Mandatory Portfolio Disclosure, Stock Liquidity, and Mutual Fund Performance *

Transcription

1 Madatory Portfolio Disclosure, Stock Liquidity, ad Mutual Fud Performace * Vikas Agarwal Kevi Mullally Yuehua Tag Baozhog Yag ** August 2013 ABSTRACT This paper studies the impact of madatory portfolio disclosure of mutual fuds o the liquidity of disclosed stocks ad o fud performace. We cosider a theoretical model of iformed tradig with differet madatory disclosure frequecies. Usig a regulatio chage i May 2004 that icreased the frequecy of madatory disclosure, we fid evidece cosistet with the model s predictios. First, stocks with higher fud owership experiece a larger icrease i liquidity as compared to other stocks subsequet to the madatory icrease i disclosure frequecy, especially for stocks disclosed by more iformed fuds or subject to greater iformatio asymmetry. Secod, better performig fuds experiece a greater drop i their abormal performace followig the regulatio chage, particularly whe they hold stocks with greater iformatio asymmetry or whe they take loger to complete their trades. Take together, our evidece suggests that madatory portfolio disclosure improves market quality by icreasig stock liquidity but imposes costs o iformed ivestors. JEL Classificatio: G14, G23, G28 Keywords: Portfolio disclosure; Stock liquidity; Mutual fuds; Fud performace * This paper has beefited from commets ad suggestios from Corad Ciccotello, Sevic Cukurova, Gerry Gay, Edith Gigliger, Christia Gourieroux, Carole Gresse, Ala Huag, Jayat Kale, Madhu Kalimpalli, Matti Keloharju, Blake Phillips, Sugata Ray, Adam Reed, Christopher Schwarz, Laura Starks, Avaidhar Subrahmayam, Ji Wag, ad semiar participats at Aalto Uiversity, Georgia State Uiversity, Louisiaa State Uiversity, Uiversity of Paris-Dauphie, Uiversity of Waterloo, ad Wilfred Laurier Uiversity. We are especially grateful to ad thak Christopher Schwarz for geerously providig us with data o the mutual fud holdigs. J. Mack Robiso College of Busiess, Georgia State Uiversity, 35 Broad Street, Suite 1207, Atlata, GA Research Fellow at the Cetre for Fiacial Research (CFR), Uiversity of Cologe. Tel: vagarwal@gsu.edu. J. Mack Robiso College of Busiess, Georgia State Uiversity, 35 Broad Street, Suite 1216, Atlata, GA Tel: kmullally1@gsu.edu. J. Mack Robiso College of Busiess, Georgia State Uiversity, 35 Broad Street, Suite 1242, Atlata, GA Tel: ytag9@gsu.edu. ** J. Mack Robiso College of Busiess, Georgia State Uiversity, 35 Broad Street, Suite 1243, Atlata, GA Tel: bzyag@gsu.edu.

2 Madatory Portfolio Disclosure, Stock Liquidity, ad Mutual Fud Performace Madatory disclosure of portfolio holdigs by istitutioal moey maagers is a vital compoet of securities market regulatio. Madated by the Securities Exchage Act of 1934 ad the Ivestmet Compay Act of 1940, portfolio disclosure provides the public with iformatio about the holdigs ad ivestmet activities of istitutioal ivestors. Academic research has utilized the disclosed holdigs to study may related topics. These topics iclude volutary portfolio disclosure (Ge ad Zheg (2006)), frot ruig ad copycat tradig activities (e.g., Frak et al. (2004), Coval ad Stafford (2007), Verbeek ad Wag (2010), Brow ad Schwarz (2012)), widow dressig behavior of disclosig istitutios (e.g., Lakoishok et al. (1991), Musto (1997, 1999), Agarwal, Gay, ad Lig (2012)), the hidig of certai positios (Agarwal et al. (2013) ad Arago, Hertzel, ad Shi (2012)), the costs of disclosure to hedge fuds (Shi (2012)), ad itra-quarter tradig (Wag (2010) ad Puckett ad Ya (2011)). Amog the madatory disclosure requiremets o istitutioal ivestors, those o mutual fuds provide perhaps the most detailed iformatio about their portfolios. 1 However, the impact of mutual fuds portfolio disclosure o their disclosed stocks ad the fuds themselves has ot yet bee examied i the extat empirical literature. To fill this gap, this study examies how madatory portfolio disclosure affects the (i) liquidity of the stocks disclosed by mutual fuds, ad (ii) mutual fud performace. Oe of the challeges i coductig such a study is that it is difficult to idetify the causal effects of portfolio 1 See Sectio I for more detailed discussio. 1

3 disclosure o stock liquidity ad fud performace. We overcome this challege by usig a Securities Exchage Commissio (SEC)-madated regulatio chage i May 2004 regardig the disclosure requiremets for mutual fuds. This chage forced mutual fuds to icrease their portfolio disclosure from a semiaual to a quarterly frequecy. We use this regulatio chage for a quasi-atural experimet to idetify the effects of mutual fuds portfolio disclosure o stock liquidity ad fud performace. We motivate our empirical aalyses usig the theoretical literature o madatory disclosure ad iformed tradig. Huddart, Hughes, ad Levie (2001) (heceforth HHL) build o the Kyle (1985) model ad study madatory disclosure of trades by iformed traders. We exted the HHL model by cosiderig differet madatory disclosure frequecies. We aalyze the impact of disclosure frequecy o stock liquidity ad iformed trader s profits ad produce several testable predictios. First, our model predicts that more frequet madatory disclosure by iformed traders improves market liquidity as measured by market depth, amely the iverse of the Kyle (1985) lambda. The ituitio is that, with madatory disclosure, the market maker ca ifer iformatio from the disclosed positios of iformed traders as well as from the aggregate order flows, which reduces the impact of iformed trades o prices. Secod, the liquidity improvemet is greater for stocks subject to higher iformatio asymmetry. Third, our model predicts that the iformed trader s profits decrease i the frequecy of madatory disclosure because the market s learig of disclosed trades limits the trader s ability to reap the full beefits of his iformatio. Fially, the magitude of the iformed trader s profit drop is positively related to both stocks iformatio asymmetry ad the umber of periods the trader 2

4 takes to complete his trades. To test these predictios of our model, we start by examiig the impact of portfolio disclosure o the liquidity of the stocks disclosed by mutual fuds subsequet to a icrease i disclosure frequecy. A large body of literature has show that mutual fuds disclosed portfolios cotai valuable iformatio. 2 Give this evidece, we expect that stocks with higher fud owership should experiece greater icreases i liquidity with more disclosure. To test this hypothesis, we employ a differece-i-differeces approach to examie the chage i stock liquidity durig the two-year period aroud May The idetificatio of our aalyses relies o a cross-sectioal compariso of liquidity chages i stocks with high mutual fud owership (the treatmet group) ad those i stocks with low fud owership (the cotrol group). Accordig to the SEC s Electroic Data Gatherig, Aalysis, ad Retrieval (EDGAR) database, a vast majority of actively maaged U.S. equity fuds (over 97%) had to switch from reportig two times to four times each year due to the regulatio chage. Our empirical aalyses focus o this sample of affected fuds. We fid that stocks with higher mutual fud owership experiece sigificatly larger icreases i their liquidity subsequet to the madatory icrease i disclosure frequecy. Moreover, the improvemet i stock liquidity is ecoomically large. For istace, a oe stadard deviatio icrease i mutual fud owership is associated with a 0.19 ad 0.08 stadard deviatio decrease i the Amihud (2002) illiquidity measure ad relative bid-ask spread, respectively. This evidece supports our model s predictio that more frequet madatory portfolio disclosure of iformed traders improves stock liquidity. 2 See Sectio II for discussio of this literature. 3

5 We corroborate this fidig by coductig several sets of placebo tests. First, we carry out cross-sectioal placebo tests by icludig other istitutioal ivestors (o-mutual fuds or hedge fuds) as cotrol groups. The uderlyig argumet is that the regulatio chage i 2004 oly applies to mutual fuds, but ot to other istitutioal ivestors. Specifically, we coduct differece-i-differece-i-differeces tests ad fid that mutual fud owership has a larger impact o stock liquidity tha that of o-mutual-fud owership or hedge fud owership after the regulatio chage. Secod, we coduct a time-series placebo test usig a alterative sample period. We choose November 2006 as our placebo evet date to avoid ay overlap with other market evets affectig stock liquidity (e.g., the Log Term Capital Maagemet debacle i 1998, the burst of the dotcom bubble i 2000, the decimalizatio of stock prices i 2001, ad the fiacial crisis from 2007 to 2009). We do ot fid similar effects of mutual fud owership o stock liquidity durig this alterative period. The results from both the cross-sectioal ad time-series placebo tests help mitigate cocers that our results are drive by a time tred i stock liquidity. Next, we test whether the improvemet i stock liquidity is larger for the stocks held by more iformed fuds ad stocks associated with greater iformatio asymmetry. To test this hypothesis, we coduct differece-i-differece-i-differeces aalyses o the (i) subsamples of fuds classified usig proxies of the likelihood of fuds beig iformed ad (ii) subsamples of stocks categorized usig proxies of the extet of iformatio asymmetry. I the fud subsample aalyses, we use two proxies for the likelihood of a fud beig iformed: (i) fud s risk-adjusted performace (Carhart (1997) four-factor alpha), ad (ii) the Daiel, Griblatt, Titma, ad Wermers (1997) (DGTW) bechmark-adjusted fud returs. 4

6 Usig both these proxies, we fid the stocks held by more iformed fuds (i.e., higher past abormal performace) experiece greater icreases i liquidity after the icrease i the disclosure frequecy. Next, i the stock subsample aalyses, we cosider three measures to proxy for the iformatio asymmetry of a stock: liquidity, aalyst coverage, ad firm size. Cosistet with our model, we fid that less liquid stocks, stocks with lower aalyst coverage, ad stocks with smaller market capitalizatio experiece a larger icrease i liquidity tha do other stocks subsequet to the icrease i disclosure frequecy. I additio, these results further help i cotrollig for time tred i stock liquidity sice the liquidity tred should ot affect the subsamples of stocks differetly. While i the aforemetioed aalyses we focus o the impact of portfolio disclosure o the disclosed stocks, we ext examie the impact of a icrease i madatory portfolio disclosure o mutual fud performace. Cosistet with our model s predictio, we fid that iformed fuds bear costs from the icrease i madatory portfolio disclosure. Specifically, better performig fuds, i.e., those i the top quartile based o past four-factor alphas or DGTW-adjusted returs, experiece sigificat declies i their abormal performace followig the 2004 regulatio chage. After cotrollig for potetial mea reversio i fud performace, the drop i abormal performace of top quartile fuds rages from 1.9% to 4.5% o a aualized basis. This deterioratio i the fud performace ca be related to our earlier fidig of a icrease i stock liquidity. Followig the regulatio chage, more iformed fuds hold more liquid stocks ad may therefore ear lower four-factor alphas. To ivestigate this issue, we calculate chages i five-factor alphas by adjustig for the Pástor ad Stambaugh (2003) 5

7 liquidity factor. We fid that about oe-fourth of the abormal performace declie ca be attributed to the chages i liquidity of the disclosed stocks. However, three-fourth of the performace declie of the top performig fuds remais. Lastly, we examie how iformed fuds portfolio characteristics ad tradig behavior affect the extet to which more frequet disclosure hurts their performace. Specifically, we study the relatio betwee the performace drop of iformed fuds ad i) the iformatio asymmetry i the stocks that they hold ad ii) the time (i.e., the umber of quarters) it takes the fuds to complete their tradig strategies. Cosistet with our model s predictios, we fid that the performace decrease is greater whe top performig fuds hold stocks that are subject to greater iformatio asymmetry or whe they take loger to fiish their trades. Our paper cotributes to the large literature that studies issues related to portfolio disclosure. To the best of our kowledge, our study is the first to examie the implicatios of portfolio disclosure o both the quality of capital markets ad idividual fud performace. For this purpose, we first provide a theoretical model allowig for madatory disclosure with differet frequecies ad geerate several testable predictios. The, we use the regulatio chage i 2004 to test these predictios ad establish causal relatios (i) betwee portfolio disclosure ad the liquidity of disclosed stocks, ad (ii) betwee disclosure ad fud performace. Our evidece suggests that a icrease i portfolio disclosure of iformed istitutioal ivestors ca improve market quality by icreasig stock liquidity, which should help reduce the cost of capital of issuig compaies as well as the trasactio costs of ivestors. This effect is similar to that of a icrease i issuer or corporate disclosure, which has bee show 6

8 to lead to more liquid capital markets (Diamod ad Verrecchia (1991), Fishma ad Hagerty (1998, 2003), ad Admati ad Pfleiderer (2000)). However, we fid that iformed fuds experiece a drop i their abormal performace ad bear substatial costs from more frequet portfolio disclosure. To the extet that madatory portfolio disclosure reveals iformatio about proprietary ivestmet strategies of moey maagers, it ca affect their icetives to collect ad process iformatio ad, i tur, affect the iformatioal efficiecy of fiacial markets (Grossma ad Stiglitz (1980)). Therefore, for policy decisios related to portfolio disclosure, regulators should weigh the beefits of a more liquid capital market agaist the costs bore by istitutioal moey maagers. The remaider of the paper is orgaized as follows. Sectio I provides the istitutioal backgroud. Sectio II discusses the related literature ad the testable predictios of our model. Sectio III describes the data ad explais the costructio of variables. Sectio IV presets the empirical aalyses of the impact of the chage i madatory disclosure o the liquidity of the disclosed stocks. Sectio V examies the effect of the regulatio chage o mutual fud performace. Sectio VI offers cocludig remarks. I. Istitutioal Backgroud Madatory disclosure of istitutioal ivestors portfolio holdigs is a key part of securities market regulatio. The SEC requires mutual fuds to disclose their portfolio holdigs through periodical filigs. Sice May 2004, the Ivestmet Compay Act of 1940 madates that idividual mutual fuds disclose their portfolio holdigs quarterly i Forms N-CSR ad N-Q with a delay of o loger tha 60 days. The other importat disclosure 7

9 requiremet, madated by Sectio 13(f) of the 1934 Securities Exchage Act, is the Form 13F that requires mutual fud compaies to disclose their aggregate holdigs (at the compay level) o a quarterly basis, with o more tha a 45-day delay. 3 Although the two owership disclosure regimes described above apply i parallel, the former requiremet typically offers much more detailed iformatio about the ivestmet of mutual fuds tha that provided by the 13F form for two reasos. First, the 13F data is at the compay level oly while the N-CSR ad N-Q data is at the idividual fud level. Sice mutual fud compaies ofte operate multiple fuds, the aggregated 13F data is less iformative. Secod, 13F forms are oly filed by large ivestors (those with more tha $100 millio i 13F securities) ad iclude iformatio oly o the large (more tha 10,000 shares ad market value exceedig $200,000) positios i the 13F securities, which cosist of equities, covertible bods, ad exchage-listed optios. 4 I cotrast, N-Q ad N-CSR forms are filed by all mutual fuds for all types of securities regardless of the fud s size or the size of the positios held i idividual securities. These requiremets make the mutual fud disclosure through N-Q ad N-CSR forms more iformative tha the 13F forms filed by mutual fud families. The disclosure requiremets for idividual mutual fuds, however, have chaged over time. Prior to May 2004, the SEC oly required mutual fuds to file their portfolio holdigs twice a year usig the semi-aual N-30D form. I May 2004, the SEC eacted a ew rule 3 Istitutios filig 13F forms ca seek cofidetial treatmet o certai portfolio holdigs which, if approved by the SEC, allows them to delay the disclosure by up to oe year. See Agarwal, Jiag, Tag, ad Yag (2013) ad Arago, Hertzel, ad Shi (2012) for details. Also, the 13F forms have always bee required o a quarterly basis ad thus do ot experiece a regulatory chage i the frequecy of madatory disclosure. 4 See for more iformatio o 13F filigs. 8

10 that chaged the N-30D form to the N-CSR form, ad required mutual fuds to complete ad file the form at the ed of the secod ad fourth fiscal quarters. 5 I additio, the ew rule also required mutual fuds to file N-Q forms at the ed of the 1 st ad 3 rd fiscal quarters, thus icreasig the reportig frequecy to four times per year. 6 To balace the beefits of more trasparecy to ivestors ad the potetial costs o mutual fuds, e.g. of frot-ruig ad copycat behavior, the SEC allowed the fuds to file the disclosure forms with a 60-day delay. Before the regulatio chage i May 2004, idividual fuds could also report their portfolio iformatio more frequetly tha what the SEC required. They could use the SEC Form N-30B2 to disclose their holdigs volutarily, i additio to the required filig of the semi-aual Form N-30D. Like Form N-30D, Form N-30B2 allows fuds to disclose their portfolio holdigs, but it is filed volutarily at the fiscal quarter eds whe the N-30D forms are ot filed. Though this is a optio that fuds had prior to the regulatio chage, we fid that oly a small umber of fuds actually used it. 7 I our empirical aalyses i Sectio III, we show that less tha 2.5% of all mutual fuds volutarily disclosed their quarter-ed holdigs usig N-30B2 before Thus, the May 2004 regulatio that icreased the disclosure frequecy affected the disclosure activities of almost all mutual fuds. 8 5 Aecdotal evidece suggests that this regulatio chage o portfolio disclosure was perhaps triggered by the accoutig scadals ivolvig Ero, Worldcom, ad Tyco, ad the esuig Sarbaes-Oxley Act i July See the SEC Fial Rule IC o May 10, 2004 at 7 Some fud compaies ca choose to disclose the largest holdigs of their fuds o a quarterly basis o their websites. For example, holdigs of Fidelity OTC portfolio are available at Such volutary disclosure by fuds will bias us agaist fidig ay impact of chage i madatory disclosure o stock liquidity ad fud performace. 8 We rely o the SEC s Electroic Data Gatherig, Aalysis, ad Retrieval (EDGAR) database to determie the volutary fuds. Iterestigly, the Thomso Reuters database shows a much larger fractio of such fuds which report more tha twice prior to May Recetly, Schwarz ad Potter (2013) discuss the discrepacies betwee the fud holdigs data from EDGAR ad Thomso Reuters. We provide more details o this issue i 9

11 II. Related Literature ad Empirical Hypotheses Our paper is motivated by two strads of literature. First, a large umber of papers have show that mutual fuds disclosed portfolios cotai valuable iformatio for ivestors (e.g., Griblatt ad Titma (1989, 1993), Griblatt, Titma, ad Wermers (1995), Daiel, Griblatt, Titma, ad Wermers (1997), Wermers (1999, 2000), Che, Jegadeesh, ad Wermers (2000), Cohe, Coval, ad Pastor (2005), Kacperczyk, Sialm, ad Zheg (2005, 2008), Alexader, Cici, ad Gibso (2007), Jiag, Yao, ad Yu (2007), Kacperczyk ad Seru (2007), Cremers ad Petajisto (2009), Baker, Litov, Wachter, ad Wurgler (2010), Ciccotello, Greee, ad Rakowski (2011), Wermers, Yao, ad Zhao (2012), ad Huag ad Kale (2013)). Therefore, ay chage i the portfolio disclosure requiremet should affect the uderlyig asset markets ad idividual mutual fuds. Secod, a strad of theoretical literature studies the impact of madatory disclosure o iformed tradig (e.g., Fishma ad Hagerty (1995), Joh ad Narayaa (1997), Huddart, Hughes, ad Bruermeier (1999), Huddart, Hughes, ad Levie (2001), ad George ad Hwag (2011)). Perhaps most relevat to our cotext is the study by Huddart, Hughes, ad Levie (2001, heceforth HHL ), which exteds the Kyle (1985) model of a iformed trader by itroducig madatory disclosure of trades at the ed of each tradig period. HHL proves the existece of a mixed strategy equilibrium i which the iformed trader adds a radom oise to a liear strategy i each period to prevet the market maker from fully iferrig his private iformatio. Such a dissimulatio strategy miimizes the loss i tradig profits due to madatory disclosure. Sectio III ad repeat our aalysis usig both EDGAR ad Thomso Reuters databases. 10

12 I this paper, we exted the HHL model to cosider differet frequecies of madatory disclosure. We provide closed-form solutios of the equilibrium, aalyze the impact of disclosure frequecy o stock liquidity ad iformed trader s profits, ad produce several testable predictios. To coserve space, we preset our model ad aalytical results i the Appedix ad iclude the proofs i the Supplemetary Appedix. First, our model shows that more frequet madatory disclosure by iformed traders improves market liquidity as measured by the market depth, or the iverse of the Kyle (1985) lambda. The ituitio is that with more frequet madatory disclosure the market maker ca ifer more iformatio from the iformed trader s disclosed positios ad his order flow. This additioal iformatio leads to a reductio i the impact of iformed trades o prices. We ote that this ituitio holds eve though the iformed trader adds radom oise to his trades, because the market maker is still able to ifer some iformatio from the oisy sigal. I our empirical settig, the icrease i madatory disclosure istituted i 2004 by the SEC affects the vast majority of mutual fuds. Based o our model s predictio, if mutual fuds are i geeral iformed, we expect that stocks with a higher mutual fud owership should experiece greater icreases i liquidity tha other stocks after the regulatio chage o madatory disclosure. Secod, our model predicts that the improvemet i liquidity depeds positively o the extet of asymmetric iformatio i the stock. Whe the isider is more iformed or whe the fudametal value of the stock is subject to greater iformatio asymmetry, the market ca lear more iformatio from icreases i portfolio disclosure, causig stock liquidity to improve more. Therefore, we hypothesize the liquidity improvemet to be greater 11

13 for stocks with higher owership by more iformed fuds as compared to stocks primarily held by fuds less likely to be iformed. We also expect that the liquidity icreases deped positively o iformatio asymmetry at the stock level. Third, our model predicts a decrease i the iformed trader s profits after a icrease i the frequecy of madatory portfolio disclosure. The uderlyig ituitio is that because the market maker lears more iformatio with more frequet disclosure, the iformed trader is less able to fully reap the beefits of his iformatio. Thus, we posit that iformed fuds are likely to experiece a drop i their abormal performace as a result of more frequet portfolio disclosure after May Fially, our model predicts that the magitude of the iformed trader s profit drop depeds positively o the extet of iformatio asymmetry i the stocks disclosed. Thus, we expect the performace declie to be larger for iformed fuds whe these fuds hold stocks that are subject to greater iformatio asymmetry. Further, our model predicts that iformed traders are hurt more whe their trades take a greater umber of periods to complete. Therefore, we expect that iformed fuds that take loger to fiish their trades should experiece a greater declie i their performace. III. Data ad Variable Costructio A. Data descriptio To determie whether or ot a mutual fud volutarily reports before the regulatio chage i May 2004, we obtai the N-30D, N-30B2, N-CSR, ad N-Q forms filed by that fud from the SEC EDGAR database. We use computer programs to parse the documets ad 12

14 obtai the mutual fud idetifier iformatio ad the filig dates of these forms. Table I reports the reportig frequecies of all fuds usig these SEC forms i each year from 1994 to 2011, the period over which data is electroically available from the EDGAR database. Pael A of Table I reports the total umber of filigs of each type of forms year by year. Our results reveal several stylized facts. First, the total umber of filigs almost doubled from 6,714 i 2003 to 12,695 i 2005 as show i the last colum. We break dow the umbers for each form type ad fid that this dramatic icrease i the total umber of filigs is completely due to the itroductio of the N-Q form i The N-Q forms accouts for about half of all filigs from 2005 oward. Secod, volutary filigs by mutual fuds usig Form N-30B2 accout for oly a small portio of the total filigs ad this umber is relatively stable over time. For istace, i 2003, out of 6,714 mutual fud filigs, oly 240 (3.6%) are N-30B2 forms. This evidece suggests that the 2004 regulatio to icrease disclosure frequecy affects the disclosure behavior of almost all mutual fuds. [Isert Table I Here] Pael B of Table I presets mutual fuds aual reportig frequecies from 1994 to The results show that most fuds file twice over the period from 1994 to 2003 ad four times from 2005 to I 2004, most fuds report two or three times because the regulatio took effect oly i the secod part of that year. These patters further cofirm that the 2004 SEC regulatio has widespread effects of more frequet disclosure o mutual fuds. To idetify the effects of the 2004 SEC regulatio chage o the stock market, we cosider all actively-maaged U.S. equity mutual fuds from the Thomso Reuters Mutual Fuds Holdigs (S12) database i 2003 ad We focus o the vast majority of fuds that do ot 13

15 disclose stock holdigs to the SEC volutarily through Form N-30B2 durig the 12-moth period from May 2003 to April We use the portfolio holdigs data from the Thomso Reuters S12 database for our empirical aalyses. We idetify the fuds impacted by the regulatio chage usig the filig frequecies we obtai from the EDGAR database as described above. We merge the Thomso S12 ad the EDGAR databases as follows. First, we collect fuds filigs iformatio, icludig fud ames ad tickers, filig form types, filig dates, ad cetral idex keys (CIKs), from the EDGAR database. Next, we use the tickers from EDGAR to match with the Ceter for Research i Security Prices (CRSP) mutual fud data. We the merge the resultig data with the Thomso S12 data usig the Wharto Research Data Services (WRDS) MFLINKS tables. 9 We are able to fid CIKs for 2,582 out of the 2,658 actively-maaged U.S. equity mutual fuds i the S12 database durig our sample period. Our fial sample cosists of 2,520 fuds that disclosed o more tha two times before the regulatio chage. 10 Mutual fud holdigs data reported to the EDGAR ad Thomso Reuters databases are ot idetical ad discrepacies betwee the two have recetly bee documeted by Schwarz ad Potter (2013). Therefore, for robustess, i additio to coductig our aalyses usig Thomso S12 data, we repeat the aalyses usig the fud holdigs data from the EDGAR database ad 9 For the S12 fuds which caot be matched this way, we use the SEC s search facilities for mutual fuds ( ad for compay ames ( to maually fid the CIK for each fud to match with the S12 data. 10 There are oly 62 mutual fuds that disclosed volutarily usig the N-30B2 form ad are thus ot affected by the regulatio i We exclude them from our aalysis that follows. 14

16 report our fidigs i the Supplemetary Appedix. 11 B. Variable costructio We costruct several stock-level variables that we use i our empirical tests. First, for each stock-moth observatio, we calculate the variable Mutual Fud Owership as the aggregate owership of all actively maaged U.S. equity fuds of the stock i that moth, scaled by the total shares outstadig of the stock at the moth ed. Whe stock holdigs are ot reported by a fud at a give moth ed, we use fud s last reported stock holdigs. While the 2004 regulatio chage affects the reportig behavior of mutual fuds, it does ot affect the disclosure frequecy of other istitutioal ivestors who disclose their holdigs through the Form 13F. We use these o-mutual-fud istitutios as a cotrol group to idetify the effects of the icrease i madatory disclosure frequecy of mutual fuds. For this purpose, we defie No-MF Istitutioal Owership as the quarterly aggregate istitutioal owership from Thomso Reuters Istitutioal Holdigs (S34), excludig mutual fuds ad asset maagemet compaies. 12 I additio, we isolate hedge fuds from the o-mutual-fud istitutios to form aother cotrol group because they are arguably the most actively maaged istitutios. We defie Hedge Fud Owership as the quarterly aggregate hedge fud owership i the Thomso S34 database. Classificatio of istitutioal ivestors ad hedge fuds follows that i Agarwal, Jiag, Tag, ad Yag (2013). We costruct our sample of stocks from the CRSP stock database. We cosider all 11 We thak Christopher Schwarz for providig us with the fud holdigs from the EDGAR database. 12 Our results are qualitatively similar if we cosider a alterative measure of No-MF Istitutioal Owership by further excludig isurace compaies because may isurace compaies (such as A.I.G.) establish trusts that operate mutual fuds (see Che, Yao, ad Yu (2007)). 15

17 commo stocks from CRSP over the period May 2003 to April We choose this period to cosist of oe year prior to ad oe year after the SEC disclosure regulatio chage i May For each stock-moth, we costruct two variables to proxy for stock liquidity: Amihud illiquidity measure, the mothly average of (the logarithm of) daily Amihud measures; ad the relative bid-ask spread measure, Rspread, the mothly average of (the logarithm of) daily bid-ask spreads scaled by mid-price. 13 We compute these measures as follows, i, t i, t i, t i, t Amihud r / P * Vol (1) Aski, t Bidi, t Rspread i, t Aski, t Bidi, t / 2 (2) where i idexes stocks ad t idexes dates, r it, is the daily stock retur, P, is the daily price, ad Vol it, is the daily volume. These proxies have bee widely used i the literature (Amihud ad Medelso (1986), Amihud (2002), ad Lesmod (2005)). We also employ several commoly used stock characteristic variables as cotrols: Mometum, i.e., the mometum variable (12-moth cumulative retur) of Jegadeesh ad Titma (1993); Book-to-Market, i.e., the ratio of book equity to market equity; ad Size, i.e., the atural logarithm of market equity. To evaluate the impact of the 2004 regulatio chage, we first compute the average of mothly variables for the 12 moths prior to May 2004 ad the for the 12 moths after May 2004 (iclusive of May 2004). Next, we compute the chages i the aual averages as the differece betwee the average after May 2004 ad the average before May We deote the resultig chage variables by the prefix Δ. We report summary statistics of the variables i Pael A of Table II. We observe that it 13 We use atural logarithmic trasformatios to mitigate the effect of ay outliers. 16

18 both the Amihud ad Rspread measures decrease after May 2004, i.e., the average stock liquidity improves from 2003 to I the year prior to May 2004, mutual fuds i our sample hold, o average, 13.9% of outstadig shares of stocks. No-MF Istitutioal Owership ad Hedge Fud Owership are 22.3% ad 7.9%, respectively. [Isert Table II Here] Fially, we use two measures of fud abormal performace: i) Carhart (1997) four-factor alpha ad ii) Daiel, Griblatt, Titma, ad Wermers (1997) (DGTW) bechmark-adjusted returs. Sice we cosider the performace oe year before ad after the regulatio chage, it is oisy to estimate the i-sample alphas based o oly 12 mothly fud returs. Therefore, we compute the out-of-sample mothly alpha usig the fud returs i each moth mius the sum product of the factor returs i that moth ad the betas estimated from the 24-moth widow edig i the prior moth, as follows: R 4 ˆ ˆ F, s t 24,, t1 (3) j, s j, t1 j, k, t1 k, s j, s k1 R ˆ F j, t j, t j, k, t 1 k, t k1 4 (4) where s ad t idicate moths, j idicates fuds, R is the mothly fud retur, ad F is the mothly returs of the four factors (excess market, size, book-to-market, ad mometum). We sum the mothly alphas to obtai the aualized alpha. For the DGTW measure, we first compute the cumulative bechmark-adjusted returs betwee two successive report dates i the Thomso S12 database ad the divide them by the umber of moths i the period to obtai a mothly measure. We the sum the mothly DGTW measure to obtai the aualized figure. We will discuss the summary statistics of these measures i Sectio V. 17

19 IV. Impact of Madatory Portfolio Disclosure o Stock Liquidity A. Regulatory Chage i Madatory Disclosure ad Stock Liquidity To empirically test the effects of the chage i fuds portfolio disclosure frequecy o stock liquidity, we estimate the followig regressios of the chages i liquidity variables. For each liquidity proxy variable y, we estimate the followig cross-sectioal regressio: y MFOw y X (5) ' i, t i, t1 i, t1 i, t1 i, t where i idicates the stock, t is the year after May 2004, y it, is the chage i liquidity from the oe year before to the oe year after May 2004, MFOw it, 1 is the lagged (i.e., oe year before May 2004) Mutual Fud Owership, yit, 1 is the lagged liquidity variable, ad X it, 1 are lagged stock characteristics, icludig Mometum, Size, ad Book-to-Market ratio. The idetificatio of the regressio i equatio (5) relies o a cross-sectioal compariso of stocks with higher mutual fud owership (the treatmet group) to those with lower mutual fud owership (the cotrol group). Equatio (5) essetially uses a differece-i-differeces approach to estimate the effect of the 2004 disclosure regulatio chage o the treatmet group. 14 The first differece is the chage i stock liquidity over the 12 moths before ad after May 2004 for the stocks. The secod differece is the differece i the liquidity chages of the treatmet ad cotrol groups. Pael B of Table II reports the estimatio results of equatio (5). Our primary idepedet variable of iterest is Mutual Fud Owership. The results show that for both liquidity measures Amihud ad Rspread, the coefficiets of Mutual Fud Owership are 14 For illustratio purposes, we discuss here the case with two groups. We actually use a cotiuous variable of the mutual fud owership i the regressio but the ituitio is the same. 18

20 egative ad statistically sigificat i all the four colums at the 5% level or better. Sice lower Amihud ad Rspread imply greater liquidity, larger fud owership is associated with greater improvemet i stock liquidity after the 2004 regulatio chage. These fidigs are also ecoomically sigificat. For istace, based o the estimates i colums (2) ad (4), a oe stadard deviatio icrease i mutual fud owership is associated with a 0.19 stadard deviatio decrease i Amihud ad a 0.08 stadard deviatio decrease i Rspread. This evidece is cosistet with our model s predictio that more frequet portfolio disclosure by iformed traders will lead to a icrease i the liquidity of the uderlyig stocks they trade. B. Cross-Sectioal ad Time-Series Placebo Tests First, we coduct a set of cross-sectioal placebo tests. The above results caot rule out the possibility that mutual fud owership proxies for istitutioal owership ad stocks with higher istitutioal owership experiece greater improvemet i liquidity after May To distiguish this alterative sceario from the effect of disclosure regulatio, we add No-MF Istitutioal Owership to equatio (5) ad estimate the followig regressio: y MFOw NoMFOw y X (6) ' ' i, t i, t1 i, t1 i, t1 i, t1 i, t Ituitively, equatio (6) uses a differece-i-differece-i-differeces approach to estimate the effect of the 2004 disclosure regulatio chage o stock liquidity. The coefficiets o Mutual Fud Owership ad No-MF Istitutioal Owership represet the differece-i-differeces effect of the owership variables o chages i liquidity as discussed before i referece to equatio (5). The differece of these two coefficiets provides a estimate of the effect of the icrease i disclosure frequecy o stock liquidity 19

21 after cotrollig for o-mutual fud istitutioal owership. Amog o-mutual-fud istitutios ot affected by the icrease i disclosure frequecy, hedge fuds are arguably more actively maaged ad better iformed. Therefore, we use them as a alterative cotrol group ad estimate the followig equatio: y MFOw HFOw y X (7) '' ' i, t i, t1 i, t1 i, t1 i, t1 i, t We report the estimatio results of equatios (6) ad (7) i Pael A of Table III. The last two rows preset the differeces i the coefficiets of Mutual Fud Owership ad No-MF Istitutioal Owership (or Hedge Fud Owership) ad the correspodig test statistics. I all but oe specificatio, we fid that the mutual fud owership has a statistically greater impact o liquidity tha does o-mutual-fud istitutioal owership or hedge fud owership. These results suggest that it is ot istitutioal owership per se, but rather the icrease i mutual fud portfolio disclosure after May 2004 that leads to the improvemet i stock liquidity. [Isert Table III Here] Secod, we coduct a time-series placebo test. Specifically, we estimate equatio (5) usig the year before ad after November 2006 as our placebo period. Note that we caot choose a period prior to the regulatio chage because of evets such as the Russia sovereig bod default ad the Log-Term Capital Maagemet debacle i 1998, the burst of the dotcom bubble i 2000, ad the decimalizatio of stock prices quotes i 2001, all of which sigificatly affected stock liquidity. Furthermore, we choose the placebo period such that it is as far away from the evet date i 2004 as possible ad ot affected by the great recessio, which started i December 2007 accordig to the Natioal 20

22 Bureau of Ecoomic Research (NBER). We first estimate the regressios as i equatio (5) for the placebo period. We the compare the coefficiets for the placebo period with those for the two-year period surroudig the 2004 regulatio chage as reported i Pael B of Table II. We report the results of this compariso i Pael B of Table III. For the sake of brevity, we keep oly the coefficiets of the mutual fud owership variables. Our results show that fud owership has a positive effect o liquidity i 2004, but has either a small or isigificat effect i The differece i the effects for the two time periods is highly sigificat, as show by the F-tests i the last row. The results of our placebo test further cofirm that the liquidity chages are drive by the icrease i madatory disclosure frequecy i Take together, the results from the cross-sectioal ad time-series placebo tests i this sectio show that liquidity improvemet is cocetrated i stocks held by mutual fuds ad ot by other istitutios ad is ot drive by a temporal factor urelated to the 2004 disclosure regulatio chage. C. Mutual Fud ad Stock Subsample Aalyses Our model predicts that icreases i stock liquidity due to more frequet disclosure should be cocetrated i i) fuds that are more iformed ad ii) stocks that have greater iformatio asymmetry. I this sectio, we use subsamples of mutual fuds ad stocks to test these predictios. First, we test whether the improvemet i liquidity is cocetrated i stocks disclosed by more iformed fuds. If portfolio disclosure cotais valuable iformatio, the a 21

23 icrease i disclosure by well-performig fuds should have greater impact tha by fuds with poor past performace. To test this hypothesis, we cosider two proxies of fuds beig iformed: i) Carhart (1997) four-factor alpha ad ii) Daiel, Griblatt, Titma, ad Wermers (1997) (DGTW) bechmark-adjusted returs. Usig these two proxies, we divide the mutual fuds ito two subsamples: more iformed, i.e., the top-quartile fuds, ad less iformed, i.e., the o-top-quartile fuds. We iclude the aggregated owership of the fuds i both groups i the followig regressio ad test the differece of the coefficiets of the two owership variables: y MFOw MFOw y X (8) top ' otop ' i, t i, t1 i, t1 i, t1 i, t1 i, t Our fidigs i Table IV show that the owership of the top-quartile fuds has a statistically larger impact o liquidity tha the owership of the o-top-quartile fuds. 15 These results support our model s predictio that the market lears more iformatio from the holdigs of more iformed fuds, which results i a greater improvemet i liquidity of the disclosed stocks. [Isert Table IV Here] Secod, we ivestigate which type of stocks experiece greater icreases i liquidity as a result of the regulatory chage icreasig the disclosure frequecy. Our model predicts that the improvemet i liquidity should be cocetrated i stocks with greater iformatio asymmetry. To test this idea, we divide our sample of stocks ito subsamples based o the top 15 Our fidigs are ot affected by the possibility that fuds may try to reduce the impact of the regulatio by the widow dressig behavior. I utabulated results, we cosider the likelihood of fud widow dressig (Agarwal, Gay, ad Lig (2012)) ad fid that the improvemet i liquidity is cocetrated i the fuds that are ot proe to widow dressig behavior. 22

24 quartiles of illiquidity (Amihud or Rspread), aalyst coverage, ad market capitalizatio. We the estimate the regressio i equatio (5) for each subsample ad compare the coefficiets of fud owership for the two subsample regressios. We report the results i Table V. As show i the table, the differeces i the coefficiets of fud owership of the two subsamples have the predicted sig ad are sigificat at the 5% level or better for all four measures of iformatio asymmetry. I particular, smaller stocks, less liquid stocks, ad stocks with lower aalyst coverage beefit more from the icrease i disclosure frequecy. This evidece is cosistet with our model s predictio that more frequet disclosure leads to higher liquidity whe there is greater iformatio asymmetry i the disclosed stocks. [Isert Table V Here] Fially, for robustess, we use mutual fud holdigs obtaied from the EDGAR database, rather tha from Thomso S12 database, ad repeat all the tests i this sectio. I the Supplemetary Appedix, we report these results i Tables B.II to B.V, which are aalogous to Tables II to V. All the results are qualitatively similar. V. Impact of Madatory Portfolio Disclosure o Fud Performace Our results i the previous sectio suggest that the market lears more whe mutual fuds are required to disclose more frequetly ad, as a result, stock liquidity improves. The icrease i liquidity reduces trasactio costs ad beefits all ivestors i geeral. We ext examie how more frequet madatory portfolio disclosure affects fud performace. 23

25 A. Mutual Fud Performace ad the Regulatio Chage Our theoretical model predicts that the iformed trader s profits decrease whe madatory disclosure becomes more frequet. The ituitio is that the market learig of disclosed trades decreases the ability of the iformed traders to fully reap the beefits of private iformatio. Cosistet with this ituitio, fud maagers argue that holdigs disclosure ca lead to frot-ruig or free ridig o their trades. Both theory ad the reactio from practitioers motivate us to examie the impact of madatory disclosure o fud performace. I particular, we cosider two measures of fuds abormal performace: Carhart 4-factor alphas ad DGTW-adjusted returs. We use the aualized values of these two variables for fuds i our sample durig the oe-year periods prior to ad after May 2004, ad the calculate the differeces to measure the performace chages. For cotrol variables, we cosider fud characteristics icludig (i) TNA, defied as the total et assets uder maagemet, (ii) Turover, defied as the average aual turover from Thomso S12 mutual fud holdigs, (iii) Flow, defied as the chage i TNA scaled by lagged TNA, (iv) Expese Ratio, defied as the total operatig expeses scaled by TNA, ad (v) Load status, defied as a idicator variable which equals oe if the mutual fud has a class with load, ad zero otherwise. Specifically, we estimate the followig regressio at the fud level: Perf TopFud X (9) ' j, t 0 1 j, t1 j, t1 j, t where j idicates the fud ad t is the year after the regulatio chage. Perf, is the chage i abormal fud performace (Carhart 4-factor alpha or DGTW-adjusted retur); TopFud jt, 1is a idicator variable that equals oe if the fud is i the top quartile based o 24 jt

26 the lagged (i the year before the regulatio chage) fud performace ad zero otherwise; X jt, 1 iclude a umber of lagged fud characteristics. Pael A of Table VI reports the summary statistics of fud performace ad other fud characteristics aroud the 2004 regulatio chage. The average aualized four-factor alphas of mutual fud icrease by 1.7% after May 2004, ad the aualized DGTW-adjusted returs drop by 0.7%. This fidig idicates o clear directio of performace chage aroud 2004 for the average fud i our sample. [Isert Table VI Here] To test our model s predictio, we examie the effect of the May 2004 regulatio chage o the performace of the top-performig fuds. Pael B of Table VI reports the results of regressios of performace chages i equatio (9). I colums (1) ad (2) of Pael B, we observe that fuds with alphas i the top quartile experiece a decrease of 9.2% i aualized alphas ad a decrease of 3.8% i aualized DGTW-adjusted returs, relative to o-top-quartile fuds. Similarly, as show i colums (3) ad (4), fuds with top DGTW-adjusted returs experiece a decrease of 3.4% i alphas ad a decrease of 11.9% i DGTW-adjusted returs. All of the coefficiets are statistically sigificat at the 1% level. A potetial cocer about the above result is that the drop i performace of top-performig fuds arises from mea reversio or other factors. To alleviate this cocer, we coduct a additioal test usig the same 2006 placebo period used i Sectio IV.B. Table VII reports the results of this test ad compares them with those i Table VI usig the differece-i-differece-i-differeces approach. We fid that top-performig fuds also experiece performace deterioratios aroud the 2006 placebo period ragig from 1.3% to 25

27 7.5% depedig o the performace measure. This evidece suggests the existece of a potetial mea reversio effect. However, after accoutig for mea reversio by subtractig the 2006 coefficiets from those i 2004, the drops i fud performace remai statistically ad ecoomically sigificat. The magitude of the performace declie, et of mea reversio effect, rages from 1.9% to 4.5% o a aualized basis. The above results are cosistet with our model s predictio that more iformed fuds bear higher costs of madatory disclosure as it hiders their ability to fully beefit from their private iformatio. [Isert Table VII Here] For robustess, we also estimate all regressios i Table VII by cotrollig for chages i fud characteristics, rather tha usig lagged fud characteristics as idepedet variables. We obtai qualitatively similar results as show i Table B.VII i the Supplemetary Appedix. B. Fud Performace ad Stock Liquidity Madatory disclosure ca lower stock returs through the improvemet i liquidity, which i tur ca lead to worse fud performace. To separate this idirect effect of chages i stock liquidity o fud performace from the direct effect of fuds dimiished ability to beefit from private iformatio, we estimate the regressios of fud performace usig alphas based o a five-factor model which augmets the Carhart (1997) four-factor model with the Pástor ad Stambaugh (2003) liquidity factor. Table VIII reports the regressio results with five-factor alphas ad compares them with those based o four-factor alphas. We fid that, after cotrollig for the impact of the 26

28 liquidity factor o fud performace, top-performig fuds experiece decreases i their five-factor alphas of 3.2% to 7.7% o a aualized basis i As i the case of four-factor alphas, we also cotrol for potetial mea reversio i the case of five-factor alphas. After subtractig the coefficiets from the 2006 placebo period, the decreases i five-factor alphas rage from 1.6% to 3.3% o a aualized basis, which is about three quarters of the correspodig decreases for four-factor alphas (from 1.9% to 4.5%). Moreover, the differeces betwee the results for four-factor ad five-factor alphas are small ad statistically isigificat i all but oe case, as show i the last colum of Paels A ad B i Table VIII. [Isert Table VIII Here] C. Fud Performace ad Iformatio Asymmetry of Stocks Our model predicts that whe tradig stocks with greater iformatio asymmetry, the iformed trader will experiece greater losses because of the revelatio of more valuable iformatio to the public. To test this predictio, we ext ivestigate whether the top-performig fuds experiece a larger drop i performace whe they also hold stocks with higher levels of iformatio asymmetry. We first calculate fud-level iformatio asymmetry measures usig stock size, aalyst coverage, Amihud illiquidity, ad relative spread by value-weightig these measures based o the amout the fud ivested i these stocks at the time of disclosure. We the create idicator variables that equal oe if a fud is i the top quartile for a fud-level measure of iformatio asymmetry. We estimate regressios of fud performace chages o 27

29 the iteractios of past fud performace ad the iformatio asymmetry variables. Based o our model s predictio, we expect the coefficiets of these iteractios to be egative ad sigificat. Table IX presets the results of these regressios. Cosistet with our model s predictios, we fid that the top-performig fuds that also hold stocks with high levels of iformatio asymmetry experiece the greatest declies i performace. For example, a top-quartile (based o past alpha) fud whose portfolio cotais small stocks suffers a additioal 1.6% (3.0%) decrease i alpha (DGTW-adj. retur) compared to a top-quartile fud that holds large stocks. Similarly, a top-quartile (based o past DGTW-adjusted retur) fud that holds small stocks experieces a additioal 5.2% (6.5%) declie i alpha (DGTW-adj. retur) relative to other top-quartile fuds. [Isert Table IX Here] D. Fud Performace ad Trade Legth The regulatio chage i May 2004 required mutual fuds to icrease their disclosure frequecy from twice a year to four times a year. Our model predicts that if a iformed trader completes his trades over a loger period, he will be more adversely affected by disclosure. This predictio implies that the regulatio chage would have a eve greater adverse effect o fuds that take loger to complete their tradig strategies. To test this predictio, we costruct a variable Trade Legth as follows. First, for each stock for a give fud-quarter, we costruct a positio-level measure by coutig the umber of cosecutive quarters over which the fud either builds or uwids the positio i that stock 28