Hedge Funds and Earnings Momentum

Transcription

1 Inernaional Journal of Economics and Finance; Vol. 9, No. 10; 2017 ISSN X E-ISSN Published by Canadian Cener of Science and Educaion Hedge Funds and Earnings Momenum Daniel T. Lawson 1 1 Eberly School of Business and Technology, Indiana Universiy of Pennsylvania, Indiana, PA, USA Correspondence: Daniel T. Lawson, Eberly School of Business and Technology, Indiana Universiy of Pennsylvania, Indiana, PA, USA. Tel: dlawson@iup.edu Received: Augus 3, 2017 Acceped: Augus 16, 2017 Online Published: Sepember 5, 2017 doi: /ijef.v9n10p40 URL: hps://doi.org/ /ijef.v9n10p40 Absrac This sudy deermines if hedge funds ake advanage of he earnings momenum anomaly. A five-facor model was used including Fama and French (1993) and Carhar (1997) facors as well as an earnings momenum facor based on Chordia and Shivakumar (2007). The average hedge fund does no ake advanage of he pos-earnings momenum drif; however, larger funds associaed wih equiy long only and equiy shor bias sraegies successfully arbirage on he earnings anomaly, conribuing 2-3% per year, respecively. In conras, funds wih even driven, fund iming, and converible arbirage sraegies end o employ a sraegy opposie o ha of he earnings momenum anomaly and suffer losses accordingly. Keywords: hedge funds, earnings momenum, marke anomalies, arbirage, behavioral finance 1. Inroducion 1.1 Savviness of Hedge Fund Managers Managers of hedge funds are generally hough of as savvy invesors ha are able o produce superior reurns due o heir sophisicaion. I is generally believed ha heir superior reurns come from exploiing marke inefficiencies ha he ypical invesor is eiher unaware of or unable o mimic. Since hedge funds are no limied o specific asse classes like muual funds or oher highly regulaed funds, and heir holdings are generally no made public, hey would be in he bes posiion o ake advanage of any marke anomalies ha may exis. Reurn anomalies associaed wih momenum, accruals and equiy financing have been well documened in he lieraure (Fama & French, 2007). This sudy specifically looks a earnings momenum, or pos-earnings-announcemen drif, which is he phenomenon firs documened by Ball and Brown in Tes of Savviness To deermine wheher or no hedge funds successfully arbirage on he earnings momenum anomaly, he four-facor model of Fama and French (1993) and Carhar (1993) is used along wih a facor capable of capuring he anomalous behavior of earnings momenum (Chordia & Shivakumar, 2007). Significan posiive loadings on he earnings momenum facor are found wih hedge funds wih equiy long only and equiy shor bias sraegies. Significan negaive loadings are found wih hedge funds wih even driven, converible arbirage and fund iming sraegies. The resuls sugges ha alhough he earnings momenum anomaly is persisen, he average hedge fund is unable o capialize on i. Moreover, he average hedge fund manager ha aemps o acively ake advanage of he earnings announcemens experiences significan negaive reurns, which bolsers suppor ha markes are highly efficien. 2. Lieraure Review 2.1 Marke Anomalies Much of he earlier lieraure on hedge funds, including Fung and Hsieh (1997), Brown, Goezman, and Park (1997), and Brown, Goezman, and Ibboson (1999), focuses on he general characerisics of hedge fund performance, revealing many differen ypes of hedge fund sraegies wih very differen reurn characerisics. Fung and Hsieh (1997) and ohers demonsrae ha mos hedge funds reurn have low correlaion wih sandard asse indices, while Fung and Hsieh (1997, 2001), Michell and Pulvino (2001), and Agarwal and Naik (2004) show cerain ypes of hedge funds have nonlinear risk reurn characerisics ha exhibi payoffs resembling opion-like feaures. Earnings momenum, or pos-earnings-announcemen drif, is he phenomenon firs documened by Ball and 40

2 ijef.ccsene.org Inernaional Journal of Economics and Finance Vol. 9, No. 10; 2017 Brown (1968) in which abnormal sock reurns coninue o drif up afer favorable earnings announcemens and down for unfavorable earnings announcemens. Subsequen sudies, including Foser, Olsen, and Shevlin (1984), Bernard and Thomas (1989), Hew, Skerra, Srong, and Walker (1996), and Booh, Kallunki, and Marikainen (1996), confirm heir findings. Loughran and Rier (1995, 2000) find ha firms issuing IPOs and SEOs have low subsequen reurns. Lakonishok and Vermaelen (1990) and Ikenberry, Lakonishok, and Vermaelen (1995) find ha sock repurchasing firms have srong subsequen reurns. Poniff and Woodgae (2005) show ha he change in ousanding shares exhibis a srong cross-secional abiliy o predic sock reurns, and Daniel and Timan (2006) and ohers provide evidence ha he issuance of shares is srongly negaively relaed o fuure reurns. Cooper, Gulen, and Schill (2007) show ha here is a srong negaive correlaion beween a firm s asse growh rae and subsequen abnormal reurns, while Fama and French (2007) find he asse growh anomaly is presen only in average reurns on microcap and small socks. If hedge fund managers are acually arbiraging marke anomalies, hen here should be significan loadings on anomaly facors associaed wih he pricing inefficiencies exploiaion sraegies. In his sudy, we specifically es wheher hedge funds successfully arbirage on he earnings momenum anomaly. 3. Mehodology 3.1 Baseline Daa We use reurn daa of 3,068 individual hedge funds from Barclay Hedge Fund DaaFeeder from January 1990 hrough December 2005, a period wih boh high and low growh. We exclude hedge funds wih less han 24 consecuive monhs of reurn daa, non-u.s. currency and reurns repored oher han in ne-of-fees. This leaves a final sample of 1,460 individual hedge funds. Fama and French (1993) and Carhar (1997) facors are obained from Kenneh French s websie and daa used o consruc he earnings momenum facor are obained from Bloomberg. 3.2 Survivorship Bias Since providing informaion o any hedge fund daabase is volunary, cerain biases need o be considered when inerpreing reurn daa. The mos noable source of bias is referred o as survivorship bias, which occurs when all or par of he reurns of liquidaed or non-reporing funds are excluded from he daa se. These graveyard funds ypically had subsandard performance, so excluding hese from he analyses produces an upward bias o a rue hedge fund porfolio. Anoher bias is he insan hisory bias (Park, 1995). Fung and Hsieh (2000) explain ha new hedge funds ypically go hrough an incubaion period where hey rade on money from friends and family, and, once hey have achieved a record of good performance, hey marke hemselves o he daabase vendors. This insan hisory bias has he poenial o upwardly bias reurn esimaes since enering hedge funds may exclude prior monhs of subsandard performance. Fung and Hsieh (2006) esimae he survivorship bias o be around 1.8% o 2.4% and he incubaion bias o be approximaely 1.5% annually. These biases will be presen in he alphas repored here. If arbiraging he earnings anomaly conribues o fund success, hen i is possible our loading esimaes on he anomaly facor may be upwardly biased. 3.3 Creaion of Earnings Momenum Anomaly Facor To creae PMN, we sor non-financial NYSE, AMEX, and NASD firms were sored ino deciles based on heir sandardized unexpeced earnings (SUE), where SUE is defined as he curren-quarer earnings per share less earnings per share four quarers ago sandardized by he earnings sandard deviaion over he pas eigh quarers. Earnings were obained from Compusa. Decile porfolios (SUE porfolios) were formed monhly by equally weighing all firms in he decile rankings, and he posiions were held for he following six monhs. This creaed porfolios wih overlapping holding periods since he porfolio in monh consiss of he posiion in he curren monh, as well as he posiions held in he previous five monhs. Each of he six porfolios received one-sixh weigh. By updaing he decile porfolios monhly, he weighs were revised on one-sixh of he securiies in he enire porfolio every monh. PMN is defined as he difference in reurns beween he highes and he lowes SUE porfolios. Chordia and Shivakumar (2006) find a srong link beween PMN and fuure sock reurns and ha PMN subsumes he effec of sock-price momenum. A posiive coefficien is expeced on PMN if he facor is being successfully arbiraged. To avoid muli-collineariy in he resuls, he PMN facor is orhogonalized o he Fama and French (1993) and Carhar (1997) facors, which allows us o view he specific conribuion he earnings momenum facor makes owards he reurns. PMN MKT SMB HML UMD = 1, 2,, T (1) where PMN is he earnings momen facor, MKT, SMB, and HML are he Fama and French (1993) marke, size, 41

3 ijef.ccsene.org Inernaional Journal of Economics and Finance Vol. 9, No. 10; 2017 and value facors, respecively, UMD is he Carhar (1997) momenum facor, and ε is an error erm. λs are coefficiens o be esimaed. The residual and inercep are hen summed o form he orhogonalized earnings momenum facor: PMNORTHANOM = λ 0 + ℇ = 1, 2,, T (2) where PMNORTHANOM is he orhogonalized earnings momenum facor. All regression resuls repored include only he orhogonalized earnings momenum values. 3.4 Model I We use equally-weighed and value-weighed porfolios of hedge funds in all regressions. In he firs se of regressions, he excess reurns of all he hedge funds are regressed on he Fama and French (1993) and Carhar (1997) facors (hereinafer, four-facor model). We hen regress he reurns of sraegy-specific hedge funds on he same four-facor model. R MKT SMB HML UMD = 1, 2,.., T (3) where R is he monhly mean equal-weighed (value-weighed) reurn of he all hedge minus he risk-free rae of reurn in monh, MKT, SMB, and HML are he Fama and French (1993) marke, size, and value facors, respecively, UMD is he Carhar (1997) momenum facor, and ε is an error erm. λs are coefficiens o be esimaed. 3.5 Model II In he second se of regressions, he excess reurns of all hedge funds are regressed on he Fama and French (1993) and Carhar (1997) facors and he earnings momenum facor (hereinafer, five-facor model). We hen regress he reurns of sraegy-specific hedge funds on he same five-facor model. R MKT SMB HML UMD PMN = 1, 2,., T (4) where PMN is he earnings momenum facor (posiive minus negaive). The monhly mean equal-weighed (value-weighed) excess reurns of he sraegy-specific porfolios are regressed on all of he same specificaions described above. 4. Resuls 4.1 Descripive Saisics Descripive saisics for he earnings momenum facor, he Fama and French (1993) marke, size and value facors, and he Carhar (1997) momenum facor are provided in Table 1. The average monhly reurn for he earnings momenum facor is 1.08%, which is similar o he value repored by Chordia and Shivakumar (2006). Table 1. Mean monhly reurns of regression facors MKT SMB HML UMD PMN Observaions Mean % Model I Regression Resuls of All Hedge Funds Table 2 repors he ne-of-fees monhly mean excess reurns of all hedge funds on an equal-weighed and value-weighed basis. Table 2 has been condensed for breviy. The complee able can be obained by conacing: dlawson@iup.edu. Coefficiens on he Fama and French (1993) facors are all posiive and saisically significan a he 1% level, while he Carhar (1997) momenum facor is posiive and significan a he 10% and 5% level for he equal-weighed and value-weighed regressions, respecively. Alphas (inercep) are posiive and significan indicaing ha he average hedge fund reurns are no fully explained by his model. The coefficien for PMN is negaive and significan a he 10% level for he equal-weighed regression and insignifican for he value-weighed regression. These resuls sugges ha he earnings momenum is no a significan sraegy employed by he average hedge fund, which is no surprising given he variey of sraegies employed by hese funds. The nex se of regressions will examine wheher he earnings momenum anomaly is aken advanage of by specific sraegies of hedge funds. Table 2. Earnings momenum facor loading Equal-weighed Value-weighed

4 ijef.ccsene.org Inernaional Journal of Economics and Finance Vol. 9, No. 10; Model II - Sraegy Specific Regression Resuls Panels A1 and A2 of Table 3 repor he regression resuls of hedge funds wih equiy long only sraegies. Table 3 has been omied for breviy. The complee able can be obained by conacing: dlawson@iup.edu. Coefficiens on he Fama and French (1993) facors are significan, while he coefficiens on he Carhar (1997) momenum facor are insignifican. In he five-facor model, he coefficien on PMN is posiive and significan a he 1% level for he value-weighed regression, which indicaes ha PMN conribues.18% per monh (.171 x 1.076, coefficien imes he mean, respecively) or 2.16% per year o he reurns of larger hedge funds ha employ an equiy long only sraegy. Panels B and C of Table 3 repor he regression resuls of he hedge funds wih equiy long bias and equiy long shor sraegies. Coefficiens on he Fama and French (1993) facors are significan (excep for HML for equiy long shor), and he coefficiens on he Carhar (1997) momenum facor are insignifican. In he five-facor model, he coefficiens on PMN are insignifican, suggesing ha hese funds do no employ a sraegy consisen wih earnings momenum. Panels D1 and D2 of Table 3 repor he regression resuls of he hedge funds wih equiy shor bias sraegies. Coefficiens on he Fama and French (1993) facors and he Carhar (1997) momenum facor are significan a he 5% level or greaer. No only are beas significanly negaive, hey are below negaive one, indicaing he shoring of high risk socks. In he five-facor value-weighed model, PMN is posiive and significan a he 10% level wih a repored coefficien of.224. This indicaes ha using PMN conribues.24% per monh (2.89% annually) o he reurns of larger funds wih equiy shor bias sraegies. Alphas are posiive and highly significan. Regression resuls for hedge funds wih equiy marke neural sraegies are repored in Panels E1 and E2 of Table 3. Coefficiens on he marke, value, and momenum facors are significan a he 10% level or greaer. Alhough numerically very close o zero, he posiive and significan coefficiens on he marke facor indicae ha hese funds are no neural in he absolue sense. However, he reurns of hese funds depend significanly less on he marke porfolio hen he average hedge fund in Table 3. For all hedge funds, he coefficiens for he marke facor are posiive and significan a he 1% level. In he equal- (value) weighed regression, he marke facor conribues.218% (.179%) per monh o reurns. For he equiy marke neural equal- (value) weighed regression, he marke conribuion is.020% (.015%) per monh o reurns. Coefficiens on he size facor are insignifican. The coefficien for PMN is posiive and significan for he equal-weighed five-facor model. This is in conras o he insignifican coefficien for PMN in Table 2. Thus, relaively smaller equiy marke neural funds appear o go long firms wih posiive earnings surprises and/or shor negaive surprise firms, a posiion ha conribues.05% per monh o reurns. Alphas are posiive and highly significan, bu somewha smaller han in Table 2. Panels F1 and F2 of Table 3 repor he regression resuls of he hedge funds wih even driven sraegies. Coefficiens on he Fama and French (1993) facors are significan a he 1% level. Coefficiens on he Carhar (1997) momenum facor are significan a he 5% level for he value-weighed regression and insignifican for he equal-weighed regression. The coefficien for PMN is negaive and marginally significan in he value-weighed five-facor model, which is differen from he insignifican coefficien for PMN in Table 2 for all hedge funds. Thus, some of he relaively larger even driven funds seem o go long firms wih negaive earnings surprises and/or shor firms wih posiive earnings surprises, a posiion ha lowers reurns by.090% per monh. Alphas are posiive and highly significan. Regression resuls for hedge funds wih fund iming sraegies are repored in Panels G1 and G2 of Table 3. Coefficiens on he marke and value facors are significan a he 1% level. The coefficiens on he size facor are significan a he 5% level for equal-weighed regressions and insignifican for value-weighed regressions. Coefficiens on he earnings momenum facor are negaive and significan a he 1% level for he equal-weighed regression, suggesing ha he smaller fund iming hedge funds employs a sraegy opposie o he earnings momenum facor and loses.15% per monh by doing so. Alphas are posiive and highly significan. Panels H1 and H2 of Table 3 repor he regression resuls of he hedge funds wih converible arbirage sraegies. Coefficiens on he Fama and French (1993) facors are significan a he 5% level or greaer and he coefficiens on he Carhar (1997) momenum facor are insignifican. The coefficiens on he marke facor are considerably smaller han hose in Table 3, indicaing ha, unlike he average hedge fund, converible arbirage funds end o be more marke neural. For he five-facor model specificaions, he coefficiens on PMN are negaive and saisically significan a he 1% level for he equal-weighed regression and a he 10% level for he value-weighed regression. This suggess ha he hedge funds wih converible arbirage sraegies end o go long firms wih negaive earnings surprises and/or shor firms wih posiive earnings surprises and he larger funds, a posiion ha lowers he larger funds by approximaely 1.57% per year. Alphas are posiive and highly significan. 43

5 ijef.ccsene.org Inernaional Journal of Economics and Finance Vol. 9, No. 10; 2017 Tables 3 show saisically significan posiive and negaive loadings on PMN for differen hedge funds. Alhough he PMN facor is insignifican for he average hedge fund (Table 2) and for some of he sraegy specific funds, i is eviden ha posiive loadings on PMN provides economically significan reurns for larger hedge funds wih equiy long only and equiy shor bias sraegies (2.16% and 2.89% per year, respecively. I is also clear ha negaive loading on PMN provides economically significan losses for even driven and converible arbirage hedge funds and smaller fund iming hedge funds. 5. Discussion This sudy uses a five-facor model o deermine wheher hedge funds successfully arbirage he earnings momenum anomaly. The resuls indicae ha he average hedge fund does no ake advanage of he pos-earnings announcemen drif. However, he observed alphas indicae ha hese funds do employ sraegies ha provide excess reurns no explained by he radiional Fama and French (1993) and Carhar (1997) facors. Larger hedge funds wih sraegies of equiy long only and equiy shor bias do appear o arbirage on he earnings momenum facor, a sraegy ha conribues 2-3% o heir reurns each year. In conras, funds wih sraegies such as even driven, fund iming and converible arbirage end o employ a sraegy opposie of he earnings momenum anomaly and suffer losses accordingly. The posiive and negaive loading on he earnings momenum facor sugges ha hese findings are boh saisically and economically significan. The resuls of his sudy are ineresing for a few reasons. Firs, i is surprising ha he average hedge fund fails o exploi an anomaly ha has been known for more han fory years. Second, and perhaps more surprising, many hedge funds employ a sraegy couner o he earnings momenum anomaly and lose money by doing so. This raises he quesion as o wheher i is possible o consisenly arbirage on he earnings momenum anomaly, wheher he earnings momenum anomaly sill exiss, and/or wheher he average hedge fund manager is as financially savvy as purpored. References Agarwal, V., & Naik, N. (2004). Risks and porfolio decisions involving hedge funds. Review of Financial Sudies, 17, hps://doi.org/ /rfs/hhg044 Ball, R., & Brown, R. (1968). An empirical evaluaion of accouning numbers. Journal of Accouning Research, 6, hps://doi.org/ / Bernard, V., & Thomas, J. (1989). Pos-earnings announcemen drif: delayed price response or risk premium. Journal of Accouning Research, 27, hps://doi.org/ / Boehme, R., & Sorescu, S. (2002). The long-run performance following dividend iniiaions and resumpions: Underreacion or produc of chance? Journal of Finance, 57, hps://doi.org/ / Booh, G., Kallunki, J., & Marikainen, T. (1996). Pos-announcemen drif and income smoohing: Finnish evidence. Journal of Business, Finance and Accouning, 23, hps://doi.org/ /j b01165.x Brown, S., Goezman, W., & Ibboson, R. (1999). Offshore hedge funds: Survival and performance Journal of Business, 72, hps://doi.org/ / Brown, S., Goezman, W., & Park, J. (1997). Condiions for survival: Changing risk and he performance of hedge fund managers and CTAs. Working Paper, New York Universiy Sern School of Business, Yale School of Managemen, and Long Island Universiy. Carhar, M. (1997). On persisence in muual fund performance. Journal of Finance, 52, hps://doi.org/ /j b03808.x Chordia, T., & Shivakumar, L. (2006). Earnings momenum and price momenum. Journal of Financial Economics, 80, hps://doi.org/ /j.jfineco Cooper, M., Gulen, H., & Schill, M. (2007). Asse growh and he cross-secion of sock reurns. Journal of Finance, forhcoming. hps://doi.org/ /ssrn Daniel, K., & Timan, S. (2006). Marke reacions o angible and inangible informaion. Journal of Finance, 61, hps://doi.org/ /j x Fama, E., & French, K. (1993). Common risk facors in he reurns on sock and bonds. Journal of Financial Economics, 33, hps://doi.org/ / x(93)

6 ijef.ccsene.org Inernaional Journal of Economics and Finance Vol. 9, No. 10; 2017 Fama, E., & French, K. (2007). Dissecing anomalies. Working Paper, Universiy of Chicago and Darmouh College. hps://doi.org/ /ssrn Foser, G., Olsen, C., & Shevlin, T. (1984). Earnings releases, anomalies and he behavior of securiy reurns. Accouning Review, Fung, W., & Hsieh, D. (1997). Empirical characerisics of dynamic rading sraegies: The case of hedge funds. Review of Financial Sudies, 10, hps://doi.org/ /rfs/ Fung, W., & Hsieh, D. (2000). Performance characerisics of hedge funds and commodiy funds: Naural vs. spurious biases. Journal of Financial and Quaniaive Analysis, 35, hps://doi.org/ / Fung, W., & Hsieh, D. (2006). Hedge funds: An indusry in is adolescence. Federal Reserve Bank of Alana Economic Review, 91, Hew, D., Skerra, L., Srong, N., & Walker, M. (1996). Pos-earnings announcemen drif: Some preliminary evidence for he UK. Accouning and Business Research, 26, hps://doi.org/ / Ikenberry, D., Lakonishok, J., & Vermaelen, T. (1995). Marke underreacion o open marke share repurchases. Journal of Financial Economics, 39, hps://doi.org/ / x(95)00826-z Lakonishok, J., & Vermaelen, T. (1990). Anomalous price behavior around repurchase ender offers. Journal of Finance, 45, hps://doi.org/ /j b03698.x Loughran, T., & Rier, J. (1995). The new issue puzzle. Journal of Finance, 50, hps://doi.org/ /j b05166.x Loughran, T., & Rier, J. (2000). Uniformly leas powerful ess of marke efficiency. Journal of Financial Economics, 55, hps://doi.org/ /s x(99) Michell, M., & Pulvino, T. (2001). Characerisics of risk and risk arbirage. Journal of Finance, 56, hps://doi.org/ / Park, J. (1995). Managed fuures as an invesmen asse (Docoral disseraion, Columbia Universiy). Poniff, J., & Woodgae, A. (2005). Shares ousanding and cross-secional reurns. Working paper, Boson College and Universiy of Washingon. hps://doi.org/ /ssrn Copyrighs Copyrigh for his aricle is reained by he auhor(s), wih firs publicaion righs graned o he journal. This is an open-access aricle disribued under he erms and condiions of he Creaive Commons Aribuion license (hp://creaivecommons.org/licenses/by/4.0/). 45