Do Shareholder Preferences Affect Corporate Policies?

Transcription

1 Do Shareholder Preferences Affec Corporae Policies? Johan Sulaeman Edwin L. Cox School of Business Souhern Mehodis Universiy January 2009 This paper benefied grealy from he coninuous suppor of my disseraion commiee, Aydoğan Alı, Jennifer Huang, Tom Shively, Laura Sarks, and Sheridan Timan. I also hank Ramin Baghai-Wadji, Will Goezmann, David Hirshleifer, Jonahan Karpoff, Ken Kavajecz, Andrey Ukhov, Kumar Venkaaraman, Rong Wang, and Toni Whied for heir helpful commens and discussion. I am also graeful for he commens received from paricipans a he 2007 Financial Managemen Associaion, 2008 European Finance Associaion and 2008 Asian Finance Associaion annual meeings and seminars a Baruch College, Baylor Universiy, Drexel Universiy, Emory Universiy, Indiana Universiy, McGill Universiy, Souhern Mehodis Universiy, Universiy a Buffalo, UC Irvine, Universiy of Colorado, Universiy of Georgia, Universiy of Hawaii, Universiy of Souh Carolina, Universiy of Texas (Ausin), Universiy of Washingon, Universiy of Waerloo, Universiy of Wisconsin (Milwaukee), and Universiy of Wisconsin (Madison). All errors are enirely my own.

2 Do Shareholder Preferences Affec Corporae Policies? Absrac This paper examines he preferences of insiuional invesors for firm policies and he relaionship beween hese preferences and firm decisions. I find ha insiuions exhibi sysemaic differences in heir preferences for financial and invesmen policies of he firms hey inves in. Furhermore, hese preferences are relaed o subsequen changes in he financial and invesmen policies of firms. In paricular, a firm is more likely o decrease (increase) is leverage raio if is curren leverage is higher (lower) han he preferences of is insiuional shareholders. Overall, he evidence suggess ha he preferences of insiuional shareholders are imporan deerminans of corporae policies.

3 I is well undersood ha differen insiuional invesors have differen invesmen syles and end o hold socks wih differen characerisics - e.g., value versus growh. In his paper, I documen ha he invesmen choices of insiuional invesors also reveal heerogeneous preferences for financial and operaing characerisics. In paricular, some insiuional invesors end o inves in firms ha are underleveraged relaive o oher firms in heir indusries wih similar characerisics, while ohers end o inves in firms ha are overleveraged. I also find ha he preferences of he insiuional invesors ha hold a firm s sock predic he firm s fuure financial and invesmen policies. Insiuional invesors can have heerogeneous preferences regarding he invesmen and financial policies of he firms hey include in heir porfolios for a variey of reasons. Invesors can have differen opinions on he economic oulook, which can affec heir preferences hrough several differen channels. For example, invesors may have differen opinions abou a firm s invesmen sraegy, preferring a more aggressive invesmen sraegy when hey have confidence in managemen and believe he economy is growing bu a less aggressive invesmen sraegy when hey have less confidence in boh managemen and he economic growh prospecs. Similarly, invesors wih a negaive opinion on growh prospecs may prefer a higher amoun of deb o reduce he availabiliy of free cash flow and he flexibiliy managers have in invesing in negaive NPV projecs (Jensen (1986)). Of course, he beneficial effec of deb in reducing managerial flexibiliy o ake value-desroying projecs comes a a cos as managers saddled wih oo much deb may underinves and pass up posiive NPV projecs due o deb overhang (Myers (1977), Sulz (1990), and Har and Moore (1995)). My analysis indicaes ha insiuional invesors display sysemaic differences in heir preferences. I measure each insiuional invesor s preference as he average characerisic 1

4 of he socks in is porfolio. The null hypohesis ha he preferences for leverage (i.e., he average porfolio deb raios) of all insiuional invesors are he same is srongly rejeced boh for he raw deb raios and when he deb raios are measured relaive o he expeced deb raios based on indusry and firm characerisics. When insiuions are sored ino quiniles based on he average leverage of heir holdings, insiuions in he op quinile hold socks ha have, on average, 13 percenage poins higher leverage han he socks held by insiuions in he boom quinile afer conrolling for various indusry and firm characerisics. I confirm using a simulaion approach ha his difference is significanly higher han wha one would observe if insiuional invesors picked socks randomly. Moreover, he heerogeneiy persiss when insiuional invesors iniiae new posiions: he new socks iniially purchased by insiuions currenly holding underleveraged socks are more underleveraged relaive o he new socks purchased by insiuions holding overleveraged socks. Afer documening he heerogeneiy in he leverage preferences of insiuional invesors, I analyze he exen o which hese preferences influence fuure corporae decisions. I firs infer a paricular firm s insiuional preference by aggregaing he preferences of all of is insiuional shareholders. I hen examine wheher his aggregaed preference is relaed o fuure policy decisions of he firm. In paricular, I ask wheher a firm held by invesors ha prefer lower (higher) leverage subsequenly makes financial decisions ha decrease (increase) is deb raio afer conrolling for various deerminans of deb raio. I find ha his is indeed he case: abou 17 percen of he gap beween he aggregaed preference of insiuional shareholders and he firm s curren deb raio is closed wihin he firs year afer conrolling for various firm and indusry characerisics. Quaniaively, he esimaed effec of insiuional preferences on capial srucure choices is as srong as he reversion o he arge leverage proxy previously used in he lieraure. 2

5 I also examine wheher his relaionship is more pronounced in firms wih cerain characerisics. In paricular, I focus on firm characerisics ha are relaed o he likelihood ha firm managers accommodae he preferences of heir shareholders. If he relaionship beween insiuional preferences and fuure firm decisions is driven by insiuional influence, his relaionship should be less pronounced when firm managers are less likely o accommodae he preferences of insiuional shareholders. Consisen wih his hypohesis, he effec of insiuional preferences is less pronounced when he firm is a consiuen of S&P 500 (and herefore has a relaively larger poenial shareholder base), larger, older, or managed by hese CEOs wih weaker career concerns due o heir age or heir dual posiion as he chairman of he board. Lasly, I examine he poenial moivaion for firms o incorporae he preferences of insiuional shareholders in heir policy choices. In paricular, I focus on wheher ignoring he preferences of insiuional shareholders has any negaive effecs on sock prices. I find ha firms ha change heir leverage raios in he opposie direcion of heir shareholders preferences experience a higher insiuional exi rae and lower sock reurns han firms ha change heir leverage raios in he direcion of heir shareholders preferences. Firms ignoring he preferences of heir shareholders underperform hose following he preferences of heir shareholders by 6.12% in he year afer he capial srucure decisions were made (afer conrolling for size, B/M, and momenum facors). This gap in sock performance is no reversed in he subsequen years, which suggess ha he inferior sock performance is reflecing a permanen change in firm valuaion and no due o shor-erm price pressures. This paper is relaed o he long line of lieraure covering invesors preferences and clieneles. Researchers in his area have documened he preferences of insiuional invesors for socks wih cerain characerisics, such as socks wih high volailiy, high-price socks, 3

6 liquid socks, and socks of large firms (Falkensein (1996), Benne, Sias, and Sarks (1998), Gompers and Merick (2001), Badrinah, Kale, and Ryan (1996) and Del Guercio (1996)). The curren paper exends his lieraure in wo ways. Firs, while he lieraure has examined he preference of insiuional invesors as a monolihic group (eiher for he whole universe of insiuional invesors or for each ype of insiuional invesors), he analysis in his paper focuses on he disinc preference of each insiuional invesor. Second, he curren paper is he firs o examine he heerogeneous preferences of insiuional invesors for financial and invesmen characerisics. 1 This paper also exends he vas lieraure on he role of insiuional invesors as poenial moniors of firm managers. 2 The mehodology used in his paper allows for an examinaion of he influence of insiuional invesors using a considerably broader sample han previous sudies ha focus on shareholder proposals or argeing by specific insiuions/insiuion groups. 3,4 Furhermore, he aggregaed preference of insiuional invesors provides a benchmark o measure he exen of hese invesors influence in firm decisions hrough boh direc acions (such as shareholder proposals or proxy voing), indirec acions (he possibiliy of 1 The curren paper is also indirecly relaed o he lieraure on dividend clieneles (Grinsein and Michaely (2005)). Alhough he curren paper does no analyze firm payou decisions, he mehodology used in his paper can be exended o examine insiuional invesors heerogenous preferences for dividend and heir poenially conflicing effecs on firm payou policy. These heerogeneous preferences can be driven by boh agency-relaed reasons (e.g., invesors may disagree on he opimal payou policy ha minimizes he poenial coss of agency issue) and ax-relaed reasons (e.g., ax-exemp insiuions may be araced by firms ha pay dividends as hypohesized by Allen, Bernardo, and Welch (2000)). 2 This monioring role can be boh direc and indirec. Indirecly, insiuional invesors can faciliae monioring hrough he marke for corporae conrol by faciliaing akeovers (Brickley, Lease, and Smih (1997), Jarrell and Poulsen (1987), and Agrawal and Mandelker (1992)). Insiuional invesors holding large sakes can also ac as direc moniors by influencing he acions of firm managers (Shleifer and Vishny (1986) and Admai, Pfleiderer, and Zechner (1994)). 3 A parial lis of sudies focusing on shareholder proposals includes Karpoff, Malaesa, and Walkling (1996), Daily, Johnson, Ellsrand, and Dalon (1996), Bizjak and Marquee (1998), Del Guercio and Hawkins (1999), Gillan and Sarks (2000), Prevos and Rao (2000), and Mavos and Osrovsky (2006). 4 Gillan and Sarks (2007) provide a recen survey of he srand of lieraure ha focuses on specific insiuional invesor groups. The sudies in his lieraure include Nesbi (1994), Wahal, Wiles, and Zenner (1995), Smih (1996), Srickland, Wiles, and Zenner (1996), Wahal (1996), and Carleon, Nelson, and Weisbach (1998). 4

7 selling (Parrino, Sias, and Sarks (2003))), and, more imporanly, unobservable communicaions beween hese invesors and firm managemen. In addiion o he leverage and invesmen decisions explored in his paper, his benchmark can be applied o oher ypes of firm decisions o gain useful insighs on he general role of insiuional invesors in corporae policy decisions. This paper is closely relaed o several recen sudies. Focusing on firm acquisiion decisions, Harford, Jener, and Li (2007) documen ha cross-holdings by a firm s shareholders affec is choice of akeover arges. In he mos closely relaed sudy o his paper, Cronqvis and Fahlenbrach (2007) documen he heerogeneiy of blockholders in erms of he financial, invesmen, and execuive compensaion policies of he socks hey hold. Alhough he curren paper also documens he heerogeneiy of invesors holdings in financial and invesmen policies, i differs from Cronqvis and Fahlenbrach (2007) in several imporan ways. Firs, his paper focuses on he relaionship beween invesors preferences and fuure changes in firm policies. Second, he wide availabiliy of insiuional holdings daa (all publicly raded U.S. socks since 1980) allows his paper o examine a considerably broader sample of firms for a longer period of ime han Cronqvis and Fahlenbrach (2007) s blockholder sample for he period. Third, he approach used in his paper allows for an examinaion of (1) he dynamics of he relaionship beween insiuional invesors and firm managers, and (2) wheher he influence of insiuional invesors has a posiive effec on firm valuaion. The remainder of his paper is srucured as follows. Secion I describes he daa and mehodology. Secion II presens and discusses he evidence on he heerogeneiy of insiuional preference for financial policies. Secion III documens he role of insiuional preference on firm financial policy choices. Secion IV documens he poenial moivaions 5

8 for firms o incorporae insiuional preference in heir financial policy choices. Secion V discusses he evidence on insiuional preferences for operaing policies. Secion VI concludes. I. Daa and Definiions A. Sample Consrucion In his sudy, I use hree main daabases: he Thomson Financial 13F Insiuional Holdings daabase (formerly known as he CDA/Specrum s34 daabase), he annual COMPU- STAT daabase and he monhly CRSP sock reurns daabase. I resric he sample o exclude financial firms (SIC codes ) and uiliies (SIC codes ) since hese indusries are highly regulaed. Alhough Thomson (CDA/Specrum) provides quarerly snapshos of insiuional invesors porfolios exraced from 13F repors filed wih he SEC 5, mos of he analyses in his sudy focus only on repors corresponding o he insiuional holdings in he fourh quarer since I use annual COMPUSTAT daa o calculae firm financial and operaing characerisics. The sample is limied o he period since insiuional holdings are only available on he Thomson daabase since Throughou he analysis, I use book leverage as my main measure of leverage. Book leverage is defined as (oal asse minus book equiy) divided by oal asse. Alhough all of he resuls are robus o using marke leverage insead of book leverage, book leverage is used as he main measure o absrac from changes in leverage ha are driven by changes 5 As saed in SEC s websie (hp:// insiuional invesmen managers who exercise invesmen discreion over $100 million or more in mus repor heir holdings on Form 13F wih he SEC [on a quarerly frequency]. [...] Secion 13(f) securiies generally include equiy securiies ha rade on an exchange or are quoed on he Nasdaq Naional Marke, some equiy opions and warrans, shares of closed-end invesmen companies, and some converible deb securiies. 6

9 in he marke valuaion of he firm s equiy, which in urn may be relaed o insiuional rading. B. Indusry- and Characerisics-adjused Leverage The lieraure documens evidence ha firms in he same indusry and hose wih similar characerisics end o have similar leverages. 6 Since insiuional invesors may have heerogeneous preferences for indusry and firm characerisics, I employ he following mehodology o absrac from he effecs of hese preferences on heir leverage preference. Firs, I esimae he prediced leverage for each firm based on is indusry and firm characerisics using a Tobi regression. The residual from his regression is he componen of leverage ha is orhogonal o hese indusry and firm characerisics. In oher words, he residual is he indusry- and characerisics-adjused leverage. Alhough heir preferences for raw leverage may parly reflec heir preferences for hese characerisics, he preference for indusryand characerisics-adjused leverage should no be affeced by hese characerisics. Consequenly, he res of he analyses use indusry- and characerisics-adjused leverage insead of he raw leverage. As suggesed by Hovakimian, Opler, and Timan (2001), a Tobi regression wih he following specificaion is employed in he firs sage: D/A f = α + β X f, (1) 6 Schwarz and Aronson (1967) documen evidence ha firms wihin he same indusry show no significan differences in deb raios. Leverage raios are also relaed o various firm characerisics; for example, leverage is negaively relaed o research and developmen expendiures (Long and Maliz (1985)), growh opporuniies (Smih and Was (1992)), and profiabiliy (Keser (1986)). 7

10 where X f is a vecor of firm characerisics ha have been used in oher capial srucure sudies: oal asses, profiabiliy, inangible asse, collaeral, marke-o-book raio, age, capial expendiure, R&D expendiure, SGA expendiure, and sock reurn for he pas wo years. To conrol for poenial ime-series variaions in he condiion of financial markes and indusries ha may affec firm financial policies, he dependen and independen variables are defined as he differences from he hree-digi SIC indusry means for a given year. 7 [Table I abou here] The esimaed coefficiens from his regression are repored in he firs column of Table I. 8 For he res of his paper, he focus will be on he adjused leverage, D/A f, which is he residual from he Tobi regression. The nex wo subsecions describe how hese adjused leverages are aggregaed o approximae he insiuional preferences a he insiuion level and a he firm level. C. Insiuional Preference Since insiuional invesors preferences for firm financial policies are no direcly observable, I infer hese preferences from heir shareholdings. In paricular, insiuion i s preference for characerisic-adjused leverage, D/A i,, is measured as he dollar-weighed average of he characerisics-adjused leverage raio of each sock in is holdings: D/A i, = f F i, ($ f i, D/A f ) / f F i, $ f i,, 7 The resuls are robus o using oher indusry definiions such as wo-digi SIC or Fama-French 49- indusry classificaion. 8 The coefficiens for marke-relaed variables (M/B and sock reurns) are considerably differen from he ones in Hovakimian, Opler, and Timan (2001) since I use book leverage insead of marke leverage. 8

11 where F i, is he se of all socks held by insiuion i a he end of year and $ f i, is he dollar amoun of insiuion i s fund allocaed o he socks of firm f. D. Firm-Level Insiuional Preference The firm-level insiuional preference is measured as he share-weighed average of he preference of each insiuion holding is socks: D/A Ins,f = i s.. f F i, ($ f i, D/A i, ) / i s.. f F i, $ f i,, where D/A i, is he insiuional preference of insiuion i as defined above. This firm-level aggregae preference will be referred o as insiuional preferred leverage raio for he res of he paper. II. Heerogeneiy in Insiuional Leverage Preference A. Heerogeneiy in Insiuional Holdings To examine wheher here is heerogeneiy in insiuional leverage preference, I sar by esing he null hypohesis ha he average deb raios of all insiuional porfolios are he same. This null hypohesis is srongly rejeced for each year in my sample boh for raw deb raios and when he deb raios are measured relaive o arge deb raios based on indusry and firm characerisics. In order o more direcly measure he degree of heerogeneiy in insiuional preference for leverage 9, I sor insiuions ino annual quiniles based on heir revealed preferences 9 An asue reader may argue ha each invesor can subsiue home-made leverage for firm-made leverage by borrowing/lending on her own. This is no necessarily rue for he insiuional invesors in my sample 9

12 (D/A i, ) and examine how heir shareholdings evolve. The average leverage of socks held by insiuions wihin each quinile is repored in Panel A of Table II. The average adjused leverage of firms held by insiuions in he op leverage preference quinile is 6.09%, which is more han 13 percenage poins higher han he average leverage of firms held by insiuions in he boom quinile. [Table II abou here] Alhough he observed heerogeneiy in he leverage of insiuional holdings is quie significan, his heerogeneiy may no be caused by he acive decisions of hese insiuions. In oher words, if insiuions were assigned socks randomly, some insiuions would end up wih high leverage socks while ohers would end up wih low leverage socks. To examine his possibiliy, I simulae he disribuion of average deb raios of insiuional porfolios by replacing each sock in each insiuion s year- porfolio wih a random sock. These hypoheical porfolios are sored annually ino leverage quiniles based on he average leverage raios of he (random) socks in hese porfolios. The ime-series average of he spread beween he op and boom quiniles is calculaed for each simulaion run. Panel A of Figure 1 repors he cumulaive disribuion of hese simulaed averages from 10,000 simulaion runs. The observed spread (13.16%) is higher han 99% of he simulaed spreads, which suggess ha he he observed heerogeneiy in he leverage of insiuional holdings is no likely o be caused by random sock selecions. [Figure 1 abou here] In order o have a beer undersanding of he exen of insiuional preference for leverage, I examine wheher insiuions whose curren holdings have high (low) leverages are because hey may have resricions on heir invesmen sraegies (e.g., hey may only hold equiy only or a cerain minimum amoun of equiy). In his sense, he degree of heerogeneiy we observe in he daa is relaed o how binding hese resricions are for he insiuions in my sample. 10

13 more likely o hold and purchase socks wih high (low) leverages in he fuure. Panel B of Table II repors he average leverage of socks held in he year following he ranking year (+1) by insiuions wihin each leverage quinile in Panel A. Afer conrolling for indusry and firm characerisics, he socks held by insiuions in he highes leverage preference quinile in he preceding year have on average 8.66% higher leverage han he socks held by insiuions in he lowes leverage preference quinile. This observed spread is higher han 98% of he simulaed spreads repored in Panel B of Figure I. 10 While he persisence in he average leverage of insiuional holdings may be driven by persisen leverage preferences of insiuional invesors, i may also be driven by he persisence in heir holdings: each insiuion may hold a relaively unchanged porfolio for several years. As leverage raios end o be similar from one year o he nex, he average deb raio of each insiuion s porfolio may be similar as well. To conrol for he persisence in holdings, I examine he average leverage of socks iniially purchased (holding iniiaions) by insiuions wihin each preference quinile. Panel C of Table II repors he average leverage of socks iniiaed in he year following he ranking year (+1) by insiuions wihin each year- leverage preference quinile. The socks purchased by insiuions in he highes quinile have on average 3.22% higher leverage han he socks purchased by insiuions in he lowes quinile. This difference is almos 25 percen of he difference beween he average leverage of firms currenly held by insiuions in he op and boom leverage preference quiniles. As his spread should be close o zero if insiuions do no ake leverage ino accoun in heir porfolio selecion decisions, he observed spread is higher han almos all of he simulaed spreads repored in Panel C of Figure I. The resuls so far sugges ha he equiy holdings and purchasing paerns of 10 As insiuional porfolios end o show some persisence, he random replacemen sock in Panel B is held consan if he sock being replaced is held by a paricular insiuion a and

14 insiuional invesors display some persisen heerogeneous preferences for firm leverage. While he discussion so far has focused on he levels of leverage, changes in leverage may also affec insiuional porfolio choices. In paricular, insiuional purchases may be relaed o expeced changes in leverage raios. To proxy for hese (unobserved) expecaions, I examine he relaionship beween insiuional purchases and he recen and fuure changes in leverage raios. In paricular, I ask wheher insiuional invesors ha currenly hold socks wih high leverage are more likely o purchase new socks ha (i) had increased heir leverage in he recen pas or (ii) end o increase heir leverage in he fuure. If insiuional invesors purchase socks based on he paern of pas (expeced) leverage changes, a relaionship beween leverage preferences and recen (fuure) changes in firm leverage should be observed in insiuional purchases. I find ha he paerns of insiuional purchases across insiuions wih differen leverage preferences are no relaed o recen, conemporaneous and fuure changes in firm leverage. Conrary o he observed heerogeneous preference for he level of deb raios, he sock selecions of insiuional invesors do no seem o be correlaed wih he expeced changes in deb raios. I find ha he paerns of recen, conemporaneous and fuure changes in firm leverage are similar for socks purchased by insiuions in he op and boom leverage preference quiniles. B. Insiuion Types As alluded o in he inroducion, insiuional preference for leverage may be a resul of heir beliefs on he economic oulook and he opimal amoun of deb o alleviae agency issues. The exen o which hese beliefs acually affec he equiy holdings and purchases of an insiuional invesor depends on he imporance of hese beliefs relaive o his invesor s 12

15 oher invesmen objecives. For example, Brickley, Lease, and Smih (1997) esablish ha insurance companies and banks may have eiher exising or poenial business relaionships wih firms. These invesors may view mainaining hese relaionships as he main objecive of heir sock invesmens. Consequenly, heir shareholdings may no fully reflec heir preferences for leverage. Addiionally, Bushee (1998) suggess ha insiuional invesors may have differen incenives depending on heir invesmen horizons and invesmen sizes. An invesor ha invess heavily in each of her holdings will only hold socks ha mach her preference; consequenly, her preference will be revealed srongly in her holdings. On he oher hand, he preference of an invesor ha aims for a diversified porfolio may no be fully refleced in his holdings since he may no be able o find enough socks ha mach his preference and has o inves in oher socks. Conrary o he clear predicion for he effec of invesmen size on revealed preferences, he predicion for invesmen horizon is ambiguous. For example, an invesor wih a long invesmen horizon may or may no reveal her preference in her porfolio. As socks in his porfolio end o be held for a relaively long ime, i is reasonable o expec hese socks o mach her preference; however, her curren porfolio may no reflec her acual preference as each long-held sock in her porfolio may have changed is characerisics over ime. Similarly, he porfolio of an invesor wih a shor invesmen horizon may no reveal his preference since socks in his porfolio may no be seleced according o his preference given ha hey are only held for a shor horizon. To examine hese predicions, I divide insiuional invesors ino wo groups based on each of he following hree classificaions. Firs, I use he classificaion of insiuional invesors 13

16 provided by Thomson Financial. Following he lieraure 11, I separae insiuions ino wo groups: Acive insiuions, which include invesmen advisers and invesmen companies, and Passive insiuions, which include bank rus deparmens and insurance companies. The acive insiuions according o his classificaion are no necessarily acivis invesors who ake over he firm as in Behel, Liebeskind, and Opler (1998) or Denis and Serrano (1996). Moreover, hey are also no necessarily acivis invesors in he mold of public pension funds as in Carleon, Nelson, and Weisbach (1998) and Wahal (1996). As repored in columns (1) and (2) of Table II, banks and insurance companies have considerably weaker revealed preferences for leverage in heir holdings and heir purchases han invesmen advisers and companies. 12 I also use wo oher disinc classificaions o capure he invesmen sraegies of insiuional invesors as moivaed by Bushee (1998). Firs, insiuions are classified based on he sabiliy of heir porfolios o capure heir invesmen horizons. For each insiuion, I calculae a Sabiliy variable which is measured as he percenage of equiy holding held coninuously over he previous eigh quarers. Sabiliy = ( ws, I ws,>0%) ws,, where w s, is he insiuional invesor s porfolio weigh in firm s a he end of quarer and I ws,>0% is an indicaor variable ha akes a value of one if sock s has been held coninuously by he insiuional invesor for he previous wo years and zero oherwise. Insiuions wih high Sabiliy end o hold similar socks for a long period, while hose wih low Sabiliy 11 Brickley, Lease, and Smih (1997), Almazan, Harzell, and Sarks (2005), and Chen, Harford, and Li (2006). 12 This evidence is also consisen wih banks and insurance companies having more homogenous beliefs abou he magniude of financial disress coss or he efficacy of deb in reducing agency issues. Addiionally, banks and insurance companies may have relaively larger holdings of corporae bonds which may reduce he incenive o reveal heir preferences for leverage hrough heir shareholdings. 14

17 end o replace socks in heir porfolio frequenly. Second, insiuions are classified based on he concenraion of heir holdings. For each insiuion, I calculae a Block variable which is measured as he percenage of holding ha makes up more han 5% of equiy under he insiuion s managemen. Block = ( ws, I ws,>5%) ws,, where w s, is he insiuion s porfolio weigh in firm s a he end of quarer and I ws, >5% is an indicaor variable ha akes a value of one if w s, is greaer han 5% and zero oherwise. Insiuions wih high Block end o have concenraed holdings, while hose wih low Block end o have diversified porfolio. I divide insiuional invesors ino wo groups based on each of hese classificaions. Insiuions wih Sabiliy greaer han 1/3 are classified as Sable insiuions, while he res of he insiuions are classified as Dynamic insiuions. Similarly, insiuions wih Block greaer han 1/3 are classified as as Concenraed insiuions and he res are classified as Diversified insiuions. 13 Consisen wih he predicion ha insiuions ha inves heavily in each of heir holdings will reveal sronger preference in heir holdings, columns (5)-(6) of Table II repor ha insiuions wih concenraed porfolios end o have considerably sronger revealed preferences for leverage in heir holdings and heir purchases. As repored in columns (3)-(4) of Table II, insiuions wih relaively shor invesmen horizon (low Sabiliy) have considerably sronger revealed preferences for leverage in heir holdings and heir purchases. This suggess ha he flexibiliy hese insiuions have in consrucing heir porfolios resuls in sronger revealed preferences han he inenion of long-horizon 13 The hreshold of 1/3 is arbirarily chosen such ha he number of insiuions in each classificaion is similar. The resuls repored here are robus o changing his hreshold. 15

18 insiuions o hold socks according o heir preference. C. Inensiy and Persisence of Insiuional Preference In order o furher examine how well insiuional invesors holdings represen heir preferences, I explore he relaionship beween he inensiy and he persisence of insiuional invesors revealed preferences. In paricular, I firs consruc a proxy for he inensiy of insiuional invesors revealed preferences, and hen es wheher his proxy is posiively correlaed wih he persisence of hese revealed preferences. I esimae he inensiy of each insiuional invesor s leverage preference by measuring he dispersion in he excess leverage raios of he socks in is porfolio. An insiuional invesor wih a srong preference for a paricular level of leverage will hold socks wih similar leverage raios. On he oher hand, he leverage raios of socks held by an insiuional invesor wih a weak leverage preference will end o look differen from each oher. Afer soring insiuional invesors ino wo annual groups on he leverage dispersion of is porfolio, I sored insiuions wihin each group ino annual quiniles based on he average excess leverage of he socks in heir year s porfolios. The wihin-quinile averages of excess leverage of (1) socks held a he end of year, (2) socks held a he end of year +1, and (3) holdings iniiaed in year +1, +2, +3, and +4, respecively, are repored in Table III. Panel A repors he averages for insiuions wih above-median porfolio leverage dispersion, while Panel B repors he averages for below-median insiuions. [Table III abou here] The descripive saisics of he soring variable are repored in he firs column ( Residual D/A of Holdings a ). Comparison of he numbers in his column for Panel A vs. Panel 16

19 B suggess ha insiuional invesors wih above-median dispersion end o have weaker revealed preferences as he gap beween he op and boom quinile is narrower for hese insiuions relaive o he low-dispersion insiuions. This paern persiss for he holdings of hese insiuions in he following year. More imporanly, his paern persiss for he socks ha hese insiuions iniially purchased in he following years. As repored in Panel C, here are significan differences in he paerns of socks iniially purchased by low-dispersion insiuions vs. high-dispersion insiuions. Low-dispersion insiuions end o show more persisence in heir revealed preferences: he correlaion beween socks iniially purchased by low-dispersion insiuions and he socks in heir porfolio is higher han he respecive measure for high-dispersion insiuions. In summary, he resul in his secion suggess ha here is some heerogeneiy in he revealed preferences of insiuional invesors for leverage. Moreover, his heerogeneiy ends o persis for (a leas) several years, in paricular for insiuional invesors wih sronger revealed preferences for leverage. III. Insiuional Preference and Financial Decisions A. Leverage Raio In his secion, he focus is on he role of insiuional preference in firms decision o change heir leverage. In paricular, I ask wheher (and o wha exen) a firm whose leverage raio is considerably differen from he preference of is insiuional shareholders aemps o reduce he difference. To answer his quesion, I assign each firm ino one of five annual quiniles based on he difference beween is indusry- and characerisics-adjused 17

20 leverage raio ( D/A f ) and he aggregae revealed preferences of is insiuional shareholders (D/A Ins,f ). If insiuional preference affecs firms capial srucure decision, a firm whose leverage raio is higher (lower) han he preference of is insiuional shareholders should reduce (increase) is leverage raio. Panel A of Table IV repors he resuls from his univariae analysis. Firms wih low leverage raios relaive o he aggregae preferences of heir insiuional shareholders (firs line) end o increase heir leverage, and vice versa. The difference in leverage change beween firms in he highes leverage quinile and hose in he boom quinile is 8.84%. Moreover, he effec is quie persisen: he difference in he leverage change in he following year (5.40%) is more han half of he difference in he firs year. [Table IV abou here] In order o conrol for he poenial effecs of oher firm characerisics on he changes in leverage raio, I employ a mulivariae regression wih he following specificaion: ( ) ( ) D/A f +1 D/A f = β D/A Ins,f D/A f + γ X f, (2) where D/A f is firm f s indusry- and characerisics-adjused leverage raio a year, D/A Ins,f is he insiuional preferred leverage raio (i.e., he aggregae revealed preferences of is insiuional shareholders), and X f is a vecor of firm characerisics ha includes financing defici, oal asse, marke-o-book raio, profiabiliy, inangible asse, collaeral, capial expendiure, R&D expendiure, SGA expendiure, and sock reurn for he pas wo years. Year fixed effecs are also included in his regression o allow for poenial ime-series variaions in he condiion of financial markes ha may affec firms financial decisions. 18

21 Panel B of Table IV repors he coefficien esimaes from he mulivariae analysis. The effec of insiuional preference on changes in leverage raios is robus o he inclusion of firm characerisics. In paricular, he poin esimae of β (0.172) in column (1) corresponds o a 4.36% difference in he change of leverage raio for a one sandard deviaion change in he difference beween he insiuional and firm leverage. Column (2) in Panel B repors he esimaes from GMM esimaion following Arellano and Bond (1991). The poin esimae for he effec of insiuional preference (0.112) is weaker han ha obained using fixed-effec regression in column (1), bu i is sill saisically significan a he 1% level. Column (3) in Panel B repors ha he effec of insiuional preference on changes in leverage raios is slighly weaker for S&P 500 firms. This suggess ha managers of S&P 500 firms are less accommodaing o he preferences of heir shareholders relaive o managers of non-s&p 500 firms. This poin will be discussed in more deail below. One may argue ha if he insiuional preferred leverage raio is in he same direcion as he arge leverage (i.e., zero adjused leverage), a leverage change in he direcion of he preferred leverage raio may be driven by he firm managers desire o adjus owards arge leverage. In order o conrol for his possibiliy, column (3) repors he coefficiens for a regression ha includes he firm s curren adjused leverage: ( ) ( ) ( ) D/A f +1 D/A f = β D/A Ins,f D/A f + θ D/A f + γ X f. (3) If firms are adjusing owards heir insiuional preference, he esimae for β should be posiive. If firms are adjusing owards he arge leverage, he esimaed coefficien on heir curren leverage, θ, should be negaive. In order o avoid mulicollineariy, only firms for which he insiuional preferred leverage raio has he same sign as bu is larger in 19

22 magniude han he firm leverage raio (see Figure II) are included in his regression. As only abou 11% of firm-year observaions are in his caegory, he sandard errors in his regression are larger han in he regression in column (1). The coefficien for he difference beween insiuional preferred and firm leverage raios in column (4) is abou half of ha in column (1) bu sill significan. This coefficien has a similar magniude as he coefficien for he curren leverage raio, suggesing ha he effec of insiuional invesors preference is abou as srong as firm managers adherence o arge leverage. The las se of resuls (column (5)) in Panel B suggess ha he effec of insiuional preference on changes in leverage raios is quie persisen. In mulivariae regressions in which he independen variable is he change in leverage raio over he subsequen year, he parameer esimae for insiuional preference is sill srongly significan. The poin esimae of β (0.119) corresponds o a 2.03% difference in he change in leverage raio for a one sandard deviaion change in he difference beween he insiuional and firm leverage. 20

23 B. Insiuion Types As he resuls in Table II sugges ha differen ypes of insiuional invesors may have differen preferences, e.g., acive insiuional invesors (invesmen advisors and managers) end o have sronger revealed preferences han passive insiuional invesors (banks and insurance companies), i is reasonable o expec ha heir preferences may have differen effecs on firm s financial decisions. Since he preferences of differen groups of insiuional invesors in a paricular firm may be similar o each oher, I employ hree mulivariae regressions wih he following specificaions o separae he effecs of he respecive preferences of hese invesors: ( ) ( ) ( ) D/A f +1 D/A f = β A D/A T ypea,f D/A T ypeb,f + β B D/A T ypeb,f D/A f + γ X f, where D/A T ypea,f D/A T ypeb,f is he aggregae preferences of firm f s T ypea insiuional shareholders, is he aggregae preferences of firm f s T ypeb insiuional shareholders, and (T ypea, T ypeb) {(Acive, P assive), (Sable, Dynamic), (Diversif ied, Concenraed)}. The coefficien esimaes from hese regressions are repored in Table V. (4) [Table V abou here] As D/A T ypeb,f appears in boh he firs and second independen variables, is esimaed effec on firm leverage change is ( β B β A ). Consequenly, he difference beween he esimaed effec of T ypea shareholders ( β A ) and ha of T ypeb shareholders is: β A ( β B β A ) = 2 β A β B. For Acive vs. P assive shareholders, he difference is (p = 0.19). While his difference is no saisically significan, he difference beween Sable and Dynamic shareholders 21

24 [0.069 (p = 0.06)] and beween Diversif ied and Concenraed shareholders [0.170 (p < 0.01)] are saisically significan and larger in magniude. While insiuions wih more flexibiliy in heir holdings (concenraed and dynamic insiuions) have considerably sronger revealed preferences for leverage, he resuls in Table V sugges ha less-flexible insiuions (i.e., sable and diversified insiuions) end o have sronger impacs on firm s financial policy. This seems o sugges ha he laer insiuions ry o make up for heir inabiliy o selec socks based on heir preferences and fully reveal heir preferences in heir shareholdings by exering heir preferences more forcefully on he firms in heir porfolios. C. Wihin-Firm Heerogeneiy in Insiuional Preferences In his secion, I explore he hypohesis ha firm managers will be more likely o accommodae he preferences of heir shareholders if hese shareholders have similar preferences, and less likely o do so if he preferences are heerogeneous. The heerogeneiy in he insiuional shareholders preferences of each firm is proxied using hree variables: (1) he iner-quarile range of he preferences of insiuional shareholders, (2) he sandard deviaion of hese preferences, and (3) he number of insiuional shareholders. A he end of year, each firm is assigned ino quiniles based on each of hese hree variables. Panel A of Table VI repors he pair-wise quinile correlaions among hese variables. [Table VI abou here] Panel B repors he coefficiens from he regression of he change in excess leverage on various firm characerisics and year fixed effecs. The dependen variable is he change in excess leverage in year +1. In addiion o he explanaory variables included in Table IV, 22

25 wo indicaor variables are included in each of hese regressions: Low Heerogeneiy which akes he value of 1 if he heerogeneiy proxy is in he lowes quinile and zero oherwise, and Medium Heerogeneiy which akes he value of 1 if he heerogeneiy proxy is in he middle hree quiniles and zero oherwise. The main variables of ineres are he ineracions beween hese indicaor variables and he gap beween he firm s insiuional preferred leverage and is excess leverage raios. More precisely, I run he following regressions: ( ) D/A f +1 D/A f = β ( ) D/A Ins,f D/A f + β L [I LowHeerogeneiy + β M [I MediumHeerogeneiy ( D/A Ins,f )] D/A f ( D/A Ins,f D/A f )] + α L I LowHeerogeneiy + α M I MediumHeerogeneiy + + γ X f. (5) Consisen wih he hypohesis ha firms whose shareholders have similar preferences are more likely o accommodae he preferences of heir shareholders relaive o firms whose shareholders have heerogenous preferences, firms wih low values of invesor heerogeneiy proxies exhibi sronger correlaions beween aggregae shareholder preferences and financial decisions. The poin esimaes of β LH sugges ha he effec of insiuional preferences on corporae policies is abou 15% higher for firms wih low values of heerogeneiy proxies relaive o he corresponding effec for firms wih high values of heerogeneiy proxies. While he coefficien esimaes of β LH are saisically significan, he economic magniude seems o sugges ha he effec of insiuional invesors preference on corporae policies is achieved mosly hrough he aggregae preference as he effec of insiuional invesors preference on corporae policies is sill relaively srong (abou 15%) for firms wih high values 23

26 of heerogeneiy proxies. This seems o sugges ha he effec of insiuional invesors on corporae policies comes hrough rading as opposed o voing or direc influencing. D. Firm Characerisics Column (2) in Panel B of Table IV repors ha he relaionship beween insiuional preference and fuure changes in leverage raios is slighly weaker for S&P 500 firms. As S&P 500 firms ends o have larger (poenial) invesor bases han oher firms, heir managers may be less likely o accommodae he preferences of heir curren shareholders. This observaion moivaes he following hypohesis: firm managers will be less likely o accommodae he preferences of heir shareholders when hey have lower incenives o do so. To formally es his hypohesis, I divide firms ino five quiniles each year based on size (i.e., oal asse) and age (i.e., number of years since firs appearance in COMPUSTAT). Firms in he op size quinile are caegorized as Large firms, while hose in he boom quinile are caegorized as Small. Similarly, firms in he op age quinile are caegorized as Old, while hose in he oher exreme quinile are caegorize as Young. 14 The indicaor variables for ( ) hese caegories are hen ineraced wih D/A Ins,f D/A f o idenify he variaion in insiuional effecs on firms from differen size/age groups. In paricular, I run he following 14 This hreshold (i.e., quiniles) is admiedly arbirary. However, he resuls are similar if hirds, quariles, or deciles are used insead. 24

27 regressions: 15 ( ) D/A f +1 D/A f ( ) D/A f +1 D/A f ( ) [ ( )] = β D/A Ins,f D/A f + β Large I Large D/A Ins,f D/A f ( )] +β Small [I Small D/A Ins,f D/A f + γ X f, (6) ( ) [ ( )] = β D/A Ins,f D/A f + β Old I Old D/A Ins,f D/A f ( )] +β Y oung [I Y oung D/A Ins,f D/A f + γ X f. (7) I hypohesize ha managers of older and larger firms are less likely o accommodae he preferences of heir curren shareholders since hese shareholders are more easily replaceable due o he larger poenial invesor bases of hese firms. [Table VII abou here] Consisen wih his hypohesis, columns (1) and (2) of Table VII repor ha he coefficiens on he ineracion erms wih Large (β Large ) and Old (β Old ) are significanly posiive. On he oher hand, he coefficiens on he ineracion erms wih Small (β Small ) and Young (β Y oung ) are negaive. The poin esimaes of β Large and β Small sugges ha he effec of insiuional preferences of corporae deb raio is reduced by 18% for large firms and increased by 48% for small firms (relaive o firms in he middle hree size quiniles). Similarly, he effec of insiuional preferences is reduced by 31% for old firms and increased by 30% for young firms (relaive o firms in he middle hree age quiniles). The previous es is an indirec es of how he relaive powers of insiuional invesors and firm managers affec heir ineracion. To more direcly es he effec of relaive powers 15 The indicaor variables are no included by hemselves in he regressions since X f includes he coninuous versions of hese variables: log(size) and log(age). The resuls are similar wheher hese indicaor variables are included or no. Similar resuls also obain if he coninuous variables are used insead of he indicaor variables in he ineracions. 25

28 on he shareholder-manager relaionship, I idenify CEOs wih weaker career concerns by consrucing he following proxy using daa from ExecuComp: 16 1, if he firm s CEO is over 60 years old and also holds he chairman posiion I Powerful CEO = 0, oherwise In order o es wheher hese Powerful CEOs are less likely o accommodae he preferences of heir shareholders due o heir weaker career concerns, I employ he following regression: ( ) D/A f +1 D/A f = β ( ) D/A Ins,f D/A f +β P owerfulceo [I P owerfulceo ( )] D/A Ins,f D/A f +γ P owerfulceo I P owerfulceo + γ X f. (8) As repored in column (3) of Table VII, he presence of a CEO wih weaker career concerns reduces he effec of insiuional preference on fuure leverage decisions. Overall, he resuls in his secion sugges ha insiuional invesors play a less significan role in corporae decisions when heir posiions are relaively weaker. E. Deb-Equiy Choice Alhough he resuls in previous secions sugges ha firms end o reduce (increase) heir leverage raios when hese raios are higher (lower) han he aggregae preferences of heir insiuional shareholders, hese resuls do no indicae wheher insiuional preferences affec firms financing choices when hey raise new funds or reire/repurchase exising capial. If insiuional preferences affec securiy choices, firms needing new capial should be more 16 Since ExecuComp only racks S&P 1500 firms from 1992, his sample is much smaller han he main sample. More imporanly, he ExecuComp sample consiss of relaively older and larger firms han he main sample. 26

29 (less) likely o issue deb han o issue equiy when heir leverage raios are lower (higher) han he aggregae preferences of heir insiuional shareholders, and vice versa. Similarly, firms wih excess capial should be more (less) likely o repurchase shares han o repay deb when heir leverage raios are lower (higher) han he aggregae preferences of heir insiuional shareholders. In order o examine he exen o which insiuional preferences affec securiy choices, I focus on firms ha issue/repurchase deb or equiy. Firms are defined as issuing (repurchasing) a securiy when he ne amoun issued (repurchased) exceeds 5% of he book value of asses. Each issuer/repurchaser is assigned ino one of hree groups (op, middle hree, and boom quiniles) based on he difference beween is characerisics-adjused leverage raio ( D/A f ) and he insiuional preferred leverage raio (D/AIns,f ) wihin each year. The firs wo columns of Panel A in Table VIII repor he probabiliy ha a firm in each group chooses o ake a leverage-increasing acion (issue deb/repurchase shares) insead of a leverage-reducing acion (issue equiy/repay deb) when i raises new funds or reduces is capial. Panel B repors he marginal effecs from a mulivariae probi regression ha includes arge leverage (from Tobi regression in Table I) and oher firm characerisics ha are relaed o he deb-equiy decision. [Table VIII abou here] The resuls in Panel A sugges ha firms are more likely o ake leverage-increasing acions (issue deb or repurchase shares) han o ake leverage-decreasing acions (issue equiy or repay deb) when heir leverage raios are below he aggregae preferences of heir insiuional shareholders, and vice versa. These insiuional effecs are weaker when firms need capial (col. 1) han when hey disribue exraneous capial (col. 2). These resuls are robus o he inclusion of firm characerisics ha may affec he firm s financing decisions 27

30 in Panel B. Moivaed by he above resuls, I also examine how insiuional preference affec firms deb-equiy choice when hey do no raise new capial. In paricular, I examine he probabiliy ha firms recapialize by increasing leverage (issuing deb o repurchase shares) insead of decreasing leverage (issuing equiy o repay deb). If insiuional preferences affec his choice, firms should be more (less) likely o underake leverage-increasing recapializaion when heir leverage raios are lower (higher) han he aggregae preferences of heir insiuional shareholders, and vice versa. The las column in Panel A repors he probabiliy ha a firm in each group chooses o ake leverage-increasing acions (issue deb and repurchase shares) insead of leverage-reducing acions (issue equiy and repay deb) when i rearranges is capial srucure. The resuls here sugges ha firms are more likely o underake leverage-increasing han leverage-decreasing recapializaion when heir leverage raios are low relaive o he aggregae preferences of heir insiuional shareholders. This resul is robus o he inclusion of firm characerisics in Panel B. Overall, he resuls in Table VIII sugges ha insiuional preferences play an imporan role in he deb-equiy decisions. However, hese preferences end o play a more significan role when firms are no in need of capial. This resul may seem counerinuiive if one expecs he exising shareholders o be he poenial source for he new capial. Consequenly, his resul seems o sugges ha firm managers ignore curren shareholders preferences because hey need o arac addiional and differen invesors when hey issue equiy. 28