high, with a t-statistic of 3.86. The lower returns when funds are value-weighted are consistent with the findings of Chen, Hong, Huang, and Kubik (2004) that small mutual funds outperform large ones. If mutual funds tend to hold risky stocks, returns that are on average greater than the market would not indicate stock picking ability. To adjust for risk, abnormal returns are calculated for each stock each month by subtracting the value-weighted return of all other stocks in its characteristic portfolio. Note that the benchmark used to calculate and abnormal return for a stock does not include its own return. This refinement is unimportant for small stocks but can make a difference for the largest stocks where characteristic portfolios can have as few as ten stocks. An abnormal return for each fund each month is obtained by taking a value-weighted average of the abnormal return of the fund’s stocks. I then calculate the average abnormal performance of the mutual fund sector by taking an equal-weighted average of each fund’s abnormal return. Panel C reports the time series average of mutual fund abnormal returns. Stocks held by mutual funds exceed the returns on their characteristic portfolios by 12 basis points per month. This is about 60% of the difference between mutual fund returns and the return on the market, confirming that funds overweight their portfolios in stocks that are risky, or at least have characteristics associated with high returns. The t-statistic on the time series average is 3.87, suggesting that mutual funds are able to find stocks that outperform other stocks with similar characteristics. As an alternative, I value-weight each fund by its holdings of U.S. stocks to get the mutual fund sector’s average abnormal return each quarter. Panel D reports the time series average of the monthly abnormal returns for the mutual fund sector. When the funds are valueweighted rather than equal-weighted, the average monthly abnormal return for the entire period falls from 17 basis points to 11 basis points. The t-statistic of 3.79 confirms that the difference in 11

returns between stocks held by funds and similar stocks is statistically significant.

C. Which Stocks do Funds Analyze Most Successfully? Any ability that funds have to pick underpriced securities may be most useful for small growth stocks. Small stocks are not followed by many professional investors and are therefore more likely to be mispriced. Much of the value of growth stocks comes from cash flows in the distant future, which require great skill to forecast. Hence in this section I examine the abnormal returns earned by funds on their investments in stocks of different size, book-to-market, and past returns. For each fund each month, I calculate the abnormal return for each of its stocks by subtracting out the return of the other stocks in its characteristic portfolio. I then calculate the value-weighted average return for stocks in each size quintile, each book-to-market quintile, and each 12-month momentum quintile. A grand average is calculated for each quintile each month across funds, with fund weights proportional to the value of their holdings of stocks in that quintile. In effect, these grand averages provide the abnormal returns an investor would earn if he bought all the stock holdings of all growth, aggressive growth and growth and income funds. Table IV shows averages and t-statistics calculated across the quarters from March 1980 through December 2005. Panel A shows abnormal returns by size quintile. Fund holdings, on average, significantly outperform their characteristic portfolios for each size quintile, with the abnormal returns increasing monotonically as firm size falls. Fund holdings of stocks in the smallest quintile outperform their characteristic portfolio by 37 basis points per month, or about 4.4% per year. Panel B shows average abnormal returns for stocks held by mutual funds across book-tomarket portfolios. Average abnormal returns increase monotonically as book-to-market quintiles decrease from value to growth. Average abnormal returns for growth stocks held by funds are 16 basis points per month. Panel C reports mean abnormal returns for each of the momentum quintiles. For the quintile of stocks with the largest returns over the last 12 months, the stocks held by funds outperform their characteristic portfolios by 25 basis points per month, or about 3% per year. The t-statistic of 3.55 indicates that these abnormal returns are reliably different from zero. Mean 12

abnormal returns decrease monotonically as past returns decrease from the winner to the loser portfolio. I next calculate value-weighted abnormal returns of all stocks held by all mutual funds for each of the 125 characteristic portfolios each quarter. That is abnormal returns of each stock in each characteristic portfolio are weighted by the value of their holdings by all funds to calculate an average abnormal return. In effect, these are the returns earned by the stock holdings of the entire growth fund sector. A time-series average is then calculated across the quarter from 1980 - 2005. For most of the remainder of the paper, I report abnormal returns for a subset of the 125 characteristic portfolios in order to save space. It is worthwhile though to get the entire picture once, so I report average abnormal returns for each characteristic portfolio in Table V. Each Panel of the table reports abnormal returns for stocks in the 25 characteristic portfolios within a size quintile. Each panel reports abnormal returns within a size quintile by book-tomarket (rows) and past returns (columns). T-statistics are reported in parentheses under the average monthly abnormal returns. T-statistics that indicate abnormal returns that are different from zero at the 5% significance level are in bold. Likewise, mean abnormal returns of 50 basis points or more per month are also in bold. Panels A and B show abnormal returns earned by funds on their investments in stocks in the largest and second largest quintiles. Abnormal returns are for the most part small and insignificant. The exception is that mutual funds do seem to earn abnormal returns on their investments in larger winners. For example, fund investments in large value firms that were winners produce mean abnormal returns of 30 basis points per month, while their investments in large growth firms with large recent returns earn mean abnormal returns of 32 basis points per month. Abnormal returns are calculated relative to the other stocks in the characteristic portfolio, so this means that funds are able to pick large winners that outperform other large winners. Across the 125 characteristic portfolios, the largest and most significant abnormal returns by far are obtained from investments in small growth stocks. Panel D shows that among the second smallest quintile of stocks, abnormal returns for winner growth stocks averaged 95 basis points per month with a t-statistic of 3.55. Monthly abnormal returns for growth stocks in the second smallest quintile that were around the median in past returns averaged 90 basis points per month, with a t-statistic of 2.20. Panel E reports results for fund investments in the smallest 13

quintile of stocks. Abnormal returns are positive and significant for all of the small growth stocks, regardless of their previous year return. The largest abnormal returns are earned by funds on their investments in small growth stock that had been winners over the prior year. Fund investments in these stocks outperformed similar stocks by 1.06% per month, or 12.72% per year. Recall that in estimating returns on their investments, I assume that funds maintain the same positions for three months after the date of the holdings from their quarterly or semi-annual report. Fund managers presumably feel that the stocks they hold are undervalued on the holding date, but may well sell them before three months have elapsed. Hence the assumption that funds own stocks from the announced holding date through three months is likely to understate the ability of funds to find undervalued securities. Likewise, a stock that is purchased within a quarter does not show up in the funds holdings until the end of the quarter. If the stock was undervalued at the time of the purchase, it is likely to have appreciated while held by the fund, but before it shows up in my data on fund holdings. To explore this issue, I examine abnormal returns separately for the first, second, and third months following fund’s holding dates. Results are reported in Table VI. To save space, I only report results for the fund holdings in the smallest and second smallest quintile of firm sizes. Panels A, B, and C report mean abnormal returns for holdings of small stocks one, two and three months after the holding date. Abnormal returns in bold are significantly different from zero at the 5% level. Panel A shows that fund holdings in small firms generally outperform other stocks in their characteristic portfolios for the first month after the holding date regardless of whether the stocks are value or growth stocks or past winners or losers. For several of the characteristic portfolios, stocks held by funds outperform similar stocks by more than 1.5% during the first month after the holding. Panel B shows that the growth stocks in the small quintile tend to continue to outperform their characteristic portfolios for the second month after the holdings date, but value stock abnormal returns fall off. Finally, Panel C shows that there is little or no significant abnormal return earned in the third month after the holding date. Panels D, E, and F present the same analysis, this time for fund holdings of stocks in the second smallest quintile. Again, abnormal returns generally decline in magnitude and 14

significance as the holding date recedes. This is what we expect - mispriced securities should move to their true values and abnormal returns should decline over time. These results also suggest that if we knew the actual date that funds traded stocks abnormal returns would be higher. Finally, the deterioration of the performance of fund holdings over time suggests that abnormal returns from small stock investments may not be earned for a long enough period to allow funds to cover the costs of trading these stocks. It is interesting to ask whether funds’ success with small growth companies is due to an ability to find overlooked gems, or an ability to avoid overpriced stocks. To test this, I first find the five stocks in each characteristic portfolio with the best returns over the next month, and the five stocks with the worst returns. I then calculate the proportion of each characteristic portfolios’ value that comes from each of the five best and worst stocks. I also calculate the proportion of total fund investment in the characteristic portfolio that is in each of the five best and five worst performing stocks. Finally, for each month, I compute differences in the proportion of fund investment in the characteristic portfolio that is in the best (worst) stocks and the proportion of the characteristic portfolio value represented by the best (worst) stocks. I then calculate a time-series average of the differences, and t-statistics based on the time-series standard deviation. Results for fund investments in the characteristic portfolios in the two smallest size quintiles and two lowest book-to-market portfolios are shown in Table VII. Panel A reports mean differences between the proportion of money invested in the five best stocks by funds and the value of the stocks as a proportion of their characteristic portfolio. If funds are particularly adept at finding big winners, we would expect these differences to be positive. Results here are mixed. Funds do generally overinvest in the best stocks in the second smallest size quintile. Differences are positive and significant for five of the ten characteristic portfolios in the second smallest size quintile. There appears to be little pattern to the return differences across either book-to-market or prior 12 month return quintiles. For stocks in the smallest quintile though, funds generally underinvest in the best stocks although the results are almost always insignificant. Throughout Table VII, the magnitude of the difference between fund holding proportions and characteristic portfolio weights is larger for the second smallest size portfolio than the size portfolio. Recall that characteristic portfolios in the smallest size quintile were made up of 129 15

stocks on average, while characteristic portfolios in the second smallest size quintile contained only 30 stocks on average. Hence the five best (or worst) stocks in the smallest size quintile might contain only 4% of the market value of the stocks in their characteristic portfolio while the best (or worst) five stocks in characteristic portfolios in the second smallest quintile should contain about 16% of the portfolios value. Thus it is not surprising that differences between proportions held by funds and characteristic portfolio weights are larger for the second smallest quintile stocks. Panel B reports differences in the proportions of worst stocks held by funds and the weights of these stocks in their characteristic portfolios. If funds successfully avoid the worst stocks, we would expect these differences to be negative. In fact all 20 are negative, with 14 of the 20 significantly less than zero at the 5% level. Furthermore, they are economically significant. If mutual funds randomly distributed their investments among small growth stocks that were prior year losers, we would expect about 4% of their money to be invested in the five worst stocks. The actual proportion is 1.54% less than the worst stocks’ weights in the characteristic portfolio. Mutual funds do especially well at stock picking among small growth stocks. The results of Table VII suggest that what funds do really well is avoiding the worst stocks in this group. Why would it be easier to find overpriced stocks to avoid than underpriced stocks to buy? It is possible that short-sale constraints result in overpriced stocks returning to their correct values more slowly than underpriced stocks, and thus that it is possible to identify more overpriced stocks than underpriced stocks.

D. Robustness Checks Throughout the paper, abnormal returns on stocks have been calculated by subtracting the value-weighted average return of other stocks in the same characteristic portfolio from the stock’s return. This could be an inadequate adjustment if, for example, the stocks that funds held differed in significant ways from the stocks that make up the characteristic portfolio. If mutual funds, for example, held only the largest of the small stocks, the return of a characteristic portfolio of small stocks could be an inadequate risk adjustment. To test for this, abnormal returns of fund holdings in each characteristic portfolio are 16

regressed on the Fama-French factors, the momentum factor, and the market risk premium. Each of these variables id obtained from Ken French’s website. Results for small growth stocks and for growth stocks in the second smallest quintile are shown in Table VII. The fourth column of the table shows the mean abnormal returns for holdings from these characteristic portfolios. These are the same as the abnormal returns in Table V. The next column provides the intercepts and their t-statistics from the regression of the abnormal returns on the market risk premium, the Fama-French factors and a momentum factor. The intercepts are very similar to the abnormal returns, and the t-statistics of the intercepts are very similar to the t-statistics of the abnormal returns. Inadequate risk adjustment does not explain the abnormal returns earned by small growth stocks held by mutual funds. I use matching firms as a second check on whether fund holdings earn abnormal returns after risk adjustment. Each month, each stock is assigned a matching stock of similar size, bookto-market, and prior 12-month returns. For each stock i and potential match m each month, I calculate the following match score: 2

Match Scorei ,m

⎛ BMi − BM m ⎞ ⎛ Sizei − Sizem ⎞ =⎜ ⎟ ⎟ +⎜ ⎝ ( BMi + BM m ) / 2 ⎠ ⎝ ( Sizei + Sizem ) / 2 ⎠ ⎛ Rtit − 12,t − 1 − Rtmt− 12 ,t − 1 ⎞ +⎜ ⎟ ⎝ ( Rtit − 12,t − 1 + Rtmt− 12 ,t − 1 ) / 2 ⎠

2

2

The matching stock for stock i is the stock with the lowest match score. The abnormal return for a stock is obtained by subtracting the return of its match from its return. It is possible for a stock to be a match for more than one stock during a particular month. It is also possible, but not necessary, for stock i to be a match for stock j during a month, and stock j to be a match for stock i. Abnormal returns are calculated for three months after a fund reports holding a stock using the same matched stock. If the match is delisted during the three months, I use the stock with the second (or third) lowest match score to calculate abnormal returns. Abnormal returns calculated using matching firms are shown for small growth stocks in the sixth column of Table VIII. We would expect these results to be slightly noisier than the

17

results using characteristic portfolios since we are comparing the return of each stock with the return of one other stock rather than a portfolio. Nevertheless, the matching firm results confirm that funds are able to find mispriced stocks among small growth companies. Abnormal returns calculated for fund holdings using the matched firm approach are positive for each of the ten small growth characteristic portfolios. T-statistics exceed 1.4 for nine of the ten portfolios, and exceed 2.0 for four portfolios. It is well known that small firms earn large returns in January (see Keim (1983)). I next recalculate average abnormal returns without Januaries. Results are show in the sixth column of Table VIII. Abnormal returns and t-statistics remain almost unchanged. The abnormal returns that funds earn on their small stock investments is not due to January. Finally, I break the sample period down into two subperiods: 1980 - 1992 and 1993 2005. Mutual funds earn positive abnormal returns on their investments in small growth stocks in both subperiods regardless of the stocks’ past returns. In the first subperiod, abnormal returns are significantly positive for five of the ten characteristic portfolios. In the second period, abnormal returns are significantly positive for seven of the ten characteristic portfolios.

E. Fund Type and the Abnormal Returns of Small Growth Stocks Throughout the paper, I use all funds that invest primarily in U.S. equities. This includes aggressive growth, growth, and growth and income funds. Growth and income funds are similar to value funds and may be expected to do poorly on investments in growth stocks or to avoid investing in them altogether. To see if this has an impact on results I calculate abnormal returns separately for aggressive growth, growth, and growth income funds. Results are shown in Table IX.

Panel A shows abnormal returns for stocks in the smallest quintile and lowest book-tomarket quintile. Panel B reports abnormal returns for stocks in the second smallest size quintile and the smallest book-to-market quintile. Results for holdings of stocks in the smallest size quintile and second lowest book-to-market quintile are shown in Panel C. Stocks in all three of these panels can be considered small growth stocks, and results are similar for all of them. Aggressive growth and growth funds seem to be able to identify small growth stocks that are 18

underpriced. This is indicated by the number of abnormal returns in boldface which indicates that the returns are significantly different from zero at the 5% level. It is possible that the statistical significance is lower for the growth and income funds because there are fewer of them. The point estimates of abnormal returns are also lower however. Growth and income funds do not, it seems, have the necessary skills for evaluating growth stocks. This raises the question of whether growth and income funds earn abnormal returns on their investments in value stocks. Panel D of Table IX reports abnormal returns for fund investments in small value stocks - that is stocks in the smallest size quintile and largest book-tomarket quintile. There is no evidence that any type of fund - aggressive growth, growth, or growth and income - is able to find underpriced securities among small value stocks.

F. Do Small Funds Earn Larger Returns Because they Invest in Small Growth Stocks?

Chen, Hong, Huang, and Kubik (2004) (CHHK) provide strong evidence that small mutual funds are better able to pick stocks than large funds. It is certainly plausible that small funds invest disproportionately in small stocks and large funds invest disproportionately in large stocks. This raises two questions. First, are the large abnormal returns earned by fund investments in small growth funds really due to the type of stock that is held, or is it due to a disproportionate amount of these stocks being held by smart small funds? Second, do small funds really do a better job of picking stocks, or is it just that they invest more in small growth stocks? To see if the superior performance of fund investments in small growth stocks is due to the type of fund that invests in these stocks, I split all funds each month into those with total assets above and below the median fund’s assets. I then recalculate abnormal returns for holdings by large and small funds of stocks in each of the characteristic portfolios. Results are reported in Table X. Panel A shows results for small growth stocks. Both large funds and small funds earn positive, statistically significant abnormal returns on their investments in small growth stocks. For example, small funds earn abnormal returns of 97 basis points per month on their investments in small growth stocks that were winners over the previous year. The t-statistic of 19

4.08 indicates that these abnormal returns are different from zero with a high degree of confidence. Large funds earn average abnormal returns of 110 basis points per month on their investments in the same stocks. The t-statistic for these abnormal returns is a highly significant 4.43. The abnormal returns funds earn on investments in small growth stocks is not due to the size of the funds that invest in these stocks. Panel B shows results for investments in large value stocks. Neither large nor small firms seem to be able to consistently earn abnormal returns from these investments. A different question is whether the superior performance of small funds over large funds is due to small funds overinvesting in small growth stocks. To examine this, I calculate the difference in abnormal returns between stocks held by funds that are smaller and larger than the median each month. I then take a time-series average and calculate t-statistics from the time series variance of the differences. Panel A of Table XI shows mean differences between the abnormal returns of stocks held by small and large fund. When a value-weighted average abnormal return is calculated using all stocks, small fund abnormal returns exceed large fund abnormal returns by 41 basis points per month. When fund abnormal returns are compared for investments in stocks in the two largest size quintilies and two highest book-to-market quintiles, as shown in the second column of Panel A, small funds earn abnormal returns that are 24 basis points per month larger than the abnormal returns of large funds. The t-statistic of 3.25 indicates that the return difference is statistically significant at the 1% level. The finding that small funds outperform large funds on their investments in large value stocks indicates that the differences in returns between the two fund type is not a result of different investment strategies. Finally, the last column of Panel A reports the mean difference in returns between small and large funds for their investments in stocks that are in both the two smallest size portfolios and the two lowest book to market portfolios. Small fund abnormal returns for these small growth stocks exceed the abnormal returns of large funds by a highly significant 78 basis points per month. Panels B and C of Table XI report differences in abnormal returns between small and large funds for investments in stock in specific characteristic portfolios. Panel B reports results forthe 25 characteristic portfolios of stocks in the lowest book-to-market quintile. Small funds outperform large funds in 22 of the 25 characteristic portfolios. The difference is positive and statistically significant at the 5% level for investments in six of the 25 characteristic portfolios. 20

Small fund investments outperform large fund investment for both large and small growth stocks. Panel C reports differences in abnormal returns for the 25 characteristic portfolios in the quintile of highest book-to-market values. Small funds earn higher returns than large funds for investments in stocks in 24 of the 25 characteristic portfolios. Differences are statistically significant at the 5% level for 3 of the 25 characteristic portfolios. To summarize, the abnormal returns earned by mutual funds on their investments in small growth stocks, and the larger returns earned by small funds than large funds are separate effects.

III. Conclusions

Mutual funds seem to be able to pick undervalued stocks. This paper examines whether mutual funds are particularly adept at finding mispricing among particular types of stocks. There are good reasons to believe that the analytical skills of mutual funds would be especially good at rooting out bargains among small growth stocks. Small stocks are likely to be followed by few analysts and familiar to few sophisticated investors. Mispricing is more likely to occur among these stocks. Growth stock values depend on cash flows in the far distant future, making them difficult to evaluate. These cash flows often depend on successful investments in new and untried technologies, on successful patent applications, or on expansion into new product lines or geographic areas. Mutual funds, who employ knowledgeable and skillful analysts, should be particularly adept at finding underpriced securities among growth stocks. The intuition that mutual funds are most likely to find mispriced securities among small growth stocks is confirmed here. Mutual fund investments in small growth stocks with large returns over the previous 12 months outperform their characteristic portfolios by a highly significant 1.06% per quarter. Mutual fund investments in small growth stocks with less favorable prior years returns outperform their characteristic portfolios by between 56 and 73 basis points per month. Investments in other stocks also seem to outperform their characteristic portfolios, but the evidence for stock-picking ability is much weaker outside of small growth 21

stocks. Small mutual funds have been shown to outperform large ones, and the results of this paper suggest that the difference may be due to the stocks they hold. I find that both large mutual funds and small ones seem to be able to pick undervalued stocks among small growth firms. Hence the strong performance of mutual fund investments in small growth stocks are not an artifact of the type of fund making the investment. At the same time, small funds earn more on their investments than large funds for small growth stocks and large value stocks. Hence the superior performance of small mutual funds over large mutual funds is not due to differences in the types of stocks that they hold.

22

References Baker, Malcolm, Lubomir Litov, Jessica Wachter, and Jeffrey Wurgler, 2005, Can mutual fund managers pick stocks? Evidence from their trades prior to earnings announcements, Working paper, New York University. Baks, Klaas, Jeffery Busse, and T. Clifton Green, 2006, Fund managers who take big bets: Skilled or Overconfident?, Working paper, Emory University. Berk, Jonathan and Richard Green, 2004, Mutual fund flows and performance in rational markets, Journal of Political Economy 112, 1269-1295. Carhart, Mark, 1997, On persistence in mutual fund performance, Journal of Finance 52, 57-82. Chen, Joseph, Harrison Hong, Ming Huang, and Jeffrey Kubik, 2004, Does fund size erode mutual fund performance? The role of liquidity and organization, American Economic Review 94, 1276-1302. Cohen, Randolph, Joshua Coval, and Lubos Pastor, 2005, Judging fund managers by the company they keep, Journal of Finance 60, 1057-1096. Coval, Joshua, and Tobias Moskowitz, 1999, Home bias at home: Local equity preference in domestic portfolios, Journal of Finance 54, 2045-2073. Cremers, Martijn, and Antti Petajisto, 2006, How active is your fund manager? A new measure that predicts performance, Working paper, Yale University. Daniel, Kent, Mark Grinblatt, Sheridan Titman, and Russ Wermers, 1997, Measuring mutual fund performance with characteristic-based benchmarks, Journal of Finance 52, 1035-1058. Frazzini, Andrea, 2005, The disposition effect and underreaction to news, Journal of Finance, forthcoming. Grinblatt, Mark, and Sheridan Titman, 1989, Mutual fund performance: An analysis of quarterly portfolio holdings, Journal of Business 62, 393-416. Grossman, Sanford, and Joseph Stiglitz, 1980, On the impossibility of informationally efficient markets, American Economic Review 70, 393-408. Hong, Harrison, Jeffrey Kubik, and Jeremy Stein, 2003, Thy Neighbor’s Portfolio: Word-ofmouth effects in the holdings and trades of money managers, Working paper, Harvard University. Kacperczyk, Marcin, C. Sialm, and L. Zheng, 2005, On industry concentration of actively 23

managed equity mutual funds, Journal of Finance 60, 1983-2011. Kacperczyk, Marcin, C. Sialm, and L. Zheng, 2005, Unobserved actions of mutual funds, University of British Columbia working paper. Keim, Donald, 1983, Size-related anomalies and stock return seasonality: Further empirical evidencce, Journal of Financial Economics 12, 13-32. Wermers, Russ, 2000, Mutual funds performance: An emprical decomposition into stock-picking talent, style, transactions costs, and expenses, Journal of Finance 40, 1655 - 1695.

24

Table I. Characteristic Portfolio Descriptions. Time series averages of monthly values over 12.1979 through 12/2005. Panel A. Size, book-to-market, and momentum quintiles Average Market Capitalizations ($ millions) Large Firms

2

3

4

Small Firms

13,086.6

1,795.3

699.6

289.3

50.6

Average Book-to-Market Ratios Value

2

3

4

Growth

1.59

0.70

0.44

0.25

0.03

Average Returns Last 12 Months Winners

2

3

4

Losers

100.8%

38.0%

15.9%

-2.5%

-27.9%

Average Number of Stocks Per Characteristic Portfolio Large Firms

2

3

4

Small Firms

13

15

20

30

129

Panel B. Average Sizes for Size and Book-to-Market Portfolios Large

2

3

4

Small

Value

8,398.1

1,748.2

696.1

281.5

29.4

2

11,983.5

1,832.0

690.3

288.3

45.5

3

13,991.0

1,783.1

695.9

287.9

55.8

4

14,806.8

1,774.9

701.8

294.4

61.1

Growth

16,253.8

1,838.4

714.0

294.3

61.1

25

Panel C. Average Book-to-Market Ratios for Size and Book-to-Market Portfolios Large

2

3

4

Small

Value

1.030

1.172

1.194

1.327

3.245

2

0.567

0.629

0.604

0.660

1.045

3

0.376

0.389

0.385

0.430

0.635

4

0.235

0.215

0.225

0.253

0.346

Growth

0.099

0.071

0.062

0.062

-0.144

26

Table II. Fund Holdings of Stocks by Characteristic Portfolio. Each NYSE, Amex, and Nasdaq stock is assigned to one of 125 characteristic portfolios. These portfolios are obtained by dividing stocks into quintiles by size, dividing each size quintile into book-to-market quintiles, and dividing each of the resulting 25 portfolios into five based on stock returns over the previous year. For this table, all momentum portfolios are combined within a given size and book-to-market quintile combination. The proportion and number of funds holding stocks in each category and the number of stocks not held by funds are calculated monthly. Time series averages are shown below for March 1980 through December 2005. Panel A. Average proportion of sample funds holding category stocks. Large

2

3

4

Small

Value

0.0748

0.0306

0.0174

0.0096

0.0014

2

0.0820

0.0335

0.0184

0.0097

0.0020

3

0.0859

0.0345

0.0200

0.0110

0.0022

4

0.0957

0.0365

0.0218

0.0123

0.0024

Growth

0.1040

0.0401

0.0238

0.0118

0.0019

Panel B. Average number of funds holding category stocks Large

2

3

4

Small

Value

68.1

28.4

16.8

9.3

1.4

2

77.7

31.7

17.6

9.2

1.8

3

84.4

32.5

19.1

10.3

2.1

4

89.6

34.0

20.1

11.3

2.2

Growth

96.4

36.9

21.5

10.9

1.9

Panel C. Average proportion of category stocks not held by any funds. Large

2

3

4

Small

Value

0.0033

0.0090

0.0300

0.0833

0.6036

2

0.0018

0.0083

0.0253

0.0839

0.5270

3

0.0056

0.0079

0.0324

0.0688

0.4974

4

0.0037

0.0078

0.0235

0.0555

0.5005

Growth

0.0016

0.0076

0.0308

0.0902

0.5836

Table III. Differences between the return on stocks held by mutual funds and returns on the value-weighted 27

market over the following month. Value of all positions by all mutual funds classified as growth, aggressive growth, or growth and income are used to determine total value-weighted return for mutual funds. T-statistics are based on the time series of differences. Equal-weighted differences between fund returns and market or characteristic portfolios returns value-weight securities held by each fund and equal-weight each fund. Value-weighted differences between fund returns and market or characteristic portfolios returns value-weight securities held by each fund and valueweight each fund. Each stock is assigned to one of 125 characteristic portfolios. These portfolios are obtained by dividing stocks into quintiles by size, dividing each size quintile into book-tomarket quintiles, and dividing each of the resulting 25 portfolios into five based on stock returns over the previous year Panel A: Equal-weighted averages of differences between fund returns and market returns. Returns Funds - Returns Value Weighted Market

T-statistic for Difference

All Months 4/80 - 12/05

0.0020

3.35

Through 1992

0.0019

2.18

After 1992

0.0021

2.55

Omitting 4th Qtr 1999 Outlier

0.0017

3.05

Panel B: Value-weighted averages of differences between fund returns and market returns. Returns Funds - Returns Value Weighted Market

T-statistic for Difference

All Months 4/80 - 12/05

0.0017

3.86

Through 1992

0.0014

2.04

After 1992

0.0021

3.51

Omitting 4th Qtr 1999 Outlier

0.0015

3.47

28

Panel C: Equal-weighted averages of the differences between fund returns and characteristic portfolio returns. Returns Funds Characteristic Port. Returns

T-statistic for Difference

All Months 4/80 - 12/05

0.0012

3.89

Through 1992

0.0010

2.35

After 1992

0.0013

3.17

Omitting 4th Qtr 1999 Outlier

0.0010

3.50

Panel D: Value-weighted averages of the differences between fund returns and characteristic portfolio returns. Returns Funds Characteristic Port. Returns

T-statistic for Difference

All Months 4/80 - 12/05

0.0011

3.79

Through 1992

0.0007

1.71

After 1992

0.0015

3.71

Omitting 4th Qtr 1999 Outlier

0.0010

3.40

29

Table IV. Fund Abnormal Returns by Size, Book-to-Market, and Momentum Quintiles Abnormal returns are calculated for each stock over each month by subtracting the return of one of 125 characteristic portfolio, matched on size, book-to-market, and returns over the previous 12 months from the return of the stock. Abnormal returns for a particular month are obtained by taking a value-weighted average of the abnormal returns of each stock, where the weights are the total amounts owned by mutual funds. I then calculate a time series average across months of abnormal returns of stocks in different size, book-to-market, and momentum quintiles. Large

2

3

4

Small

0.0007 (2.59)

0.0018 (3.82)

0.0022 (3.99)

0.0032 (4.48)

0.0037 (4.82)

Value

2

3

4

Growth

0.0004 (1.00)

0.0006 (1.46)

0.0013 (3.15)

0.0014 (3.13)

0.0016 (3.10)

Winners

2

3

4

Losers

0.0025 (3.55)

0.0014 (2.98)

0.0012 (3.08)

0.0003 (0.69)

-0.0003 (-0.50)

30

Table V. Mean monthly abnormal returns for fund positions in stocks in different characteristic portfolios. Abnormal returns are calculated for each stock each month by subtracting the return of one of 125 characteristic portfolio, matched on size, book-to-market, and returns over the previous 12 months from the return of the stock. Abnormal returns for a particular month are obtained by taking a value-weighted average of the abnormal returns of each stock, where the weights are the total amounts owned by mutual funds. I then calculate a time series average across months of abnormal returns of stocks in characteristic portfolios. Panel A. Large Stocks Winners

2

3

4

Losers

Value

0.0030 (2.63)

-0.0001 (-0.07)

0.0005 (0.54)

-0.0002 (-0.18)

-0.0006 (-0.48)

2

0.0006 (0.52)

-0.0001 (-0.12)

0.0011 (1.26)

0.0003 (0.29)

-0.0018 (-1.52)

3

0.0012 (1.08)

0.0012 (1.07)

0.0018 (1.86)

-0.0001 (0.90)

0.0006 (0.51)

4

0.0033 (2.18)

0.0016 (1.44)

-0.0001 (-0.07)

0.0010 (0.90)

-0.0011 (-0.88)

Growth

0.0032 (2.00)

0.0018 (1.44)

0.0010 (0.92)

-0.0002 (-0.16)

-0.0002 (-0.19)

Panel B. Second Largest Size Quintile Winners

2

3

4

Losers

Value

-0.0010 (-0.83)

0.0018 (1.41)

0.0019 (1.63)

-0.0007 (-0.51)

-0.0001 (-0.05)

2

0.0033 (2.32)

0.0034 (2.81)

0.0005 (0.46)

-0.0003 (-0.29)

0.0011 (0.81)

3

0.0017 (1.18)

0.0026 (2.03)

0.0026 (1.84)

0.0006 (0.50)

0.0006 (0.39)

4

0.0045 (2.83)

0.0035 (1.95)

0.0022 (1.64)

0.0005 (0.34)

-0.0009 (-0.60)

Growth

0.0060 (3.16)

0.0005 (0.29)

0.0025 (1.74)

0.0010 (0.66)

-0.0009 (-0.54)

31

Panel C. Third Largest Size Quintile Winners

2

3

4

Losers

Value

0.0007 (0.45)

0.0010 (0.73)

-0.0010 (-0.86)

-0.0017 (-1.17)

-0.0023 (-1.44)

2

0.0039 (2.22)

0.0017 (1.38)

0.0036 (2.76)

0.0000 (0.02)

0.0018 (1.03)

3

0.0007 (0.45)

0.0012 (0.84)

0.0019 (1.31)

0.0010 (0.75)

0.0028 (1.60)

4

0.0069 (3.49)

0.0028 (1.88)

0.0006 (0.35)

0.0037 (2.53)

0.0026 (1.40)

Growth

0.0061 (3.50)

0.0070 (4.16)

0.0038 (2.71)

0.0002 (0.15)

0.0003 (0.15)

Panel D: Second Smallest Size Quintile Winners

2

3

4

Losers

Value

-0.0001 (-0.05)

-0.0011 (-0.88)

-0.0002 (-0.15)

0.0006 (0.35)

0.0015 (0.75)

2

0.0026 (1.66)

0.0018 (1.28)

0.0019 (1.34)

-0.0006 (-0.42)

0.0023 (1.21)

3

0.0019 (1.10)

0.0035 (2.30)

0.0008 (0.60)

0.0006 (0.43)

0.0030 (1.75)

4

0.0042 (2.40)

0.0050 (3.13)

0.0050 (3.74)

0.0034 (2.18)

0.0033 (1.91)

Growth

0.0095 (3.55)

0.0049 (3.13)

0.0090 (2.20)

0.0063 (3.38)

0.0035 (1.82)

32

Panel E. Small Stocks Winners

2

3

4

Losers

Value

0.0014 (0.81)

-0.0005 (-0.29)

-0.0005 (-0.23)

0.0043 (1.55)

0.0046 (1.24)

2

-0.0017 (-1.11)

0.0005 (0.37)

0.0004 (0.29)

0.0034 (1.93)

0.0011 (0.46)

3

0.0031 (1.27)

0.0040 (3.25)

0.0005 (0.37)

0.0034 (2.08)

0.0039 (1.61)

4

0.0075 (3.75)

0.0032 (2.56)

0.0022 (1.35)

0.0062 (3.45)

0.0072 (2.79)

Growth

0.0106 (4.85)

0.0059 (3.59)

0.0056 (2.88)

0.0073 (3.41)

0.0066 (1.84)

33

Table VI. Abnormal Returns of Stock in Various Characteristic Portfolios the 1st, 2nd, and 3rd Month after Holdings are Published. Abnormal returns are calculated for each stock each month by subtracting the return of one of 125 characteristic portfolio, matched on size, book-to-market, and returns over the previous 12 months from the return of the stock. Abnormal returns for a particular month are obtained by taking a value-weighted average of the abnormal returns of each stock, where the weights are the total amounts owned by mutual funds. I then calculate a time series average across months of abnormal returns of stocks in characteristic portfolios. Panel A. Abnormal Returns for the Small Stock Quintile for the 1st Month. Winners

2

3

4

Losers

Value

0.0095

0.0038

0.0147

0.0197

0.0243

2

0.0019

0.0093

0.0061

0.0146

0.0205

3

0.0103

0.0105

0.0111

0.0113

0.0121

4

0.0197

0.0089

0.0084

0.0173

0.0112

Growth

0.0189

0.0242

0.0142

0.0123

0.0155

Panel B. Abnormal Returns for the Small Stock Quintile for the 2nd Month. Winners

2

3

4

Losers

Value

-0.0034

0.0043

-0.0012

0.0031

0.0157

2

0.0068

-0.0035

0.0019

-0.0013

0.0063

3

0.0068

0.0079

0.0042

0.0082

0.0129

4

0.0168

0.0092

0.0042

0.0087

0.0099

Growth

0.0113

0.0146

0.0156

0.0151

0.0025

Panel C. Abnormal Returns for the Small Stock Quintile for the 3rd Month.. Winners

2

3

4

Losers

Value

0.0010

-0.0092

-0.0022

-0.0023

-0.0109

2

-0.0056

0.0023

-0.0021

-0.0033

-0.0071

3

-0.0018

-0.0035

-0.0004

-0.0011

0.0060

4

0.0033

0.0047

-0.0003

0.0058

0.0025

Growth

0.0075

0.0071

0.083

0.0060

-0.0006

34

Panel D. Abnormal Returns for the Second Smallest Quintile for the 1st Month. Winners

2

3

4

Losers

Value

0.0071

0.0048

0.0063

0.0004

0.0134

2

0.0084

0.0050

0.0072

0.0017

0.0119

3

0.0119

0.0045

0.0058

0.0094

0.0126

4

0.0153

0.0135

0.0099

00065

0.0210

Growth

0.0225

0.0134

0.0089

0.0066

0.0151

Panel E. Abnormal Returns for the Second Smallest Stock Quintile for the 2nd Month. Winners

2

3

4

Losers

Value

-0.0000

-0.0035

0.0034

0.0043

0.0053

2

0.0071

-0.0027

0.0028

0.0048

0.0064

3

0.0034

0.0024

0.0053

0.0012

0.0021

4

0.0085

0.0077

0.0097

0.0109

-0.0027

Growth

0.0130

0.0058

0.0080

0.0041

0.0115

Panel F. Abnormal Returns for the Second Smallest Stock Quintile for the 3rd Month. Winners

2

3

4

Losers

Value

-0.0022

0.0006

-0.0039

-0.0034

-0.0011

2

-0.0009

-0.0018

0.0011

0.0021

-0.0003

3

-0.0015

0.0075

-0.0026

-0.0024

0.0010

4

0.0034

0.0041

0.0029

0.0071

0.0017

Growth

0.0011

0.0025

0.0141

0.0044

0.0050

35

Table VII. Differences between the proportion of fund holdings in the five best (worst) stocks in characteristic portfolios and the weight of the stocks in the characteristic portfolio. Each month, the five best and worst performing stocks are identified for every characteristic portfolio in the two smallest size and two lowest book-to-market quintiles. The weight of these stocks in the characteristic portfolio is then subtracted from the total proportion of funds’ investment in the characteristic portfolio that is in the stocks. A time series average difference is calculated using the months from March 1980 through 2005. T-statistics are based on the standard deviation of the monthly differences. Panel A. Difference between percentage of top five performing stocks held by funds and their percentage of the characteristic portfolio. Size

Book to Market

4

Winners

2

3

4

Losers

4

0.0112 (1.34)

0.0410 (4.03)

0.0331 (3.55)

0.0170 (1.78)

0.0297 (2.66)

4

Growth (5)

0.0284 (2.64)

0.0323 (3.28)

0.0156 (1.79)

0.0073 (0.80)

0.0135 (1.34)

Small (5)

4

0.0086 (1.39)

-0.0009 (-0.23)

-0.0022 (-0.66)

0.0016 (0.33)

-0.0031 (-0.89)

Small (5)

Growth (5)

-0.0010 (-0.22)

0.0052 (0.77)

-0.0055 (-1.35)

-0.0083 (-2.22)

-0.0041 (-1.22)

Panel B. Difference between percentage of bottom five performing stocks held by funds and their percentage of the characteristic portfolio Size

Book to Market

4

Winners

2

3

4

Losers

4

-0.0173 (-2.28)

-0.0179 (-2.14)

-0.0103 (-1.31)

-0.0096 (-1.12)

-0.0272 (-3.59)

4

Growth (5)

-0.0408 (-5.18)

-0.0313 (-4.25)

-0.0271 (-3.88)

-0.0232 (-3.07)

-0.0397 (-5.19)

Small (5)

4

-0.0074 (-2.63)

-0.0049 (-1.81)

-0.0023 (-0.63)

-0.0016 (-0.38)

-0.0065 (-1.48)

Small (5)

Growth (5)

-0.0154 (-5.56)

-0.0157 (-8.25)

-0.0102 (-2.45)

-0.0098 (-2.82)

-0.0115 (-2.36)

36

Table VIII. Robustness Checks Mean monthly abnormal returns to holdings of mutual funds, calculated in various ways, are reported along with t-statistics. Abnormal returns are calculated for individual stocks by subtracting the return on its characteristic portfolio from the stock return. For the base case, a value-weighted average abnormal return is calculated for mutual fund holdings each month. The weights are the total amount invested in each stock by all mutual funds that month. The base case incorporates returns for all months over 1980 - 2005. These abnormal returns are the same as those shown in Panels D and E of Table V. The intercept of abnormal returns is obtained by regressing monthly base case abnormal returns on the market premium, HML factor, SMB factor, and momentum factor from Ken French’s website. Matching firm excess returns are obtained by subtracting from stock returns the return of a stock matched on size, book-to-market, and momentum. The closest match of al CRSP stocks is used. Quintile

Monthly Abnormal Returns with T-statistics in Parentheses

Size

BM

Momentum

Base Case

Intercept of Abnormal Returns on FF Factors, Momentum, and Rm - Rf

Small

Growth

Winner

0.0106 (4.85)

0.0097 (4.18)

0.0091 (4.27)

0.0101 (4.67)

0.0115 (3.65)

0.0097 (3.21)

Small

Growth

2

0.0059 (3.59)

0.0064 (3.60)

0.0030 (1.58)

0.0056 (3.37)

0.0054 (2.20)

0.0067 (3.03)

Small

Growth

3

0.0056 (2.88)

0.0040 (1.94)

0.0042 (2.09)

0.0053 (2.76)

0.0052 (1.75)

0.0060 (2.42)

Small

Growth

4

0.0073 (3.41)

0.0056 (2.47)

0.0035 (1.46)

0.0072 (3.33)

0.0090 (2.73)

0.0060 (2.21)

Small

Growth

Loser

0.0066 (1.84)

0.0078 (2.09)

0.0133 (3.39)

0.0062 (1.74)

0.0067 (1.09)

0.0061 (1.70)

4

Growth

Winner

0.0095 (3.55)

0.0081 (2.92)

0.0042 (3.25)

0.0088 (3.25)

0.0090 (3.46)

0.0093 (2.00)

4

Growth

2

0.0049 (3.13)

0.0051 (3.10)

0.0007 (0.85)

0.0047 (3.00)

0.0045 (1.98)

0.0053 (2.43)

4

Growth

3

0.0090 (2.20)

0.0078 (1.83)

0.0064 (1.56)

0.0087 (2.14)

0.0037 (1.75)

0.0146 (1.85)

4

Growth

4

0.0063 (3.38)

0.0043 (2.25)

0.0027 (1.87)

0.0060 (3.24)

0.0034 (1.54)

0.0097 (3.26)

4

Growth

Loser

0.0035 (1.82)

0.0035 (1.77)

0.0019 (1.47)

0.0032 (1.67)

0.0022 (0.84)

0.0053 (1.87)

37

Matching Firm Excess Returns

No Januaries

1980 - 1992

1993 - 2005

Table IX. Abnormal Returns Earned by Funds with Different Investment Objectives Stocks are sorted into 125 characteristic portfolios based on five size quintiles, five book-tomarket quintiles, and five momentum portfolios. Abnormal returns are calculated for each stock each month by subtracting the value-weighted average return of other stocks in the characteristic portfolio from the stock return. An average abnormal return is calculated for each characteristic portfolio based on value-weighting of holdings across all funds with a given investment objective. A time series average abnormal return is calculated for each characteristic portfolio and fund investment objective using all months from March 1980 through December 2005. Time series averages for selected characteristic portfolios are shown below. Averages in bold are significantly different from zero at the 5% level. Panel A: Small Growth Stocks (5th Size Quintile, and 5th Book-to-Market Quintile) Winner

2

3

4

Loser

Aggressive Growth

0.0129

0.0050

0.0034

0.0096

0.0007

Growth

0.0106

0.0058

0.0069

0.0067

0.0109

Growth and Income

0.0027

0.0078

0.0013

0.0045

0.0019

Panel B: Second Smallest Growth Stocks (4th Size Quintile, and 5th Book-to-Market Quintile) Aggressive Growth

0.0098

0.0048

0.0141

0.0072

0.0047

Growth

0.0092

0.0059

0.0041

0.0064

0.0044

Growth and Income

0.0069

-0.0070

0.0088

0.0030

0.003

Panel C: Small Stocks (5th Size Quintile) Second Lowest Book-to-Market Quintile Aggressive Growth

0.0087

0.0030

-0.0012

0.0040

0.0083

Growth

0.0081

0.0030

0.0037

0.0075

0.0087

Growth and Income

0.0038

0.0013

-0.0049

-0.0025

0.0049

Panel D: Small Value Stocks (5th Size Quintile, and 1st Book-to-Market Quintile) Aggressive Growth

-0.0032

-0.0007

0.0028

0.0080

0.0042

Growth

0.0019

-0.0010

0.0004

0.0047

0.0060

Growth and Income

0.0077

-0.0059

-0.0056

-0.0019

0.0024

38

Table X. Comparison of monthly abnormal returns for large and small funds. Abnormal returns are calculated for each stock each month by subtracting the return on the characteristic portfolio composed of stocks with similar size, book-to-market ratios and returns over the previous 12 months. A weighted average of the abnormal stock returns is calculated using the total dollar investment by mutual funds as weights. An equal-weighted average of the quarterly results is then calculated. T-statistics in parentheses are based on the time-series standard error. Panel A. Abnormal returns from investing in small growth firms. Funds Smaller than Median

Funds Larger than Median

Book-to-Market Quintile

Return Last 12 Months Quintile

Mean Ab. Return

T-statistic

Mean Ab. Ret.

T-statistic

Size Quintile Smallest

Lowest (Growth)

Highest

0.0097

4.08

0.0110

4.43

Smallest

Lowest (Growth)

2

0.0060

3.01

0.0058

3.15

Smallest

Lowest (Growth)

3

0.0049

2.19

0.0059

2.89

Smallest

Lowest (Growth)

4

0.0056

1.95

0.0083

3.69

Smallest

Lowest (Growth)

Lowest

0.0100

2.22

0.0053

1.35

4

Lowest (Growth)

Highest

0.0081

3.20

0.0096

3.43

4

Lowest (Growth)

2

0.0036

1.75

0.0050

3.05

4

Lowest (Growth)

3

0.0046

2.77

0.0095

2.16

4

Lowest (Growth)

4

0.0066

3.18

0.0061

3.17

4

Lowest (Growth)

Lowest

0.0040

1.70

0.0033

1.67

39

Panel B. Abnormal returns from investing in large value firms. Funds Smaller than Median

Funds Larger than Median

Mean Ab. Return

T-statistic

Mean Ab. Ret.

Highest

0.0021

1.90

0.0030

2.66

Highest (Value)

2

0.0005

0.42

-0.0001

-0.05

Largest

Highest (Value)

3

0.0011

1.20

0.0005

0.53

Largest

Highest (Value)

4

-0.0003

-0.31

-0.0002

-0.18

Largest

Highest (Value)

Lowest

-0.0011

-0.78

-0.0006

-0.47

2

Highest (Value)

Highest

0.0001

0.07

-0.0011

-0.92

2

Highest (Value)

2

0.0045

3.58

0.0016

1.24

2

Highest (Value)

3

0.0013

1.07

0.0019

1.60

2

Highest (Value)

4

-0.0006

-0.42

-0.0007

-0.53

2

Highest (Value)

Lowest

0.0019

0.83

-0.0005

-0.27

Size Quintile

Book-to-Market Quintile

Largest

Highest (Value)

Largest

Return Last 12 Months Quintile

40

T-statistic

Table XI. Proportion of total money invested in U.S. stocks that is put into various portfolios by funds with assets greater than and less than the median. REDO Panel A. Small growth stocks. Proportion of Money Invested Return Last 12 Months Quintile

Funds Smaller than Median

Funds Larger than Median

Lowest (Growth)

Highest

0.0032

0.0010

Smallest

Lowest (Growth)

2

0.0027

0.0012

Smallest

Lowest (Growth)

3

0.0018

0.0008

Smallest

Lowest (Growth)

4

0.0012

0.0006

Smallest

Lowest (Growth)

Lowest

0.0005

0.0002

Smallest

4

Highest

0.0037

0.0016

Smallest

4

2

0.0039

0.0020

Smallest

4

3

0.0032

0.0018

Smallest

4

4

0.0024

0.0013

Smallest

4

Lowest

0.0013

0.0007

4

Lowest (Growth)

Highest

0.0048

0.0026

4

Lowest (Growth)

2

0.0048

0.0031

4

Lowest (Growth)

3

0.0048

0.0032

4

Lowest (Growth)

4

0.0038

0.0027

4

Lowest (Growth)

Lowest

0.0026

0.0019

4

4

Highest

0.0048

0.0031

4

4

2

0.0043

0.0032

4

4

3

0.0044

0.0033

4

4

4

0.0039

0.0031

4

4

Lowest

0.0033

0.0028

Size Quintile

Book-to-Market Quintile

Smallest

41

Panel B Large value stocks. Proportion of Money Invested Return Last 12 Months Quintile

Funds Smaller than Median

Funds Larger than Median

Highest (Value)

Highest

0.0113

0.0131

Largest

Highest (Value)

2

0.0112

0.0130

Largest

Highest (Value)

3

0.0127

0.0140

Largest

Highest (Value)

4

0.0111

0.0127

Largest

Highest (Value)

Lowest

0.0102

0.0118

Largest

2

Highest

0.0144

0.0178

Largest

2

2

0.0147

0.0172

Largest

2

3

0.0167

0.0190

Largest

2

4

0.0161

0.0180

Largest

2

Lowest

0.0145

0.0158

2

Highest (Value)

Highest

0.0043

0.0047

2

Highest (Value)

2

0.0044

0.0046

2

Highest (Value)

3

0.0049

0.0051

2

Highest (Value)

4

0.0042

0.0043

2

Highest (Value)

Lowest

0.0042

0.0045

2

2

Highest

0.0059

0.0061

2

2

2

0.0053

0.0053

2

2

3

0.0057

0.0059

2

2

4

0.0052

0.0054

2

2

Lowest

0.0050

0.0053

Size Quintile

Book-to-Market Quintile

Largest

42

Table XI. Abnormal Returns of Small Funds Minus Abnormal Returns of Large Funds Returns of all stocks within a fund are value-weighted, and the abnormal returns of individual funds are value-weighted to produce and abnormal return for small funds and large funds. Small have assets less than the median, while large have assets greater than the median. Abnormal returns are calculated by subtracting the return on a stock’s characteristic portfolio from the return of the stock. Panel A.. All Stocks

Large Value Stocks

Small Growth Stocks

0.0041 (7.41)

0.0024 (3.25)

0.0078 (5.12)

Panel B. Growth Stocks Winner

2

3

4

Loser

Large

0.0041 (1.54)

0.0034 (1.89)

0.0006 (0.33)

0.0021 (1.35)

0.0044 (2.20)

2

0.0003 (0.07)

0.0085 (2.96)

-0.0037 (-1.49)

0.0049 (1.81)

0.0079 (2.52)

3

0.0074 (2.28)

-0.0033 (-0.94)

0.0037 (1.05)

0.0010 (0.28)

0.0052 (1.52)

4

0.0103 (1.98)

0.0049 (1.12)

-0.0002 (-0.04)

0.0042 (0.98)

0.0054 (1.36)

Small

0.0075 (1.64)

0.0035 (0.57)

0.0053 (1.07)

0.0045 (0.90)

0.0280 (2.08)

Winner

2

3

4

Loser

Large

0.0005 (0.28)

0.0024 (1.38)

0.0021 (1.29)

0.0017 (0.93)

0.0004 (0.16)

2

0.0046 (1.86)

0.0039 (1.55)

0.0030 (1.08)

0.0023 (0.96)

0.0058 (1.67)

3

0.0001 (0.03)

0.0012 (0.53)

0.0003 (0.12)

0.0091 (2.85)

0.0042 (1.06)

4

0.0052 (1.49)

0.0002 (0.05)

0.0028 (0.89)

0.0102 (2.68)

0.0126 (2.19)

Small

0.0001 (0.03)

-0.0004 (-0.11)

0.0079 (1.66)

0.0114 (1.54)

0.0160 (1.65)

Panel C. Value Stocks

43

Fig. 1a. Number of aggressive growth, growth and income, and growth funds each month.

1400 1200 1000 800 600 400

44

20040131

20010131

19980131

19950131

19920131

19890131

19860131

0

0

19830131

200

Fig.1b. Value of U.S. shares held by aggressive growth, growth and income, and growth funds filing each month in billions of dollars.

1400 1200 1000 800 600 400

45

20040131

20010131

19980131

19950131

19920131

19890131

19860131

0

0

19830131

200