Momentum Investment Strategies, Portfolio Performance, and Herdinj Study of Mutual Fund Behavior Mark Grinblatt, Sheridan Titman, Russ Wermers The American Economic Review, Volume 85, Issue 5 (Dec., 1995), 1088-1105. Your use of the ISTOR database indicates your acceptance of JSTOR's Terms and Conditions of Use. A copy of ISTOR’s Terms and Conditions of Use is available at hup://www,jstor-org/abouvterms.himl, by contacting JSTOR at jstor-info@umich.edu, or by calling JSTOR at (888)388-3574, (734)998-9101 or (FAX) (734)998.9113. No part of a JSTOR transmission may be copied, downloaded, stored, further transmitted, transferred, distributed, altered, oF otherwise used, in any form or by any means, except: (1) one stored electronic and one paper copy of any article solely for your personal, non-commercial use, or (2) with prior written permission of JSTOR and the publisher of the article or other text Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission, The American Economic Review is published by American Economic Association, Please contact the publisher for further permissions regarding the use of this work. Publisher contact information may be obtained at bup:/svww jstor.org/journals/aca. hia. The American Economic Review (©1995 American Economic Association ISTOR and the ISTOR logo are trademarks of ISTOR, and are Registered in the U.S. Patent and Trademark Office. For more information on JSTOR contact jstor-info@umich.edu, (©2001 JSTOR upswww jstor.org/ Mon Ape 23 16:05:46 2001Momentum Investment Strategies, Portfolio Performance, and Herding: A Study of Mutual Fund Behavior By MARK GRINBLATT, SHERIDAN TITMAN, AND Russ WERMERS* This study analyzes the extent to which mutual funds purchase stocks based on their past returns as well as their tendency to exhibit “herding” behavior (ise. buying and selling the same stocks at the same time). We find that 77 percent of the mutual funds were ‘momentum investors,” buying stocks that were past winners; however, most did not systematically sell past losers. On average, funds that invested on momentum realized significantly better performance than other funds, We also find relatively weak evidence that funds tended to buy and sell the same stocks at the same time. (JEL G14, G23) ‘The amount of wealth managed by institu- tional investors has grown considerably over the past 20 years. Due perhaps to decreased trading costs, brought about by the termination of fixed commissions in May 1975, these in- stitutional investors have become much more active traders and, as a result, have become increasingly important in terms of setting mar- ket prices.' The growing influence of institu- tional investors has led to increased scrutiny both by policymakers and by journalists, who tend to believe that these investors trade ex- cessively and move in and out of stocks in a herd-like manner. This tendency to invest with the herd, in combination with the alleged ten- dency of institutions to follow momentum- based fads by buying past winners and selling past losers is of concern, since this behavior could potentially exacerbate stock- price volatility. Momentum trading strategies and herding, behavior are also used by academics to mo- tivate models of seemingly irrational mar- kets. Fischer Black (1986) and Brett Trueman (1988) provide reasons why insti- tutional investors may trade excessively, and ‘a number of recent theory papers provide ra- tionales to explain why institutional inves- tors would analyze the same groups of stocks and trade in the same direction.’ In addition, J. Bradford De Long et al. (1990) describe ‘what they call “‘positive-feedback traders,” ‘who have a tendeney to buy stocks after they * Grinblat: John E. Anderson Graduate School of “Management, University of Califia, Los Angeles, Box 951481, Lox Angeles, CA 90095-1481; Titman: Depart rent of Finance, Caroll School of Management Beston College, Chestnut Hil, MA 02167; Wermers: Division of Finance and Economics, Graduate School of Businss and ‘Administration, University of Colorado, Boulder, CO 80309. The authors are grateful Narasimhan Jegaesh, a5 well aso seminar participants at National Taiwan Uni ‘verity, National Chinese University, Osaka University University of California at Berkeley, University of Chi 1g9, Stanford University, UCLA, University of Tenas, and Yale University for comments on earlier drat "Iastutional holdings are now about 50 percent of to tal equity holdings the United Stats, while institutional rang, when added to member trading, accounted for bout 70 percent of total NYSE volume in 1989 (Rober ‘Schwartz and James Shapiro, 1992) 1038 perform well. Our study provides empirical evidence on the trading patterns of fund managers by ex- amining the quarterly holdings of 155 mutual funds over the 1975-1984 period. We char- 2 These papers include Robert J. Shiller and John Pound (1989), Michael Breanan (1980), David S. Schart- stein and Jeremy C. Stein (1990), Josef Lakonishok et (1991), Abhgit Banerjee (1992), Sushil Bikhehandani et 1 (1992), Kenneth A. Froot etal (1992), and David Hirsleiferet al (1994),VoL. 85 No.5 acterize the portfolio choices of these funds to determine the extent to which they pur- chase stocks based on their past returns and the extent to which they “herd,” that is, the extent to which the group either predomi: nantly buys or predominantly sells the same stock at the same time. We then examine the extent to which herding and momentum in- vesting affect the performance of the funds." If either irrationality or agency problems gen- crate these trading styles (as discussed, for example, by Scharfstein and Stein [1990]), then mutual funds that exhibit these behaviors will tend to push the prices of stocks that they purchase above intrinsic values, thereby re- alizing lower future performance. However, if this type of behavior arises because in- formed portfolio managers tend to pick the same underpriced stocks, then funds that ex- hibit these styles should realize high future performance. Our analysis of momentum investing and performance is also motivated by two previous studies (Grinblatt and Titman, 1989a, 1993), which indicate that, atleast before transaction costs, a number of mutual funds eared sig- nificant risk-adjusted abnormal returns. This ‘observed performance is not related to known anomalies that involve cross-sectional differ- cences in expected returns, like the small-firm effect. However, before we conclude that these abnormal retums are generated by either su- perior information or analysis, we would also Tike to rule out the possibility that the observed abnormal performance was generated by ex- ploiting. time-series anomalies. Specificall we would like to determine the extent to which “ein Friend etal (1970) were perhaps the ist to ‘examine the wading paters of mutual funds, They found, lamong other things that there was a tendency of some ‘mutual funds to follow the prior investment choices of their more soecesfal counterparts. Alan Kraus and Hans R Stoll (1972) examined the tendency of mia funds and bank rusts to buy an sel the sme stock the same time but didnot ind evidence of herding beyond that due to-chance. Lakonishok et al (1992) examined the amount ‘of herding exhibited by pension fund managers. They found only weak evidence of the funds ether buying ot selling in herds (above chance oocurrenses) and weak relation between herding in stocks and the pat returns of the stocks GRINBLATT ET ALL: MOMENTUM INVESTMENT AND HERDING 1089 the observed performance was generated by the simple momentum strategy of buying past ‘winners and selling past losers, as described in Narasimhan Jegadeesh and Titman (1993). This simple strategy would generate abnormal performance with either of the Grinblatt and Titman (1989a, 1993) performance measures, as well as with any of the more traditional measures. ‘The paper is organized as follows. Section I describes the data, while Section IL de- scribes the methodology used to compute the degree of momentum (or contrarian) invest- ing behavior exhibited by a fund, Section III presents empirical results on momentum in- ‘vestment styles and performance. Section IV investigates the tendency of the funds to en- gage in herding behavior and also considers the relation of herding behavior to momen- tum investing and performance. Finally, Section V summarizes and concludes the paper. 1 Data Quarterly portfolio holdings for 274 mutual funds that existed on December 31, 1974, were purchased from CDA Investment Technolo- gies, Inc. of Silver Springs, Maryland. These ‘mutual fund data, used previously by Grinblatt and Titman (1989a, 1993) to examine fund performance, include 155 funds that existed during the entire 10-year time period of De- cember 31, 1974, to December 31, 1984.‘ Cen- ter for Research in Security Prices (CRSP) monthly returns for each NYSE- and AMEX- listed stock held by the funds were computed bby compounding returns in the CRSP daily re- tums file. Over-the-counter (OTC) stocks and fixed-income holdings were treated as missing values in a manner that we will describe shortly. “The analysis in Grnblat and Titman (1989) and Ste- phen J. Brow and William N. Goctzmann (1995) indi ‘ates that this fund survival requirement has ony small effect on inference tess of performance abilities. In out Inter analysis of the herding” of nds into individual stocks, we expand our sample to incl all 274 funds, Which includes nonsurvivors1090 THE AMERICAN ECONOMIC REVIEW Methodology A. The Momentum Measures ‘A momentum investor buys past “winners”” and sells past “losers."” A contrarian investor does the opposite. Our measure of momentum investing is CD MAREE OH DR ves ‘where W, i the portfolio weight on security j at date t, and Ry, i the return of security JG = 1...) from date 1 ~ k to date t= ‘k+ 1, the historical benchmark period. ‘This statistic is designed to measure the de- {gree to which a fund manager tilts is portfolio inthe direction of stocks that have experienced high returns in some historical benchmark pe- riod, and away from stocks that have experi- enced low returns. Since this measure equals the difference between two portfolio returns ‘uring the benchmark period, a positive mea- sure means that, on average, the fund's current portfolio had higher returns than the portfolio that the fund would have held had no portfolio revisions been made, Since the mutual fund holdings are only available quarterly, while stock returns. are available monthly, further modifications ofthe measure given by equation (1) are needed to aurive at the measures of momentum that are implemented. Given that we have 41 quarters of holdings, with three monthly retums per quarter, equation (1) is modified as follows:* 1 ; a BEES om var Q) om Since the most recent returns are probably of the greatest interest to portfolio managers, * Using monthly retums rather than quarterly retums reduces the problem of missing returns. For example. if fetums for the Bocing Corporation common stock are svailable fom the CRSP for January and Februar, Dut ‘ot Mare, then Boeing drops out of our momentum in ‘esting measur in only one observation out of 120 (using ‘monthly returns), instead of ene ou oF 40 (using quater returns) See aso Footnote 6 DECEMBER 1995 1 and k = 2 are the two measures that we will focus on, although we will present some re- sults for k > 2. We will refer to equation (2) as “lag-0 momentum’* (LOM) when k = 1, and as “‘lag-I_ momentum”? (LIM) when k B. Statistical Inference ‘As described by equation (2), the differ- enced portfolio weights were updated every calendar quarter, while the differenced port- folio return was generated each month.” This process resulted in a time series of 120 ‘monthly retum differences for each mutual fund. Ifthe return differences associated with these measures are serially uncorrelated under the null hypothesis of no momentum invest- ing, then inference-testng forthe significance of the measures is simple. Testing whether the momentum measure has a mean value of zero is identical to test of whether two given port- folios (with dynamic weight vectors) have the ‘same mean return, We employ many cross-sectional regres- sions in our analysis, mainly of fund perfor- mance on fund characteristics. Statistical significance cannot be inferred from the cross-sectional r and F statistics typically re- ported in such regressions, since the regression residuals are correlated across mutual funds. ‘Thus, we use alternative rand F tests that are derived from a time-series procedure (see GGrinblatt and Titman, 1994) ‘The diferenced weights ae identical for any three ‘months inthe same calendar quater, except when eum data are not available for one oF more scans fn some {Gut no al) ofthe tree months. For example, f returns for the Bocing Corporation are avalale in January and Februar, bu notin March, then the iferenced portfolio ‘weight Boeing is set to ero only for March, and iis ‘entcal for January and February "Since most portfolios of imerest, such a5 value- weighted portfolios, have changing weights the ordinary ests that are usualy applied i these ests te technically ‘appropiate. However if secunties returns are scilly tincorrelated the cenal-imit theorem san be applied and asymptotic : tests apd chi-square tess are valid for noo- ‘normal portfolio etums. Given the length of our time se fies, thete asymptotic test tatstcs are veal identical tothe and Fsatistice used here and ave negligibly dif- {erent significance level,VOL. 85 NO. 5 C. Modifying the Measures to Eliminate “Passive Momentum Investing" ‘The portfolio weights of winning (losing) stocks increase (decrease) even if the number Of shares held stays constant. In this ease, the lag-0 momentum measure would indicate momentum investing for buy-and-hold in- vestment strategies. To correct this, we cal- culate the weights using the average of the beginning- and end-of-quarter share prices." For consistency, we make the same modifi- cation for all momentum measures, even though passive momentum investing affects only the lag-O momentum measure.” D. Extensions of the Measure We will also use decomposed versions of the LOM measure, called “Buy LOM” and “Sell LOM.” A high Buy (Sell) LOM measure for a fund means that it bought winners (sold losers) strongly, on average. These two measures are the decomposition of equation (2) into partial sums: (Ga) Buy Lone DEL Oi Bar Rd (3b) sett Lom " Ifthe beginning-f-quarer pice was not available for Agven secunty ina given quater, the end-ofsuarer price Was used, and vice versa " Because each fund's tok holdings af observed only quarterly in our data St iis tempting to think that LOM may spurious be nonzero because of fund performance It is ue that when the fund manager can achive superior ‘returns the actual poroio weighs are corelaed with ft {ure returns However, since LOM uses diferenced portfolio ‘weights a bis only aries when the portfolio revisions are ‘redominandy a the begining of the qarer (spurously Indicating momentum vesting) oa heen ofthe quarter (Gpuriouly indicating contrarian investment behavin. Since there is no a prion reason to believe that portfolio revisions tat occur for the purpose of achieving superior performance should our closer to the beginning of at ter than send, we donot Believe that bias exit GRINBLATT ET AL.: MOMENTUM INVESTMENT AND HERDING 1091 Here, we subtract means from returns in or- der to have measures that asymptotically ap- proach zero under the null hypothesis of no momentum investing. The monthly return from 12 months ahead for security j is used as a proxy for R..° We also use a similar decomposition of LIM into ““Buy LIM" and **Sell LIM.” While the LOM and LIM statistics are ap- propriate measures of the extent to which past returns affect the total holdings of a fund, we also use a turnover-adjusted LOM (TALOM) which measures the extent to Which past returns affect portfolio trades, in- dependent of the number of trades made by ‘a fund during a time period. This measure is sziven by (4) TaLow ‘The turnover-adjusted measure (TALOM) is the LOM measure, normalized so that the changes in weights of the stocks purchased (and the changes in weights of the stocks sold) sum to I during each quarter. The re- sults give a more accurate picture of the av- erage difference in past returns between stocks purchased and stocks sold across all quarters by a mutual fund, since a constant $1 is invested and shorted each quarter. A ‘mutual fund that trades very little, but buys past extreme winners and sells past extreme losers, will have a very high TALOM mea- sure, even though the unmodified LOM mea sure will be relatively small. Analogous to Buy LOM and Sell LOM, ““Buy TALOM"” and “Sell TALOM’ decompose TALOM into terms having Wy > W,sy-5 and Way < Way Fespectively. " Admitedly. this is 2 noisy proxy for the expected retum, but since thet are large numbers of stocks and time Periods average inthe measures we report the noisiness Of the proxy has negligible effect on oar rests,1092 THE AMERICAN ECONOMIC REVIEW DECEMBER 1995 “Tans I—Mowmavrusewvesting Susttany Srarisnics Fo SaMru oe 155 Murua. FoNos (QuaRTERLY FUND ‘Porrrouo Howes AR ror Tie Praioo Decramer 31,1974, TaRouc DeceMneR 31, 1984) ‘Venture capital Aggresive- Grown Special special ‘Tol sample ‘growth Balanced Growth income Income pirpose situations Suatisic (= 15) = 45) = 10) W= 4) W= 3) W= 1) WH) WH 3) LOM (pereendquaner) 074 135029089082 «O17 005095 attic 1os6r* 980% 383 HOTT 633 163 033317 Percentage postive 768 889 HDI G7H LS 657657 Wileoxon probably 0.0001 00001048 0.0001 _OOr_—030 O64 Oh statistic (LOM in every category = 0}: F = 51.57** FF2 sats (LOM is equal across categories: F = 18.24% Buy LOM (percenuquaner) «103,183 050,107 06t_ 0.6) 028170 tats pose 200% 176) 265H* 214 DORE 1386277 Percentage positive Sa 3832S SB S2 RS 430 STS Wileoxon probaly 010 om oa? 02306) 01s 003 1 stats (Bay LOM in every category = 0}: F = 5.85°* 2 statistic (Buy LOM isequlatoss categories) F = 008 Sell LOM (percenvquaner) -0.29-—-040- 012-013-031 051-006-060 aati “08 = 088-085 037. 1.181465" 060 1.16 Percentage postive 5308 500,500 «S08 SOK RS S7S 46.7 Wileoxon probaly ost 100100 Ost ost 06s oor 04 1 statistic (Sell LOM in every category = 0): F = 0.62 2 statistic (Sell LOM is equal across categories): F = 002 LIM (percenguarter) 030 083-002 043-001 003 ao 108 tats S46" 4st 033 G16" 101036218 aie Percentage postive 587689400780 35.1 4525.7 6H:7 Wilcoxon probaly 0.005 ouoi 0.44 0.001 002078 O_O FF statistic (LIM in every category = 0): F'= 12.16" 2 sats (LIM is equal across categories): F = B.41** Buy LIM (percentquarer) 0.85 134032 0830456038 stats 22" mee 11320150 2olF 080 Percentage positive 57538349258 SOR 500 ASB Wilcoxon probably 003 ouz ost 010 0st__t00 023 FF sats (Boy LIM in every category = OF = 3.36 2 sats (Buy LIM is equal across categories) F = 0.06 Sell LIM (percenvguaner) 044-072-023. 028-037, 041, 005-073 statis Hse 178) 127 “0.97 <177 037-160 Percentage positive 4542467500428 SDD. ATS Wilcoxon probaly 047 oo 09 os! 068 087 __ 087 FF stats (Sell LIM in every category = 0): F = 191" F2 ase (Sel LIM is equal across categories): F = 006 ‘TALOM (percenuquarer) 20733906298 os Ondo 250, "stats D5 97s Lis 927 ATI" 036 0.372424 Percent positive 73250073548 7691000 «66.7 Wilcoxon probability 2.0001 0001097 _o.o01 029 0010.05 O64 FF stats (TALOM in every category = 0): P = 30:39¢* F2 dase (TALOM is equal across categories): P= 930°VOL. 85 NOS GRINBLATT- ET Al.: MOMENTUM INVESTMENT AND HERDING 1093 ‘Tans 1—Contnue. ‘Venture capita” Aggressive- Growth Special special- ‘Total sample growth Balanced Growth income Tncome purpose situations Stats W=153)_ WH 45) = 10) W= 4) We 3 = TY C Buy TALOM (percenvquaner) 231 «364 ——«129~=«TR._SCOGKSCOSS.COMR aS P atatntic Tos 256132 2g 089096 om .95" Percentage positive SIT 567,922 525 M7 4D 58 Wileoxon probability 063006 oslo? 0d? 04400002 statistic (Buy TALOM in every category = 0): F = 2.48 -F2 Statistic (Buy TALOM is equal across categories): F = 006 Notes: The LOM statistic isthe measure of momentum investing based on stock turns inthe sme quarter a the portfolio revisions. The LIM statistic isthe measure of momentum investing based on stock retoms inthe quarter before the Pottfolio revisions. “TALOM' i the LOM statistic, sith portfolio revisions normiized so tht I of stocks are bought ach quarter and $1 are sold For each category above, weighted portfolio ofall funds in that eae the time-series mean and statistic are calulated for tha portfolio across probability thar the absolute valve ofthe Wileoxon-Mann-Whitny rank 2 statistic ls srater than the absolute vue of the observed: under the nll hypothesis Statistically significant at the [percent level AIL Results A. Summary Data on the Degree of ‘Momentum Investing Table 1 presents the average LOM measure for the entire sample, as well as for various investment-objective categories." According to this table, about 77 pereent of the mutual funds, 119 out of 155, buy “winners” and/or ° The mutual fonds in this study were subdivided in seven investmentobjctive categories, according 0 thet ‘ated objectives, Aggressive-growth abd growth funds in ‘vest inthe common stock of growth companies, with the ‘primary aim of achieving capital gains instead of dividend. Income. Growth-income fans seek o provide both capital ‘ain; and a steady stream of income by buying the shares of high-yielding conservative stocks. Balanced funds in ‘eatin both stocks and bonds, intending to provide capital fins and income while preserving principal, Income ands seek to povide high curent ncome by baying gv: ernment and corporate bonds as well as high-yielding ‘ommon and prefered stock. Finally special-purpose sind ‘venture-capitalspecia-situstions nds, 8 ther names Suggest, have very specilized strategies that vary from fund to Yond. These two categories representa vey sll portion of our sample sell “losers,”” as defined by the LOM measure. ‘The average LOM measure for all 155 funds over the 10-year period is 0.74 percent per quarter, indicating that, on average, the stocks held by a fund at the end of a given quarter had returns 0.74-percent higher, during that quarter, than the stocks held at the end of the previous quarter, which was highly statis- tically significant." F tests. strongly reject both that the average LOM is equal across investment-objective categories and that it is zero across categories. In unreported results, we also find that funds with the greatest ten- dency to buy winners in the frst five years of the sample period are more prone to buy win- ners in the second five years of the sample pe- riod, indicating that some managers follow consistent “styles.”” Table 1 also provides the average Buy LOM and Sell LOM measures for each cate- gory. Note that the results for the Buy LOM ‘measure are largely similar to the LOM " Nonparametric est results, designated by Wilcoxon probabilities, generally agree with the standard statistics.09 THE AMERICAN ECONOMIC REVIEW results. For example, the aggressive-growth, growth, and growthincome categories have the highest average LOM measures among the five investment-objective categories with nonnegligible numbers of funds, while the aggressive-growth, growth, and income cat- egories have the highest Buy LOM measures. Note, however, that the Sell LOM measures are insignificant (at the 5-percent level) for every category, and a joint test of signifi- cance cannot reject that the seven average Sell LOM measures are all equal to zero. Therefore, momentum investing appears to bbe almost entirely driven by funds buying winners, and not by selling losers. The results for LIM, Buy LIM, and Sell LIM are much the same. The LIM measure is significantly positive, on average, indicating that fund managers had a tendency to select stocks based on superior returns over the prior quarter. The average one-quarter-lagged mo- mentum measure is about 0.30 percent per ‘quarter, suggesting that, on average, the most recent quarter's returns were more important ‘determinants of portfolio choice (as shown by LOM) than the returns realized in the more dis- tant past (as shown by LIM). In unreported regressions, we found a strongly positive cross-sectional correlation between the LOM and LIM measures of the funds. The regres- sion of LIM on LOM gave a coefficient of 0.48, with a time-series ¢ statistic of 5.5; the reverse regression of LOM on LIM gave aco- efficient of 0.88, with a time-series statistic of 10.0)."" As with the LOM measure, the aggressive-growth and growth categories had, fon average, the highest levels of LIM: measured momentum investing, which is due, in par, to a larger percentage of these funds trading on momentum, relative to other funds (about 89 percent and 82. percent of aggressive-growth and growth funds, respec- tively, followed momentum strategies, accord- ing to their LOM measures). Despite its statistical significance, the 0.74- percent (0.30-percent) quarterly return for the LOM (LIM) measure seems economically in- These two regressions imply 4 conelation of 0.65, between LOM and LIM. DECEMBER 1995, significant, The results for the tumover- adjusted measures, TALOM and Buy TALOM (also shown in Table 1), provide a more dra- ‘matic confirmation of the momentum invest- ing behavior of the funds. For both measures, the average difference between buy and sell portfolios across all 155 funds was about 2 percent per quarter, confirming that buying ‘winners is the chief method of momentum in- vvestment. In unreported results, we found that the top 10 funds, ranked by TALOM, bought portfolios of stocks with returns that were ‘more than 8 percent (quarterly) greater than the portfolios of stocks they sold, on average; the top 25 had a difference of about 6 percent. ‘The TALOM results in Table 1 also confirm that the aggressive-growth and growth funds ‘were much more likely to have traded on mo- ‘mentum than funds in other large categories: their larger LOM measures were not primarily due to higher tumover than other categories (since TALOM is adjusted for tumover). Again, results from nonparametric tests gen- ‘erally agree with the standard f-statistic results. In results not reported in Table 1, we also ‘computed Buy LOM measures for partitions of stocks in the portfolio based on the market capitalization of the stocks, in order to mea- sure the relative contribution of buying win- ners in different size deciles to the overall LOM measure. For all objective categories, and for the total sample of funds, buying large capitalization past winners provided almost all of the contribution to the observed momentum- investing behavior. We also found no signifi- cant evidence of selling past losers in any size decile. B. The Relation between Momentum Investing and Superior Portfolio Performance In this subsection, we examine the extent t0 which a fund’s tendency to hold past winners relates to its performance. As mentioned ear- lier, past research CJegadeesh and Titman, 1993) suggests that stocks that perform rela- tively well over a 3~6-month period tend 10 realize relatively good performance during the next year. Hence, mutual funds that hold stocks that performed well in the recent pastVoL. 85 NO. 5 should realize better performance than those funds that hold stocks that did not perform well In order to measure mutual fund perfor- ‘mance, we employ the method developed by Grinblatt and Titman (1993), which does not require that we select a benchmark portfolio. The performance measure (a) developed in that paper uses a four-quarter change in (un- modified) portfolio weights and multiplies the differenced weights by a future return; that is, leas TL EE £ Ose Sor onces 6) « With this measure, the benchmark used to ‘adjust the return of a portfolio for its risk in a given month is the current month's. return eared by the portfolio holdings four quarters prior to the current quarter's holdings. There- fore, a represents the mean return of a zero- investment portfolio. If the systematic risks of the current and benchmark portfolios are the same from the point of view of an investor with no selectivity or timing abilities (as de- fined by Grinblatt and Titman [1989b]), the performance represented by a should be insig- nificant for that investor.* ‘We first split our sample of 155 funds into ‘momentum and contrarian investors, based on the sign of LOM and LIM, and examined the performance of these two subgroups. Table 2A compiles mean LOM measures for the total sample and for the five largest investment- “objective categories. For the sample of all 155 "The weights ofthis zro-investment portfaio repr sent the difference between the vector of fund petolio ‘weights the current period andthe vector of fond prt folio weights four quarters eae, We found that ana ternative performance measure which wses one-quarter Tag rather than a four quater lag revealed relatively ite performance, on average, which indicates thatthe stocks picked by these funds performed wellnthefllowing fout ‘quarter, and not simply inthe fist are the stocks were Tel. Tis finding rules ou the possiblity that Funds may be affecting thei measured performance by heavily buy ing (or selling) the same stock during consecutive qua * Grinblat and Titman (1998) provide evidence that ‘the two portfolios have the same market bets. GRINBLATT ET AL: MOMENTUM INVESTMENT AND HERDING 1095 funds, the 119 funds using momentum invest- ‘ment strategies clearly outperformed the 36 funds using contrarian strategies over the ten-year period. The performance of the mo- mentum investors averaged about 2.6 percent per year, while the contrarians had ‘an insignificant’ average performance of about 0.1 percent per year. Similar results held for most of the individual investment- objective categories Table 2B repeats this analysis with LIM as the momentum investing measure. The 91 LIM momentum investors outperformed the 64 LIM contrarians by about 1.8 percent per year, on average. The LIM momentum inves- ‘ors actually had slightly higher performance than the LOM momentum investors (although there is a large degree of overlap between the two). The LIM contrarians also achieved bet- ter performance than the LOM contrarians, and their performance was statistically significant (but relatively small in magnitude). Table 2 also shows that the investment- objective categories having the best perfor- mance are those that most strongly used a momentum strategy in selecting stocks (see the “total”? columns). Of the three categories with significant (at the 99-percent confi- dence level) performance (aggressive growth, growth, and income funds), two have signifi- cantly positive LOM and LIM measures. In fact, among the five major categories, the aggressive-growth category ranks first in performance, LOM, and LIM, while the growth category ranks second in each of these three measures." Interestingly, these ° At ist glance, it seems surprising that the LIM ‘wend-following and contrarian potolios both outperform {heir LOM counterparts, However, this follows from the fact that LIM classifies fewer funds as trend followers ‘We expect thatthe sample of LIM wend fllowers will contain stonger uend-followers than the LOM end: Tollowers. For this reason, the average returns ofthe LIM ttend-followers ar higher. In atin, since the LIM con- tearias inclade some ofthe fund that were classiid as ‘tend-followers by the LOM criteria, we also expect the average retums ofthe LIM contains to be higher than {he expected returns of the LOM contrarans "The special-purpose (SP) and the veature-capta” specal-stuations (VS) categories each had only thee finds.‘THE AMERICAN ‘Tanue 2—Mean ECONOMIC REVIEW DECEMBER 1995 Porrroue Sranstics ‘end “Toul Momentum Conarins 1 Based on LOM wats OW = 155) (N= 119) = 36) Lon (pecemiquater) 078 e030 1096) 078) S206) Peformance (peremheu) 208261 Gor) Gas) Prfrmanc of diferent ‘ania pens i 246 Gey 2B, Rayed om LM ine: (N= 155) WW = 91) w= 64) Lim gecemiguner) 0X08 033 46%) sles) C1936) Pvfrmance(peweminea) 208279097 Gite) Gin) am) Peformance of dienes ‘nfo pre a ut Mean pris Aagrenive arom = “Teal _Momentam Conrarians Teal Momentum Contarans W= 5) (=a) Was) = WHO) las as oat 09a ©) GODT!) C29) GB) GI) C2 340395, oo oat 635") GS) oy (005) a2 0s 2 (008) W245) Wea) w=) w= 1) Ww 0s) 098-038 -e2 035 Gein) Gear) C248) COS) a5) 34039236 ots G35) Gia) Gio (00 19) 146 o2s %0 02, ‘Notes: Far cach category, the an were parted ino wo ss those with postive momenta investing mess Mower) and ‘howe witha negative measure ("Contrarian") Equally weap polos woe ten farmed ad time series me ‘shown inthe tle The ifeenced potion te mamentum-avsing poo dst the coon pont. Number in arose ae sat. Sisal Spica a the 10 percent eve ately nian atthe Spree Sasa Sgiean a the 1 percent evel categories also tend to have the highest amount of portfolio turnover and tend to be the smallest in terms of the size of total as- sets managed. Only the income-fund cate- gory shows significant performance and an insignificant level of momentum investing Other categories of funds show some degree ‘of momentum investing, but their levels are relatively small. Regression results, all of which control for investment objective, are found in Table 3. ‘The first two regressions show a strong cor- relation between performance and momentum investing, whether measured with LOM or LIM, For the sample of all 155 funds, the es- timated regression coefficient of 1.27 for the first regression indicates that an increase of 1 percent in momentum investing, according to the LOM measure, increases performance by about 1.27 percent." Multiple regressions of performance on LOM and LIM and of performance on LOM, LIM, L2M, L3M, and L4M show that LOM provides the main explanatory power. This is not surprising, given the high correlation be- tween LOM and the momentum measures ‘based on longer lags. The next two regressions "Note tha the tumover adjusted LOM (TALIM) was not included as one ofthese momentum investing met sures because it is nota metic of the tendency ofa fond to choose a portfolio based on past returns of stocks, CTALOM was only used to compare the tendency of fonds to invest on momentum, without regard to differences in the intensity of trading across fundsVou. 85 No. s “Tanue 2—Estended GRINBLATT ET AL.: MOMENTUM INVESTMENT AND HERDING 1097 =) WA 36) WH 8) WH) WH 28) os 117-3902 st oI) UDKS) CI) aI) Gates, 2a) 2m ose O19 om 25) asi) (lad) (178) 999 ED 226 ai =) W=B) Wath Won w= 1H os 07) 045-008 039-028 (16%) BOB) 30) CLO) IH) CABG) 2a 2491S Oaks see) Casey (198) 36 Is on aay (2) show that LOM and Sell LOM do not explain performance, after controlling for Buy LOM (similar results are shown for Buy LIM and Sell LIM). This finding is consistent with our prior finding that momentum investing was concentrated in buying large-capitalization winners. The regression in the last row con- firms that Buy LOM explains performance bet- ter than Buy LIM. In unreported results, we compared the hypothetical gross (not risk-adjusted) port- folio returns of momentum investors and contrarians with a market benchmark, the value-weighted CRSP index (with daily div- idend reinvestment). In calculating gross returns, we assumed that the portfolio indi- cated by the beginning-of-quarter holdings ‘was held constant until the end of the quar- ter, when the weights were updated. We found that, in general, momentum investors realized higher gross returns than contrari- ans, and both realized higher returns than the CRSP index. For example, the mean gross return of the 119 LOM momentum investors Momeni Comin (a oo: 0a 05 03) Base) C2 ese) ai) to to was 17.9 percent per year, while that of the 36 contrarian investors was 17.2 percent per year.” The mean return of the CRSP index ‘uring this period was about 14.7 percent per year. From Table 2A, the average difference between the risk-adjusted performance of momentum and contrarian investors was 2.5 percent (per year) during this period, which was higher than the average difference be- tween their gross returns (0.7 percent per year). Contrarians held more priced risk in their portfolios by holding smaller stocks than momentum investors. IV. The Herding Behavior of the Mutual Funds ‘The preceding analysis indicates that mutual funds show a tendency to buy stocks based on "The top 20 percent of LOM momentum investors had an average hypothetical gross return of 19.0 percent per year, while the bottom 20 percent (the most contrarian investors) had an average return of 175 percent pt Yea.1098 THE AMERICAN ECONOMIC REVIEW DECEMBER 1995 “Tyo 3—Ceoss-SecrionAl: REGRESSIONS ACROSS ALL 155 PNDS, Deets VaRIABL? Independent Le vaable © Ww o ow uM lar Ls 1 066 Gorey 2) 233) (1.00) uM 120037 ox 202%) 03) 058) LM LM uM Buy LoM 94 Lo bas 165) 326) easy) sell LOM, 063 (os Buy LIM Lay oe ey on sell LIM ous 02 ®: 008 on 039 03904] rR 308 24603619 366 sas 00303) @003)__@o02)__08) oon, ‘Notes: In each repression, the time-series average fund performance (in percenea) is regressed, cross-sectional, on the time-series average momentum investing measures (tm perceaUquanet- For example, n repression (), the tme-seres ‘mean performance is regressed, across funds, onthe time-series mean LOM measute. The method of compatng and F Stites is based on a me-series procedure (se Grialat and Titman (1984] for details). Separate Gummy intercepts ‘were used for funds in different investnet objective caegores to contro for diferences in now momentun- investing. related performance across categories. Therefore, the common intercept was fixed at reo. The fsa ae piven ih Patenheses beneath coeficient estimates; mambers in parentheses Beneath F tatstics are p values "Statistically significant at the 10-percet level «Statistically si their past returns. This, by itself, suggests that mutual funds should show some (possibly weak) tendency to herd (ie., buy and sell the same stocks in the same quarter). For exam- ple, we would expect to observe more mutual funds buying than selling those stocks that have recently increased in price. In this see- tion, we examine this tendency to herd more generally ‘Asa starting point, we replicate the analysis of Lakonishok, Andrei Shleifer, and Robert W, Vishny (1992) (henceforth, LSV) on our sample of mutual funds. LSV calculated a sta- tistic, described by equation (6), that mea- sures the average tendency of pension funds either to buy or (o sell particular stocks at the same time (6) UHM, = |p. — Pil — Elpis ~ Pil ‘where p,, equals the proportion of funds, trad- ing in stock i during quarter ¢, that are buyers;VoL. 85 No. s Fr, which is the expected value of p,,, is cal culated as the mean of p,, overall stocks dur- ing quarter. Therefore, 7, i the proportion of fund trades in quarter # that are buys, for the average stock. We refer to the statistic given by equation (6) as the “unsigned herding measure” (UHM) to distinguish it from what we later describe as the ‘signed herding mea- sure'* (SHM), which separates buy and sell herding Table 4A presents the mean unsigned herd- ing Statistics, averaged over all stock-quarters (see the “total” column) and averaged over {wo subgroups of stock-quarters (see. the “buy” and “sell” columns). Membership of fa stock-quarter in one of the subgroups was determined by whether the set of all funds was buying or selling the stock during the quarter to a greater degree than would be expected with random buying and selling (ie. stock i during quarter 1 was considered to be a “buy herding” stock-quarter if p,, > 7: similarly, stock “sell herding” categorization occurred when p,, < ;). This partition allows us to determine whether herding was stronger on the buy side than on the sel side of institu- tional trades. Analogous to LSV, the stati tics given in Table 4 are from the perspective of individual stocks (instead of from a fund perspective and are based on the entire sam- ple of 274 mutual funds (including nonsur- Vivors) that existed on December 31, 1974 Inaddition, we segregated the stock-quarters by whether they had a retumn among the top 50 percent of NYSE and AMEX returns dur- ing the quarter, or among the bottom 50 percent “The herding statistic of 2.5 percent (in Table 4A under the “total” column for all 274 funds) is the unsigned herding measure, averaged over all NYSE and AMEX stock” ‘quarters (where trades by at least one fund oc- ‘curred in that stock) during the period from December 31, 1974 to December 31, 1984 ‘This overall herding measure can be thought of as meaning that, for the average stock- ‘quarter, if 100 funds traded in that stock- quarter, 2.5 more funds traded on the same side of the market than would be expected under the null hypothesis that the stocks were picked independently. This overall GRINBLATT ET AL; MOMENTUM INVESTMENT AND HERDING 10% level of herding does not seem economically significant, and itis similar to the mean level that LSV found for pension funds, 2.7 percent. Not surprisingly, Table 4A. shows that the set of all funds exhibits more herding in buy- ing past winners than in buying past losers. However, herding that occurs on the sell side, although positive, appears to be less related to past returns. These findings are consistent with the average fund being a momentum investor that buys past winners but does not systemat- ically sell past losers, which results in several funds herding into (but not out of) the same ‘groups of stocks based on their past-quarter returns. ‘The average herding measure for the set of all funds appears to be small. Two explana- tions for this are examined in Table 4. The first has to do with the possibility that we are mea- suring herding over a sample of investors that is too broad. For example, by definition, all investors cannot be buying and selling as a herd, since, in the aggregate, the buys must ‘equal the sells. As a result, if our sample of mutual funds is representative of a large frac tion of trading, then we would not expect to find much evidence of herding. However, herding may exist among various subsets of the mutual funds. For this reason, we also ap- ply the LSV herding measure [equation (6)] {o measure imbalances between buys and sells of the smaller subgroups represented by the investment-objective categories. The results in Table 4A indicate that we find even less evi- dence of herding within investment-objective category subgroups. ‘A second reason why we may not have found strong evidence of herding is that the herding measure was aggregated across all stock-quarters, including those with very little trading by the mutual funds. Intuitively, it makes sense to condition the herding measure fon the number of funds trading in the stock during the particular quarter. It is certainly much more meaningful to analyze the ten- dency of funds 10 be either simultaneously buying or selling a particular stock that several funds are trading in a particular quarter than a stock which only a few funds are trading. Be- cause of this, we present the average herding110 ‘THE AMERICAN ECONOMIC REVIEW DECEMBER 1995 “TABLE 4 Mean Henoine Srarisics, Seontcatin ty PASt-QUARIER-RETURN DECILES AND By "Buy" o "Sei" HERDING “Tota sap of i ‘Aggesive gh w=) Bane Sande N= 1) Stic Bey Set Tey Se Talay Selo 1X Mean Hering Stic (Percentage) for Voune of Trade = I Pa ers Pe er er er er ‘Nonberoftock-qunies) 641) OSE) CRISS) (SMD) HIRT) 12ET) «1.308) m0) G78) Pa winners Pe ey er en ee) Tamer stock-genes) (LAK) (1288) 277) 4) (698) UASH)_LD)_ IBN) ITO) Ait soxk-quatr Ta 325005) 32a) ae 207-04 (Suniefsock-qures) @LOR) GOST) INS) CARI2 3817) 76K) BHM ABN) GTS) 1B. Moun Mending Stic (Percentage) for Volume of Trade = 5 Pa oer B49 kok sk 783 gas ata ‘amber ofsockguites) Gt) GORD GM (SB) HD) 285) SH) HIS) Past ier ash 447453650 S08 550-126 ‘Mimberofsockquines (407!) (4367) BH) 2H) THD) ST) All sock-quarers aa 46h AMCs BIDAR (umber of sickquanes) (740) (8284) IS) 40D) (460) ae) TBO) GR) Mean Herding Sait Percete for Volume of Tre 10 Pt ners sas 525381 MSS wo “Number ofstockquaner) (180) (1853) 008) (106) 8) Pa winners Som 538439} 67 TH ss ota (Nanberofsock-qerem) (488) cast) G90) 2a wT) Alt soxk-quters Sa sak ss) ASD 7M (Gambro sockqaren) —GOF) G88) 6s) G3) tex: ast quarried a those ous ring he ume grr the por revisions, vial ering attics re ‘Sued a p= ipl! Elp lp), whee p= the poperton of fans bin he sn tock caring the piven ater among al ‘nde ht wad hat stock rng ht quar Fp and Ep Ip arc unit he nl hypothesis of mo ito ering. The “mean hing sate” the nvrge of he nia herding sacs Ss ne ah ns sch om gren easy, THe laid "ais he mean heding stati acted ove il tock sre ving a est he vue of rae by the fds ‘nica nthe part Buy cele the average ve nly Use tock quarters whee p> Ep) hat the proper of bays ‘spent ha th expected poparton of bye "Sell ick othe avrg ely aver hose ck are fr which p= Bip) Pu esate th seks tang ps furs the lower $b percent smng al NYSE an AMEX tock ring he ten quart wpa wire ae hoe having ps returns nthe per 0 perce ‘measures over all stock-quarters with at least five active funds in Table 4B, and over all stock-quarters with at least ten active funds in Table 4C, Panels B and C of Table 4 show that, when ‘we limit our analysis to stock-quarters with at least five or ten trades, respectively, evi- dence of herding increases significantly. For example, for the entire sample of funds, the average herding measure is about 5.5 percent when we include only stock-quarters with at Teast ten funds active. Note that, when at least ten funds were active, funds in the ob- jective categories with the highest average performance (aggressive growth, growth, and income [see Table 4A]) showed the greatest tendency to herd in the average stock-quarter. ‘The next step in our analysis is to characterize individual funds by the extent to which they “go with the crowd."" In order to measure a particular fund's tendency to herd, we first develop what we call a “signed” stock herding measure (SHM), defined below, which provides an indication of whether a fund is “following the crowd’” or “goingVOL. 85 NO.$ ——GRINBLATT ET AL “Tantn 4 Extended, ‘Growth—loome funds Growth fads Y= 81) wes) 06) 1h seo sk Ls (IS) st) 338) (SI) 5525) Ms 728%) G17) 84s) 8857) SMA, Cr rT) tat (33) 83TH NHI) 40985) 23st) aan e328 Bi) «1.5, (6) 0) 3% 450 3453.98 420s 613030) __ 1) __85)_500, 417 4714393 ith a3) (442) A391) RI) sa 4x9 a0 am sae Ga) Gi) G3) 7H Gao) mu 4g $98 ayaa G3) 28) Ge) is) Gr esta) _—S8 3a 473 (Sih) rs) aw sn against the crowd” in a particular stock during a particular quarter: (7) SHM, 1, * UHM, ~ El, x UHM.,] where SHM,, = 0 if fewer than 10 funds traded stock during quarter #. Otherwise, (0 wine < elm. — => Ele pi and eal od yer fe ting quae 18 8) > lps Bi and te ad i eee he fond M1 tn 81> ln, i de tdi eo ck ing ron — 3 > ln, ~ and fond eye (th nd MOMENTUM INVESTMENT AND HERDING not Income ands W = 31) 270) 75) (5459) Gk) Got) 6198) “i 293 Oe G81) (6H) 11855) a6) ila) no 13607 a) (5) 400 6a) Sk 8) (182) G3) v4 os 0 “3 S38 3 ub 25) on 3% us) 38) Note that SHM,, = 0 if a stock-quarter shows negative herding or if only a small number of funds have traded it, since there is no mean- ingful way in which the fund ean herd (or in- vest against the herd) in these cases. Also, 1, = Lif the fund trades “with the herd” in stock i during quarter t, and J,, = -1 if the fund trades “against the herd” in that stock- quarter. The second term in SHM,, is calcu- lated under the null hypothesis of no herding by the funds in the stock-quarter (above that due to chance)” ® Under the null hypothesis of independent trading de ‘sions among funds, the numberof ading funds that are buyers is binomially distibuted. We can caleulte the Value of E(7 x UHM) for stock Fin quarter starting with {he following known binomial parameters:102 THE AMERICAN ECONOMIC REVIEW DECEMBER 1995 ‘Taste 5—Meas Portrouio Sraristics ‘Venture capital Toul Agaresive Growth Special: special iow Balanced Growth income Income situations Measure W245) WW) Wes) W=3) Wald, w= FM 10506 089 OO O12 083) percent) GGT) GAH) 9H) HIS) ERIN 4G") (192) TB) Lom O74 125029, owe oa2 700s os (percentiquarer) (10.96%) (@80%*) GAB) (LOTIE*) 33%) (163) (033) GIT) Performance 208 340 Okt 888 021 266 (percentyeas) — GAG) GSS) OB) HH) LIS) — oH) — OI) (14) ‘Note: For each category above an equally weighted portfolio of al funds in that category is formed. Then forthe fund (EHIM), we calculate the portfolio. weighted “signed herding measure” (SHB) of the stocks held by 3d potolio athe end ofa quater, less the portfolio weighed SHM ofthe stocks held at the beginning ‘ofthat quarter, based onthe herding measute of he stocks during that quar. Finally, theme series mean and 3 far calcbltedacros sil 40 quarters. For the LOM measure, the same procedure is followed, but the potolio-weighed ‘tock retums ae used instead ofthe portolioweighed herding meas (giving atime series of 120 months of at) For the performance measure, we calculate foreach quater the portfolio-weighted stock retums (ofthe next quartet) ‘base othe end-of quarter poli, less the portoio-weighted returns (the next quater) based onthe portfolio held four quarters previously. Ths procedure gives a time series of 111 months of data Time-sries# statistics are given in parentheses "Statistical Sais significant atthe 10-percent level lly significant atthe L-percent level ‘The fund herding measure for an individual fund (FHM) is then calculated by substituting the signed herding measure in place ofthe stock return in equation (2) (for k = 1); that i AAs the above equation illustrates, a positive (negative) portfolio revision is multiplied by a positive (negative) SHM if the set of all funds bought (sold) heavily in a given stock during a given quarter, giving a positive con- tribution to that fund's FHM. Conversely, a positive (negative) portfolio revision is mul- Liplied by a negative (positive) SHM if the set of all funds sold (bought) heavily in a given hos) SHM he numberof fonds wading stock 1 quarter the proportion of trading fonds i the population that te buyers, estimated as deseibed for equation (6), Note that inthe above expectation UHM = UHM(,), Where p =the proportion of funds trading in stock quater (Cp that are buyers. Then, for stock Fin quarter f Bu UHM 2p — UHM (PPP) ep UHM Pe) where forthe discrete values that p can assume, oio=(Q)eru- ar stock during a given quarter, giving a negative contribution to that fund's FHM. Hence, funds that tend to buy (sell) when other funds are also buying (selling) will be characterized as herders by this measure. Table 5 presents the fund herding results. All categories of funds showed highly signif- icant levels of FHM, and unreported F tests strongly rejected that the average FHM fund herding measure is equal across categories, or that it is zero for all categories. We can inter- pret the reported 0.84 value for FHM as mean- ing that, ifthe average fund traded 10 percent Of its portfolio each quarter, it bought stocks that, on a portfolio-weighted average, hadVOL 85NO.S ——GRINBLATT ET AL.: MOMENTUM INVESTMENT AND HERDING 103 “Taste 6—Cross-SecrionAt REGRESSIONS OF FUND PERFORMANCE Growth Special Independent Growth “income Income variable wei) W=3)_ W=13) [A. Cross Sectional Regressions of Fund Performance on Fund Herding Measure (FHA) Constant 299 oat 025-1728 2.79 Gor) 033 045) E216) HH) HD) 126) FAM Lot 040 0302972781609 67998 ex) Os) 0) GT) 28) CI 8) 3B) ‘Adjuned R= 025 0.0001, 00202 ~~02S ~~ B. Cross Sectional Regressions of Fund Performance on LOM and FHM: Constant, 265 020 049035239 ts Ga) 030) 67) 053) 230 Lom Las os 002, 108122286 ez) aay 002) ST) OT rum 012 “032 028 0610-02 (020) 03) 02 oy 17) F 439 110 008 600A? 266 oon 3 096) 003) 1) OT) Adjusted R= 039 on om 00s 09 ‘Notes: In panel A. for each category the benchmark-fre fund performance measure (percet/Yea) is regressed, across funds, on the fund herding measure (FHM, as percentage). In panel B, foreach category the benchmark-free fund performance measure (percent eat i regressed, across funds, onthe fund omentum investing measure (LOM, percent! ‘qarer) and on FHM. The method of computing # and P sass is given in Grinlat and Taman (1994). For the Fegressions across all 15S Tunds, separate dummy intercepts were used for funds in different investment objective cate ores to contol for diferences in non-reressor-elated performance across categories. Therefore, the common intercept tras fixed a 200 for those regressions. Stent a ses ae in parentheses below the coefficient estimates the numbers in parentheses beneath the Fsaistics are the snscinted p vals. “Insufcint data * Statistically significant atthe 10-percen level Significant tthe S-prcent evel Significant atthe I-percent level about 8.4-percent excessive buying by all funds, or sold stocks that, on a portfolio- weighted average, had 8.4-percent excessive selling by all funds (or some combination of these two extreme outcomes); while an aver- age aggressive-growth fund trading 10 percent each quarter bought (sold) stocks with about 10.5-percent excessive buying (selling) by the set of all 155 funds. Note that aggressive-growth funds had the highest average levels of FHM, LOM, and per- formance, while the growth funds were second in all three categories (among the five major fund categories). Funds that invest on mo- ‘mentum are more likely to invest in herds and are more likely to perform. ‘Table 6A presents results for ross-sectional regressions of performance on FHM. The results show that fund performance is signif cantly correlated with the tendency of a fund to herd (FHM). This finding by itself would support the idea in some theoretical herding ‘papers that informed investors have a tendency to herd (Brennan, 1990; Froot et al., 1993; Hirshleifer etal, 1994). However, this is due to the high correlation between the tendencyios THE AMERICAN ECONOMIC REVIEW to herd and the tendency to buy past winners, which was confirmed in an unreported regres sion of FHM on LOM. Table 6B shows that, at the margin, FHM does not significantly ex- plain fund performance, given the explanatory power already provided by LOM. Therefore, fon average, performing funds tend to buy past winners, with herding in past winners appar- ently occurring as a result V. Conetusion ‘This paper characterizes some of the investment strategies of mutual funds and an- alyzes how these strategies relate to realized performance. The evidence indicates that mu- tual funds have a tendency to buy stocks based ‘on their past returns, and that they tend to buy and sell the same stocks at the same time (i.e herd) in excess of what one would expect from pure chance. The average level of herding and ‘momentum investing was statistically signifi- cant, but not particularly large. However, there ‘was a significant degree of cross-sectional dis- persion across funds in their tendency to buy past winners and to trade with the herd. ‘The tendency of individual funds to buy past winners as well as to herd was shown to be highly correlated with fund performance over ‘our period of study. The relation between the tendency to buy past winners and performance ‘was especially strong. On average, those funds following momentum strategies realized sig- nificant excess performance, while contrarian funds realized virtually no performance. The relation between a fund’s tendency to go with the herd and its performance was less con- vincing, and it largely disappeared after con- trolling for the fund’s tendency to buy past winners. This research provides some insights about the extent to which mutual funds are able to profit from their security-analysis ef- forts. The positive relation between momen- tum trading and performance suggests that the positive performance of mutual funds observed in Grinblatt and Titman (1989a, 1993) may have been at least partially gen- ‘erated by a simple trading rule rather than by superior information. This suggests that if the momentum profits observed in Jegadeesh DECEMBER 1995 and Titman (1993) disappear in the future, then the performance of these funds is likely to diminish, REFERENCES Banerjee, Abhijit. ““A Simple Model of Herd Behavior."” Quarterly Journal of Eco- nomics, August 1992, 107(3), pp. 797 817. Bikhchandani, Sushil; Hirshleier, David and Welch, Ivo. “A Theory of Fads, Fashion, Custom, and Cultural Change as Informa: tional Cascades.” Jounal of Political Economy, October 1992, 100(5), pp. 992— 1026. Black, Fischer. “Presidential Address: Noise."* Journal of Finance, Suly 1986, 41(3). pp. 529-44, Brennan, Michael J. “Presidential Address: Latent Assets.” Journal of Finance, July 1990, 45(3), pp. 709-30. Brown, Stephen J. and Goetzmann, William N. “Performance Persistence."” Journal of Fi- nance, June 1995, 50(2), pp. 679-98, De Long, J. Bradford; Shleifer, Andrei; Summers, Lawrence H. and Waldmann, Robert J. “Pos- itive Feedback Investment Strategies and Destabilizing Rational Speculation.”* Jour- nnal of Finance, Sune 1990, 45(2), pp. 379— 96. Friend, Irwin; Blume, Marshall and Crockett, Jean, Mutual funds and other institu: tional investors. New York: McGraw- Hill, 1970, Froot, Kenneth Aw; Scharfstein, David 5. and Stein, Jeremy C. “Herd on the Street: In~ formational Inefficiencies in a Market with Short-Term Speculation."” Journal of Fi- nance, September 1992, 47(4), pp. 1461— 84. Grinblatt, Mark and Titman, Sheridan D. “Mutual Fund Performance: An Analysis of Quarterly Portfolio Holdings."* Journal of Business, July 1989a, 62(3), pp. 393~ 416. . “Portfolio Performance Evaluation: Old Issues and New Insights."” Review of Financial Studies, 1989b, 2(3), pp. 393— 421 ——.. “Performance Measurement withoutVoL. 85 NO. s Benchmarks: An Examination of Mutual Fund Returns.” Joumal of Business, Sanu- ary 1993, 66(1), pp. 47-68. “A Study of Monthly Mutual Fund Returns and Performance Evaluation Tech- niques.’* Journal of Financial and Quanti- tarive Analysis, September 1994, 29(3), pp. 419-44 irshleifer, David; Subrahmanyam, Avanidhar and Titman, Sheridan, D. "Security Analysis and Trading Patterns When Some Investors Receive Information Before Others." Jour- nal of Finance, December 1994, 49(5), PP. 1665-98. Jegadeesh, Narasimhan and Titman, Sheridan D. “Returns to Buying Winners and Selling Losers: Implications for Stock Market EI- ficiency.”” Journal of Finance, March 1993, 48(1), pp. 65-91 ‘Kraus, Alan and Stol, Hans R, “Parallel Trading by Institutional Investors.” Journal of Fi- nancial and Quantitative Analysis, Decem- ber 1972, 7(5), pp. 2107-38. Lakonishok, Jose; Shleier, Andreis Thaler, Ri- chard H. and Vishny, Robert W. “Window GRINBLATT ET Al: MOMENTUM INVESTMENT AND HERDING 110s Dressing by Pension Fund Managers.”” ‘American Economic Review, May 1991 (Papers and Proceedings), 81(2), pp. 27-31 Lakonishok, Josef; Shleifer, Andrei and Vishny, Robert W. “The Impact of Institutional Trading on Stock Prices.” Journal of Fi- nancial Economics, August 1992, 32(1), pp. 23-43. ‘Scharfstein, David S. and Stein, Jeremy C. “*Herd Behavior and Investment.”” American Eco- nomic Review, Sune 1990, 80(3), pp. 465— 79. Schwartz, Robert and Shapiro, James. “The Challenge of Insttutionalization forthe Eq- uty Market” Working paper, New York University, 1992. Shier, Robert J. and Pound, John. “Survey Ev- ‘dence on Diffusion of Interest and Infor- mation Among Investors."” Journal of Economic Behavior and Organization, Au- ust 1989, 12(1), pp. 47-66. ‘Trueman, Brett. “A Theory of Noise Trading in Securities Markets.” Journal of Finance, March 1988, 43(1), pp. 83-95.