Market Timing and REIT Capital Structure Changes

Transcription

1 IRES IRES Working Paper Series Marke Timing and REIT Capial Srucure Changes Ying LI Universiy of Wisconsin Muhammad Faishal bin IBRAHIM Deparmen of Real Esae Naional Universiy of Singapore Seow Eng ONG Deparmen of Real Esae Naional Universiy of Singapore Joseph T. L. OOI Deparmen of Real Esae Naional Universiy of Singapore

2 Marke Timing and REIT Capial Srucure Changes Ying LI Universiy of Wisconsin Muhammad Faishal bin IBRAHIM Naional Universiy of Singapore Seow Eng ONG * Naional Universiy of Singapore Joseph T. L. OOI Naional Universiy of Singapore November 5, 2007 Key words REITs, capial srucure, marke iming *Corresponding auhor. Deparmen of Real Esae, Naional Universiy of Singapore, 4 Archiecure Drive, Singapore seong@nus.edu.sg We hank wo anonymous referees for heir excellen suggesions.

3 Marke Timing and REIT Capial Srucure Changes Absrac This paper seeks o deermine if REIT equiy and deb offerings depend on capial marke condiions. We examine he fundamenal mechanism behind he marke iming behavior. Two differen hypoheses of equiy marke iming he informaion-based marke iming and marke inefficiency-based marke iming are esed. We furher es for backward and forward looking behavior marke iming behavior in deb offering. In addiion, we examine he securiy choice of REITs o evaluae he marke iming heory. The empirical resuls from 1993 hrough 2004 suppor he informaion-based marke iming heory for equiy offerings and backward looking marke iming in deb issuances. While he evidence from he securiy choice analysis largely favors he marke iming heory, some suppor is also esablished for he rade-off and pecking order heories. We conclude ha here is no single heory ha fully explains he capial srucure of REITs. 2

4 1. Inroducion Recen evidence in capial srucure shows ha he iming of financing decisions depends on capial marke condiions (Graham and Harvey, 2001; Baker and Wurgler, 2002). Real Esae Invesmen Truss (REITs) are ypically excluded in sudies for general socks as hey are special invesmen vehicles. However, i is precisely because REITs are special vehicles ha make hem an ineresing es bed for alernaive capial srucure explanaions. This is so for wo reasons. Firs, because REITs face higher dividend payou requiremens, hey have o look o exernal funding for capial invesmen and asse acquisiions and capial srucure adjusmens (Brown and Riddiough, ). Second, he higher angibiliy of REIT asses and beer predicabiliy of cash flows imply ha less asymmeric informaion is expeced from REITs compared o general socks. Given ha asymmeric informaion is fundamenal o capial srucure heories, evidence from he REIT indusry would provide even sronger suppor for informaion-based heories. By sudying marke iming in REIT equiy and public deb issuances, his paper seeks o make hree conribuions o he exan lieraure. Firs, we examine he shor-run and longrun performance of REITs afer equiy issues o es wo marke iming hypoheses informaion-based marke iming heory (Korajczyk, Lucas and McDonald, 1991) and marke inefficiency based marke iming heory (Loughran and Rier, 1995). The asymmeric informaion framework posulaes ha beer informed managers choose o issue equiy when he sock prices are overvalued and repurchase when sock prices are undervalued. If managers could ime he marke when hey issue equiy, he sock prices should be increasing before equiy offerings. Sock prices are expeced o decline afer he announcemen or issues of he new shares if marke iming is done well (Korajczyk, Lucas and McDonald, 1991; 1992) 2. Anoher varian of he informaion-based hypohesis is ha managers choose o issue equiy when he marke is bes informed of he qualiy of he firm so as o reduce he adverse selecion problem. The marke inefficiency-based marke iming heory is suppored by he long-run reurn anomaly afer equiy issues as documened by Rier (1991) and Loughran and Rier (1995). The long-run performance of REIT seasoned equiy offerings (SEOs) has been examined by 1 Brown and Riddiough (2003) analyze he public financial offerings of equiy REITs and find ha equiy offerings are more likely o be used for invesmen, while deb offerings are normally used for adjusing he capial srucure. 2 A negaive price reacion on equiy issue is also consisen wih he pecking order hypohesis. Differeniaion beween he marke iming and pecking order heories hinges on marke reacion o deb issuance, and his is examined in a subsequen secion. 3

5 Howon, Howon and Friday (2000). They find long-run underperformance in REIT SEOs when comparing he holding period reurn of equiy issuing REITs wih non-equiy offering REITs. As mos of he REITs (91 ou of 123) in our sample have a leas one equiy offering during he sudy period, we op o use he Fama and French (1993) and Cahar (1997) four facor model o calculae long-erm reurns. This approach is preferred over he Howon e al. s approach in ha we conrol for relevan risk facors. Our second conribuion is o examine he facors influencing he choice of deb offering by REITs. In paricular, we appeal o he idea ha he curren level of ineres raes in relaion o he pas hisory of ineres raes maers in deb marke iming, and ha managers are backward-looking. We also es he forward-looking deb marke iming hypohesis by examining wheher ineres raes rise afer REITs issue deb, alhough he power of his es is limied given he falling ineres rae environmen over he sudy period from 1993 hrough The hird conribuion is o es for marke iming in REIT equiy and deb issuance agains compeing capial srucure heories by examining he securiy choice when financing decisions are made. The rade-off heory implies ha all capial financing decisions are made o adjus he leverage raio o an opimal raio, and he opimal leverage raio is correlaed wih bankrupcy coss. The pecking-order heory suggess ha should firms need exernal capial, hey would use deb firs. The marke iming heory focuses on he capial marke condiions and implies ha firms would choose a paricular securiy when i is relaively cheap. As moivaed earlier, REITs offer a good es ground for securiy choice because REITs are subjec o high dividend payou requiremens, and limied reained earnings availabiliy. Such resricions effecively limi he securiy choice of REITs o issue eiher equiy or deb. As such, we differeniae beween he differen capial srucure heories by examining he securiy choice of REITs beween equiy and deb. This allows us o conduc a cleaner es of differen capial srucure heories. Our resuls suppor he asymmeric informaion marke iming hypohesis, raher han he marke inefficiency-based marke iming. Our resuls have wo imporan implicaions. Firs, alhough REITs are relaively more ransparen han indusrial firms, REITs managers sill ime equiy issuances. The implicaion is ha asymmeric informaion and adverse selecion sill exiss, bu REITs exhibi less obvious iming paerns. Second, he absence of long-run reurn anomaly afer REIT SEOs means ha marke efficiency is no violaed. This is differen from general socks and an earlier sudy by Howon, e al. (2000), in which he 4

6 marke inefficiency-based marke iming is suppored raher han he informaion-based marke iming. We also demonsrae ha REITs end o issue deb when he curren ineres rae level is low relaive o is pas hisory. In addiion, alhough marke condiions are imporan in he securiy choice of REITs, hey are no he only facors ha deermine capial srucure changes. Firm characerisics, such as size, leverage, deb capaciy, ec, have significan effecs on REIT financing aciviies. The rade-off heory and he pecking-order heory also find parial suppor from our analysis. The nex secion reviews he heories on capial srucure and provides a brief lieraure review. Secion 3 describes he daa for his paper, and he following secions conain he empirical resuls. 2. Capial Srucure Theories and Empirical Evidence Trade-Off heory The rade-off heory is based on he capial srucure irrelevance concep firs developed by Modigliani and Miller (1958). In he real world of axes, bankrup coss and leverage, however, he opimal capial srucure is deermined by a rade-off beween he ax benefis of deb and bankrupcy risk. Empirical evidence ha suppors he rade-off heory and he opimal capial srucure is found in Schwarz and Aronson (1967), Bradley, Jarrell and Kim (1984), among ohers. Pecking Order Theory Myers and Majluf (1984) developed he pecking-order heory where informaion asymmery beween firm managers and invesors is he primary deerminan of capial srucure changes. Managers ac in he ineres of exising shareholders, and choose no o issue equiy when firm prospecs are good. Invesors, who have a sparser informaion se han firm managers, will regard equiy issues as negaive signals. Consequenly, managers avoid issuing equiy if hey have inernal capial or hey could issue deb. The pecking order for managers is firs inernal capial, hen deb, and equiy as he las resor. Howe and Shilling (1988) examine he sock price reacions o he announcemen of a relaively small sample of new securiy offerings by REITs from , and find a posiive sock price reacion o deb offerings, and a negaive sock price reacion o equiy offerings. Similarly, Francis, Lys and Vincen (2004) documen negaive reacion o REITs 5

7 equiy offerings, alhough he reacion is more favorable han ha for indusrial firms. This reacion is aribued o asymmeric informaion. Marke Timing Theory The marke iming heory posulaes ha managers make financing decisions according o capial marke condiions (Graham and Harvey, 2001). Managers will issue equiy when heir socks are overvalued and issue deb when ineres raes are low. The long-run relaive underperformance of socks afer iniial public offerings or secondary equiy offerings (Rier 1991; Loughran and Rier 1995) is consisen wih he marke iming heory. The marke iming heory depars from radiional capial srucure heories no only because i examines capial srucure changes from he capial marke perspecive, bu i also relaxes he assumpion of marke efficiency. Equiy (deb) marke iming means ha decisions o issue equiy (deb) depend on sock prices (ineres rae levels). The concep of an opimal leverage raio is relegaed o a secondary role in he marke iming heory. In oher words, he securiy choice of a firm depends on marke condiions raher han some pecking order. Research o undersand he underlying reason for equiy and deb marke iming is somewha inconclusive. Equiy marke iming gains suppor from wo differen hypoheses: informaion-based marke iming (Lucas and McDonald, 1990; Korajczyk, Lucas and McDonald, 1991; Korajczyk, Lucas and McDonald, 1992) and marke-efficiency based marke iming (Rier, 1991; Loughran and Rier, 1995). The informaion-based marke iming hypohesis focuses on he variaions in adverse selecion coss of equiy offering arising from asymmeric informaion. Since managers issue equiy only when sock prices are high or overvalued, equiy offerings should precipiae a fall in sock prices. In Korajczyk, Lucas and McDonald (1991; 1992), asymmeric informaion is no fixed over ime, and firms end o issue equiy when he marke is mos informed abou he qualiy of he firm, for example, afer earnings releases. The marke-efficiency based equiy marke iming hypohesis is predicaed on posoffering long-run sock performance. Underperformance afer IPO (Rier, 1991) and longrun underperformance afer equiy issues (Loughran and Rier, 1995) sugges ha managers 6

8 capialize on he inefficiency in capial markes by iming he marke. 3 The evidence on marke inefficiency-relaed marke iming is mixed (Fama and French, 1988; Kohari and Shanken, 1997). Furher, here are mehodological concerns in he es for long-run anomaly (Fama, 1998; Eckbo, Masulis, and Norli, 2000; Brav, Geczy, and Gompers, 2000). Baker and Wurgler (2002) show empirically ha marke iming has persisen effecs on capial srucure, and ha he capial srucure is he cumulaive oucome of pas aemps o ime he equiy marke. In paricular, hey find ha leverage has a srong negaive correlaion wih marke-o-book raio, which is he usual measure for marke iming opporuniies. Since he rade-off heory and pecking-order heory could no explain his long-run relaionship, marke iming appears o be he only plausible explanaion. However, he underlying reason for his resul is no clear in Baker and Wurgler (2002). In fac, boh equiy marke iming hypoheses could explain he negaive long-run correlaion beween leverage and marke-o-book raio. Under he informaion-based marke iming, he marke-o-book raio is inversely relaed o adverse selecion, and emporary flucuaions in he marke-o-book raio measure variaions in adverse selecion. The markeinefficiency iming heory assumes ha managers ime he equiy marke when hey perceive ha invesors overvalue he firm. They would use he marke-o-book raio as a proxy of marke misevaluaions. Since marke-o-book raio could represen boh adverse selecion and marke misevaluaions, i is difficul o differeniae which marke iming hypohesis dominaes. Empirical work on deb marke iming is provided by Guedes and Opler (1996) where hey show ha he mauriy of deb issues is negaively relaed o he erm spread. Baker, Greenwood and Wurgler (2003) also suppor marke iming in deb issuances in ha firms end o issue long-erm deb when inflaion is low and erm spread is narrow. On he oher hand, firms issue shor-erm deb when inflaion is high and erm spread is wide. Barry, e al. (2005) invesigae he deb issuing decision of firms in relaion o hisorical ineres raes, and conclude ha he amoun and number of deb issues are higher when curren ineres rae is lower compared o hisorical ineres raes. REIT Capial Srucure 3 Apar from marke iming behavior before equiy issues, oher corporae decisions such as spin-offs, deb offerings and sock splis, also exhibi long-run abnormal reurns (Desai and Jain, 1999; Spiess and Affleck- Graves, 1999; and Ikenberry, Rankine and Sice, 1996). 7

9 Because of he special characerisics of REITs, i is difficul o make REITs fi ino some capial srucure heories. For example, as REITs do no need o pay ax, issuing deb is no as appealing o REITs as prediced by he rade-off heory. A he same ime, REITs managers have much less discreion because of he higher informaion disclosure requiremen, and his may reduce he power of he asymmeric informaion and signaling in explaining he capial srucures of REITs. On he oher hand, he high dividend payou requiremen on REITs may lead o he fac ha REITs managers need o seek for exernal finance. As a consequence of he higher reliabiliy on exernal capial, REITs managers may have o lower he cos of capial and do some marke iming. Previous sudies on REIT capial srucure changes include Brown and Riddiough (2003). They find ha equiy offerings are more likely o be used for invesmen, while deb offerings are normally used for adjusing capial srucure. While hey esablish ha a firm s asse, liabiliy and income srucure is relaed o he REIT s decision o issue public deb or equiy, marke iming heory is no explicily esed. Feng, Ghosh and Sirmans (2007) adop he Baker and Wurgler (2002) mehodology o examine he relaionship beween marke-o-book and leverage raios of REITs. They find ha REITs wih high marke-o-book raios have high leverage raios and hisorical markeo-book has long-erm persisen impac on curren leverage raio. The pecking order heory is believed o be more relevan in explaining REITs capial srucure heories in ha sudy. Their sudy suffers from he same issues as Baker and Wurgler (2002), in ha he wo versions of marke iming could no be differeniaed. 3. Daa Sources This sudy covers all equiy REITs repored by he Naional Associaion of Real Esae Invesmen Trus (NAREIT). Quarerly REITs cash flow daa comes from Compusa daabase and daily price daa are exraced from CRSP. We exrac he daes of firs public announcemen of quarerly earnings from he Compusa daabase and all REIT equiy and deb issues 4 (excluding IPOs) from he Securiies Daabase Corporaion (SDC) daabase. Long-erm ineres rae is obained from he U.S. Federal Reserve Board and equiy REITs price index from NAREIT. Afer screening for coninuous rading and daa availabiliy on Compusa and CRSP daabases, 123 equiy REITs are capured in his sudy. The sudy period is from January 4 We focus on public deb issuance as hese are observable and arguably more marke-oriened. 8

10 1993 o March 2004, a oal of 135 monhs. The final sample conains 400 SEOs from 90 REITs. The final sample of REITs deb offerings includes 318 deb issues made by 55 equiy REITs from January 1993 o March 2004 (see Table 1 for deails) 5. We employ boh ime series and cross secional regressions in our analysis. The ime series variables (denoed wih a subscrip) capured in our analysis are: NAREIT : NAREIT equiy REIT index reurn in quarer INT : Long-erm (10 year) ineres rae in quarer S & P : S&P 500 composie index reurn in quarer TS : Term-spread beween long-erm ineres rae and shor-erm ineres rae (3-monh T-bill rae) in quarer Cross secional variables also include EQUITY, DEBT, TS, and INT (wihou he subscrip and defined as of he end he previous quarer before he equiy/deb offering), and he following: LEV: Leverage raio of he REIT a he end of he previous quarer ASSET: Size of he REIT a he end of he previous quarer (log value of oal asses). MB : Marke-o-book raio = marke value of asses/ oal asses = (marke value of equiy + long-erm deb + preferred sock + deb in curren liabiliies)/oal asses DC: Deb capaciy dummy variable ha shows wheher he deb capaciy is reached when he new equiy/deb issue is made. Here he proxy for deb capaciy is he exisence of curren raed deb ousanding. If here is raed deb ousanding, hen i is believed ha firms have he abiliy o borrow from he public deb marke, and have smaller probabiliy o have reached he deb capaciy compared o firms who mainly borrow from banks or oher financial inermediaries (Lemmon and Zender, 2004). STK_AP: REIT sock price appreciaion in he previous quarer SVAR: Variance of he sock reurns in he pas year. The saisical summary of seleced cross secional variables is lised in he following ables. The vas majoriy of he deb issues by equiy REITs are fixed-rae issues. 5 The final sample considers muliple deb issues made by a REIT on he same day as a single issue. 9

11 Year Table 1: Equiy and deb offerings by year Number of Equiy Offerings Amoun of Equiy Offered ($ mil) Number of Deb Offerings Amoun of Deb Offered ($ mil) Toal REITs capured in he year

12 Table 2: Saisical Summary AVMB Sock NAREIT S&P INT (%) TS (%) LEV Reurn (per quarer) reurn (per quarer) reurn (per quarer) Mean Sandard Variaion AVMB : Average of he M/B raio of all he equiy REITs in quarer ; NAREIT : NAREIT equiy REIT index reurn in quarer ; INT : Long-erm (10 year) ineres rae in quarer ; S&P : S&P 500 composie index reurn in quarer ; TS : Term-spread beween long-erm ineres rae and shor-erm ineres rae (3-monh T-bill rae) in quarer ; LEV : Leverage raio of he REIT a he end of he previous quarer ; Table 3: Correlaion Marix beween Time Series Variables AVMB NAREIT S&P INT TS AVMB NAREIT S&P INT TS 1 11

13 Table 4: Correlaion Marix Beween Cross Secional Variables LEV ASSET DEBT DC STK_AP SVAR LEV ASSET DEBT DC STK_AP SVAR 1 LEV: Leverage raio of REIT i; ASSET: log value of oal asses of REIT i; NAREIT: NAREIT equiy REIT index reurn a end of previous quarer; INT: Long-erm (10 year) ineres rae a end of previous quarer; S&P: S&P 500 composie index reurn a end of previous quarer; TS: Term-spread beween long-erm ineres rae and shor-erm ineres rae (3-monh T-bill rae) a end of previous quarer; DC: Deb capaciy dummy variable ha shows wheher he deb capaciy is reached when he new equiy/deb issue is made; STK_AP: REIT sock price appreciaion in he previous quarer; SVAR: Variance of he sock reurns in he pas year. 4. Hypoheses from Informaion-based Marke Timing In his secion, wo sub-hypoheses under he informaion-based marke iming heory are esed. The firs sub-hypohesis, cenral o he informaion-based marke iming heory, is ha sock prices are expeced o fall on he equiy issuance announcemen since equiy offerings are regarded as negaive signals (Asquih and Mullins, 1986; Masulis and Korwar 1986; Barclay and Lizenberger, 1988). The second sub-hypohesis predics ha equiy issues end o cluser afer earnings are announced o miigae negaive sock price reacion. Sub-hypohesis 1: Sock prices fall on equiy issue announcemen We calculae he gross and abnormal sock reurns for he announcemen day. The abnormal reurn is he gross sock reurn minus he CRSP equally weighed reurn, excluding dividend. Table 5 Panel A shows ha here is significan negaive reurn when equiy issuance is announced. This resul is consisen wih he asymmeric informaion beween REITs invesors and managers explanaion for negaive sock reacions o equiy offerings. In conras o he significan negaive marke reacions o equiy issues, he sock price reacions o deb issues are insignifican (Table 5 Panel B). We do no find significan posiive reacions as documened by Howe and Shilling (1988) afer Such changes may come from he srucural change of REITs during ha period.

14 Table 5: Even day reurns Panel A: Equiy offering announcemens Average T-saisic N Gross reurn on he announcemen day *** Abnormal reurn on he announcemen day *** Panel B: Deb issue announcemens Average T-saisic N Gross reurn on he announcemen day Abnormal reurn on he announcemen day Wha is clear is ha invesors consider REIT equiy offerings o be negaive news. In conras, deb issues do no have significan marke impac. For REITs invesors, he deb issues of REITs may simply indicae ha firms are aking advanages of he low ineres rae level, and he deb issues are no correlaed wih invesmen opporuniies. The preceding evidence suppors he informaion-based marke iming for equiy REITs 6. For compleeness, we also compare he sock price reacions o equiy issues of REITs wih common socks. Daa of secondary equiy offerings of general socks ha have similar marke value and offering size over he sudy period are colleced. The common sock sample comprises 943 observaions. Boh gross reurn and abnormal reurn on he announcemen day are calculaed. The resul is shown in Table 6. Table 6: Even day reurn on announcemen of common sock equiy issues Average T-saisic N Gross reurn on he announcemen day *** Abnormal reurn on he announcemen day *** Our resul is robus o issuing aciviies wih differen purposes. Wheher he issuing proceeds are used for invesmen or no do no have effec on he marke reacions o boh he equiy and deb issues. 13

15 The negaive reurns on equiy issuances by common socks are similar o ha for REITs. To see wheher REITs and common socks have he same sock price reacion behavior o equiy issue announcemen, a simple -es is carried ou. We find ha alhough REITs and common sock reurns have difference variances, he differences in means are no saisically significan. Table 7: Differences in Sock Reacion for Equiy Offering by REITs and Common Socks - -es (unequal variances) Gross Reurn Abnormal Reurn T Sa P(T<=) one-ail Criical one-ail P(T<=) wo-ail Criical wo-ail Sub-hypohesis 2: Equiy issues cluser afer earnings releases This sub-hypohesis is based on he ime-varying asymmeric informaion in equiy issues. As menioned earlier, informaion-based marke iming predics ha managers issue equiy when heir socks are overvalued. Managers realize also ha negaive marke reacions o equiy offerings reduce he wealh of exising shareholders. Assuming ha asymmeric informaion beween invesors and managers is ime-varying, managers should issue equiy when he marke is mos informed of he qualiy of he firm and when he cos of adverse selecion is a is lowes. Quarerly earnings announcemens provide imporan informaion, and we expec clusering in equiy issues shorly afer earnings releases (Korajczyk, Lucas and McDonald, 1991). The lag beween equiy issues and he laes quarerly earnings announcemen is calculaed for each seasoned equiy issue (Figure 1). We find ha abou 50% of equiy issues occur wihin 40 days afer he quarerly earnings announcemen. The majoriy of issues occur no laer han 55 days afer he laes quarerly earnings announcemen. Saisical es also shows ha mean of he lag beween equiy issues and he laes quarerly earnings announcemen is less han 45 days. This is basically consisen wih he findings in Korajczyk, Lucas and McDonald (1991). As moivaed before, i would also be ineresing if we could compare REITs wih general socks in esing sub-hypohesis 2. We calculae he ime lags beween equiy issues 14

16 and he laes quarerly earnings announcemen for general socks ha have similar marke value and issuing size (Figure 2). The resul in Figure 2 is essenially an updae of he earlier in Korajczyk, Lucas and McDonald (1991) sudy. The wo disribuions in he ime lags are largely similar. Figure 1: Disribuion of ime lags beween equiy issues and he laes quarerly earnings announcemen for REITs 12% 10% Percenage in all equiy issues 8% 6% 4% 2% 0% 0~6 7~13 14~20 21~27 28~34 35~41 42~48 49~55 56~62 63~69 70~76 77~83 84~90 >90 Days afer quarerly earnings announced Figure 2: Disribuion of ime lags beween equiy issues and he laes quarerly earnings announcemen for common socks Disirbuion of ime lags for common socks 12% 10% 8% Percenage 6% 4% 2% 0% 0~6 7~13 14~20 21~27 28~34 35~41 42~48 49~55 56~62 63~69 70~76 77~83 84~90 >90 Days afer quarerly earnings announced 15

17 One quesion arising from he above analysis is: does he informaion conained in he earnings announcemen affec he ime lag? I is conceivable ha managers issue equiy shorly afer a good earnings announcemen repor and avoid doing so when earnings are poor. To find wheher his is rue, we assume ha he marke is efficien and he marke have fully reaced o he negaive informaion on he earning announcemen day, and divide he whole sample ino hree subgroups by he price reacion o earnings announcemen: significanly posiive, significanly negaive and insignifican reacions. We hen analyze he ime lag beween earnings announcemen and equiy issuing announcemen for earnings announcemens wih significanly posiive reurns and negaive reurns respecively. If he equiy issues follow posiive earnings announcemen more closely han for negaive earnings announcemen, hen he informaion in he earnings announcemen would be an imporan deerminan of ime lag. Wha we find (resul is no shown here bu available on reques) is ha he disribuion of ime lag beween posiive and negaive news is no subsanially differen. In oher words, he informaion conained in he earnings announcemen does no influence he ime lag beween earnings announcemen and equiy offering. A corollary o he ime-varying asymmeric informaion hypohesis is ha he closer he issue announcemen follows an earning release, he more informed is he marke, and consequenly he price drop a he issue day is expeced o be smaller. In Korajczyk, Lucas and McDonald (1991), his hypohesis is suppored using general socks daa. We es his using REIT daa. A simple regression is run o es his relaionship where he announcemen day reurn is regressed on he days beween earnings release and equiy issue announcemen dae and a consan erm. The resuls (available on reques) show ha here is no significan relaionship beween he lag of days and he price drop a he announcemen day, in conras wih general socks. This is differen from he findings abou general socks, for which here exiss a significan relaionship beween he wo variables. In his secion, we find ha marke reacions o REIT capial issues suppor he exisence of he asymmeric informaion in he marke iming of REITs. Suppor for he pecking-order heory is no srong, alhough asymmeric informaion exiss. When we compare he equiy marke iming paern of REITs wih general socks, we find ha he imevarying asymmeric informaion found in general socks is no as srongly suppored in he REIT indusry. One possible reason may be ha REITs operae under less asymmeric informaion. This is consisen wih he findings in sub-hypohesis 1. 16

18 5. Marke Efficiency and Marke Timing in REIT Equiy Offerings The inefficiency-based marke iming explanaion has been suppored by prior research. Pos IPO long run underperformance for general socks is believed o challenge he concep of marke efficiency. Managers ime secondary equiy offerings (SEO) o ake he advanage of marke inefficiency. Asymmeric informaion beween invesors and managers is no required here. Wha maers is he marke iming abiliy of managers. In oher words, long-run underperformance from REIT SEOs would challenge he noion marke efficiency for REITs. Generally researchers use wo mehodologies o calculae long-run performance. The firs mehod is o compare reurns of SEO firms wih non-seo firms (Rier, 1991). The second mehod is he Fama and French (1993) hree facor model. Since mos of he REITs issued equiy over he sudy period, he second mehod is applied o es for abnormal reurns. A four-facor model ha includes he momenum variable in Carhar (1997) is esimaed. The porfolios are formed comprising REITs ha issued equiy over he previous five years. The sudy period is January 1993 o March The average monhly reurn of REITs ha have issued equiy in he pas five years is calculaed for each monh wih he porfolio rebalanced monhly. Boh value-weighed and equally-weighed monhly reurns are calculaed. Nex he abnormal monhly reurn for each porfolio is regressed agains he hree Fama and French (1993) equiy risk facors, Marke, SMB, and HML, Carhar s (1997) momenum variable, and he reurn on he NAREIT index minus he risk-free rae. The NAREIT index reurn is added o mimic he risk in reurns specifically relaed o he real esae indusry (Buimer, Hyland and Sanders, 2005). We firs use he above four facors in he regression, hen he NAREIT index is included. The regression equaion is REIT Rf + β MOMENTUM 5 = α + β 1 ( NAREIT Rf ) + β 2 ( Marke Rf ) + ε + β SMB + β HML 3 4 Equaion (1) REIT : The value-weighed (equally-weighed) reurn in quarer of REITs wih equiy offering in he las 5 years; Rf : Risk-free rae in quarer ; NAREIT : Reurn of NAREIT equiy REIT index in quarer ; Marke : Value-weighed CRSP reurns in quarer ; 17

19 SMB : HML : Momenum : The difference beween he reurns on small and big sock porfolios wih abou he same book o marke equiy in quarer ; The difference beween he reurns on high book-marke firms and low book-marke firms in quarer ; The high momenum sock reurn minus low momenum sock reurn where momenum is measured based on pas one-year reurn. Here SMB HML and Momenum are calculaed on all socks in CRSP. No long-run underperformance is deeced as he consan erms are insignifican (resuls in Table 8). When NAREIT reurn is included in he regression, he explanaory power of he equaions increase, boh in equal and value weighed porfolios. A he same ime, adding NAREIT reurn ino he equaion decreases he explanaory power of oher variables. An ineresing side quesion is he effec of firm size. Graham and Harvey (2001) noe ha large firms pracice marke iming more frequenly han smaller firms. As such, i may be possible ha we do no find evidence of underperformance because small REITs are no good a marke iming. So could he phenomenon hold for large REITs? We sor he daa by oal marke value and calculae he long-run performance for large REITs (op quinile). The resuls (Panel B of Table 8) coninue o indicae ha here is no evidence of long-run underperformance for large REITs 7. One major difference beween he resuls in Panel A and Panel B is ha he variable NAREIT-Rf becomes insignifican in Panel B. This may be because he NAREIT index is no a good proxy for large REITs, so he NAREIT index reurn is no significan in Panel B. Alernaively, we may inerpre ha large REITs behave more like general socks, bu no smaller REITs. Our resul is consisen wih Buimer, Hyland and Sanders (2005), in which long-run underperformance from REITs IPO daa is no found. Since boh posiive and negaive conceps exis in our resul, he argumen in Fama (1998) is valid ha he finding of longrun underperformance is no robus. Thus he marke efficiency is no violaed 8. In summary, we examine he equiy marke iming behavior of REITs by esing he informaion-based and marke inefficiency-based marke iming hypoheses. Our resuls show 7 A similar es is also carried ou for small REITs and no evidence of underperformance is found. 8 In Howon, Howon and Friday (2000), i is found ha equiy REITs underperformed an index of non-issuing REITs for hree years. Using he same ime period ( ) wih value-weighed sample and he four-facor model as above, we could no find underperformance. 18

20 ha asymmeric informaion exiss beween REITs managers and invesors, alhough i is less obvious compared o general socks. By conras, marke efficiency is no challenged by REIT SEOs. 6. Marke Timing in REIT Deb Issuances Graham and Harvey s survey (2001) concludes ha companies ry o ime marke ineres raes by issuing more deb when hey feel ha ineres raes are low. Barry, e al. (2005) find empirical suppor for his relaion. Specifically, we es wo perspecives of deb marke iming: backward-looking iming and forward-looking iming. These are esed in he nex wo hypoheses. Backward-Looking Hypohesis: The amoun of deb issues of REITs is correlaed wih he level of ineres raes and heir relaion o hisorical raes. To es his hypohesis, we define he dependen variable as he oal amoun (log value) of public deb issued by individual REIT in a quarer (AMT). If here are several deb issues in he same quarer, hen he oal amoun is calculaed. Independen variables include ineres rae and oher variables. 19

21 Table 8: Long-run performance wih Fama-French Four Facor Model Panel A: Full Sample Inercep Marke-Rf SMB HML Momenum Equally Weighed Sample ( ) *** (6.456) *** (6.209) *** (7.119) (-1.280) (0.6509) (0.581) (1.755) (0.577) (-0.809) *** (15.146) Value Weighed Sample (-1.556) *** (7.278) *** (5.156) *** (7.651) (-0.780) (0.107) (0.858) (-1.071) (0.133) (0.273) *** (24.441) NAREIT-Rf Adjused R-square Panel B: Large REITs Inercep Marke-Rf SMB HML Momenum Equally Weighed Sample (-0.368) *** (6.045) *** (3.457) *** (5.738) (-0.967) (0.337) *** (6.009) *** (3.304) ** (5.679) (-0.934) (0.553) Value Weighed Sample (0.505) (0.097) *** (2.174) (1.521) (-0.532) (0.273) (0.095) *** (2.122) (1.507) (-0.525) (0.059) NAREIT-Rf Adjused R-square (wih saisic in parenhesis) [***: significan a 1% confidence level] Rf : Risk-free rae in quarer ; Marke : Value-weighed CRSP reurns in quarer ; SMB : Difference beween he reurns on small and big sock porfolios wih abou he same book o marke equiy in quarer ; HML : Difference beween he reurns on high book-marke firms and low book-marke firms in quarer ; Momenum : High momenum sock reurn minus low momenum sock reurn where momenum is measured based on pas one-year reurn; NAREIT : NAREIT equiy REIT index reurn in quarer. 20

22 Barry, e al. (2005) find ha he amoun and number of deb issues are higher when curren ineres rae is low compared o hisorical ineres raes. To es his, boh he ineres rae levels and is rank relaed o hisorical raes are included as independen variables. Here we adop he concep of he ineres rank (IR), which is he ranking of he curren ineres level relaive o hisorical raes (Barry, e al., 2005). The key idea is ha managers form expecaions on he basis of mean reversion in ineres raes and fuure ineres raes are likely o fall a some ime in he fuure if he curren ineres raes are high, vice versa. If so, he hisorical rank of he curren ineres rae is an imporan indicaor of deb marke condiions. To calculae he hisorical rank of he curren ineres rae (IR), we use he following formula: IR = ( X Min) *10, Max Min where X is he curren ineres rae. Max and Min are he highes and lowes ineres raes in he pas en years respecively. IRANK is assigned he value of: 1 if 0 IR 1 2 if 1 < IR 2 10 if 9 < IR 10. As noed earlier, Guedes and Opler (1996) find ha erm spread is negaively relaed o deb issues, and erm spread is also included in he equaion. The equiy REIT price index reurn from NAREIT and S&P 500 index reurn are also included o conrol for he equiy marke condiions, which may influence deb issuance. In addiion, we include size and leverage o conrol for firm characerisics. A simple OLS regression is run o es his hypohesis and he resuls are given in Table 9. Boh he level of ineres rae (INT) and he hisorical rank of ineres rae (IRANK) have negaive effecs on he amoun of deb issues, bu ineresingly, i is only he ranking of he ineres rae ha has a significan effec on he amoun of issues. The lack of saisical significance for level of ineres raes (INT) is in conras wih he resuls in Barry, e al. (2005). 9 9 This difference could be due o wo reasons. Firs, Barry, e al (2005) use monhly daa insead of quarerly daa as in our analysis. A finer pariioning of he daa se could pick up greaer variaions in ineres raes. Second, he downward rending ineres rae environmen over he sudy period could also explain why deb issuance is no as sensiive o ineres rae levels. 21

23 We can also find ha larger REITs end o issue more deb and REITs wih higher leverage raio issue less deb, boh of which are consisen wih our expecaions. Overall, he empirical resuls suppor he backward looking deb iming hyphesis. Table 9: Effec of Ineres Rae and Ineres Rank on REIT Deb Offering variable coefficien -saisic inercep *** INT IRANK *** TS *** NAREIT S&P ASSET *** LEV ** R square: ***: significan a 1% confidence level **: significan a 5% confidence level *: significan a 10% confidence level Dependen variable is amoun of deb issued by REITs per quarer. IRANK: ranking of prevailing ineres rae level in quarer of issue; ASSET: log value of oal asses of REIT i a end of previous quarer; NAREIT: NAREIT equiy REIT index reurn a end of previous quarer; INT: Long-erm (10 year) ineres rae a end of previous quarer; S&P: S&P 500 composie index reurn a end of previous quarer; TS: Term-spread beween long-erm ineres rae and shor-erm ineres rae (3-monh T-bill rae) a end of previous quarer; LEV: Leverage raio of REIT i a end of previous quarer. Forward-Looking Hypohesis: Deb issuing decisions are followed by ineres rae movemens ha demonsrae manager s abiliy o ime fuure ineres raes. This hypohesis differs from he previous hypohesis in ha we es he forwardlooking iming abiliy of REITs managers. Pu differenly, if he ineres rae rises afer deb issues, hen forward-looking marke iming exiss. To es his hypohesis, we examine wheher ineres raes rise on average afer deb issues. We define I n I 0 as he long-erm ineres rae a n monhs afer he issuance minus he long-erm ineres rae a he issuance. 22

24 Table 10: Ineres Raes Subsequen o Deb Issues (I: % per annum) I 6 I 0 I12 I 0 I18 I 0 I 24 I 0 I 30 I 0 Average ** ** *** *** saisic Table 10 shows ha ineres raes do no rise afer issuances. Insead, hey even fall afer issuances. This resul indicaes ha managers of REITs do no have he forwardlooking iming abiliy which could allow hem o issue deb before ineres raes rise. As noed earlier, he above analysis mus be qualified in ha ineres rae decreased over he sudy period. 7. Securiy Choice of REITs Offerings Now ha we have esablished marke iming in REIT equiy and deb offerings, we proceed o es he marke iming heory agains wo oher capial srucure heories hrough analyzing he securiy choice beween deb and equiy in REIT offerings. Each capial srucure heory has differing assumpions and explanaions wih regard o he securiy choice beween equiy and deb. We resric our es o he informaion-based marke iming heory. As moivaed earlier, he key difference beween he informaion-based marke iming heory and he pecking-order heory is wheher here is a pecking order in he financing decision. In conras, he effec of he marke condiions on securiy offerings is considered o be imporan in marke iming heory. We use a probi model o es he securiy choice of REITs from 1993 hrough The dependen variable is zero if equiy is issued in one quarer, and one if deb is issued, condiional securiy being issued for he quarer. Independen variables include proxies of cos of capial and firm characerisics. The independen variables include 10 : (1) Marke o Book raio (M/B): The rade-off heory implies ha higher M/B raio is correlaed wih higher growh opporuniies and higher coss of financial disress. Thus less deb is used when he M/B is high, i.e. deb is negaively correlaed wih M/B. The 10 Here all he variables are calculaed using daa for he previous quarer. We also exend our analysis by using he previous wo quarers and he previous four quarers bu he resuls are qualiaively unchanged. The regression resuls are available on reques. 23

25 pecking-order heory saes ha high growh firms wih high M/B raio need more exernal capial. Since deb is preferred o equiy, high growh firms would end up wih high leverage raio and more deb. So deb is posiively correlaed wih M/B. In he marke iming heory, M/B raio is negaively correlaed wih deb and leverage. (2) Toal asses (ASSET): According o he rade-off heory, size relaed variables, such as oal asses are inverse proxies for volailiy and for bankrupcy coss. The larger he firm, he higher he leverage raio. The implicaion from he pecking-order heory is no very clear. This is because on he one hand, larger firms migh have more asses in place and suffer a greaer loss should adverse selecion happens; on he oher hand, larger firms may have less asymmeric informaion compared o small firms. The marke iming heory has no specific assumpions abou he size of firms and he securiy choice. (3) Leverage (LEV) (4) Sock price appreciaion (ST_AP): The marke iming heory assumes ha he higher he sock price appreciaion in he recen period, he higher probabiliy ha equiy is issued by firms o ake advanage of he appreciaion. (5) Toal deb (DEBT) (6) S & P 500 reurn (SP) (7) NAREIT reurn (NAREIT) (8) Ineres rae (INT): Deb marke iming implies ha firms choose deb if he curren ineres rae is comparaively low, so a lower ineres rae increases he probabiliy of a deb issue. (9) Term spread (TS): If he long-erm ineres rae is considered o be low compared o he shor-em rae, hen firms would choose more long-erm deb according o he marke iming heory. (10) Deb capaciy (DC): If he deb capaciy is reached, DC equals one, oherwise, zero. According o he pecking-order heory, when he deb capaciy has been reached, firms will choose equiy. Oherwise, deb is preferred o equiy. (11) Sock variance (SVAR): The rade-off heory implies ha firms wih more volaile cash flows face higher expeced coss of financial disress and herefore use less deb. The implicaion from he pecking-order heory is ha firms wih volaile socks are firms abou which beliefs are quie volaile. They would suffer more adverse selecion, and will use more deb and have higher leverage raios. (12) Propery ype dummy variables. 24

26 As here may be more han one deb offering by an individual REIT in he same quarer, muliple issues are aggregaed. A he same ime, only one ype of issue for each REIT is allowed in each quarer (equiy or deb) in he probi model. Cases where boh equiy and deb are issued in one quarer are excluded. In oal, 20 REITs issue duplicae securiies, and hese are excluded from he sample, and such excluded daa represens less han 10% of he enire daa sample 11. Afer reducion of hese issues and oher issues which do no have relevan daa, he final sample comprises 310 equiy issues and 211 deb issues. The resul of he probi model 12 in Table 11 shows ha sock price appreciaion (STK_AP) and long-erm ineres raes (INT) have significan effecs on he equiy/deb choice of REITs. The negaive coefficiens mean ha REITs prefer equiy o deb when he sock prices are increasing and when long-erm ineres is high. Again, he marke iming heory is srongly suppored in our securiy choice analysis. Size (ASSET) also has a significan influence on he securiy choice of REITs. The probabiliy of deb issues increases in REIT size, i.e., larger REITs has a higher likelihood o issue deb raher han equiy. This evidence suppors he rade-off heory, as larger REITs end o have lower bankrupcy risk and hus could afford higher deb levels. The resul on deb capaciy consrain (DC) shows logically ha here would be less deb offering when deb capaciy is reached. This lends suppor o no only he pecking-order heory, bu also he rade-off heory. Leverage raio (LEV) has significanly negaive effecs on he probabiliy of deb issues. We also noe ha he oal amoun of deb (DEBT) has an negaive bu insignifican effec. Since he leverage raio is no relevan for he marke iming heory, suppor is found for he rade-off heory. We also find ha REITs prefer deb o equiy when he erm-spread (TS) is high, alhough he effec is no significan. A high erm-spread can be explained by a higher expeced ineres rae in he fuure, and REITs would use more deb a presen o avoid fuure financial needs. As leverage raio is posiively correlaed wih bankrup risk, managers of REITs would no choose deb issues when he curren leverage raio is high, or when here is a lack of good invesmen opporuniies. However, REITs ha have larger holding asses could issue more deb. I is reasonable o infer ha higher leverage raio is considered o be 11 The resul of he probi model does no change when we include he ne amoun of hese 20 issues in our sample. If we use he ne issue, hese 30 issues reduce o 6 ne equiy issues and 14 ne deb issues. 12 A logi model is also esimaed and he resuls are similar. 25

27 risky for REITs, and his may be he reason why mos REITs keep heir leverage raio in an accepable level (such as 40% o 60%) 13. Propery ype also has an imporan influence on he securiy choice of REITs. Healh care and residenial REITs have higher probabiliy o offer deb relaive o REITs from oher secors. This could be due o sable cash flows and asses values of healh care and residenial REITs, which in urn increase he deb capaciy and reduce he public deb coss. A number of variables urned ou o be insignifican. We briefly discuss he implicaions here. M/B raio is ofen used as an indicaor of firm valuaion. High M/B raios indicae overvaluaion while low M/B raio sugges undervaluaion. If managers wish o ime he marke by issuing equiy when socks are overvalued, hen higher M/B raio will lead o more equiy issues. The resuls show ha higher M/B raio increases he probabiliy of deb issues (alhough only a 11% significance), bu no equiy issues. This is conrary o he predicion of he marke iming heory. One possible explanaion for his resul is ha M/B raio is no necessarily a good indicaor of he valuaion of REITs. Someimes i is believed ha M/B raio could no differeniae beween he growh of he firms and he overvaluaion of he firms (Lee, Myers and Swaminahan, 1999; Kohari and Shanken, 1997). If he M/B raio is a proxy for growh opporuniies raher han sock valuaion, hen he preference for deb when REITs have access o high growh opporuniies suppors he pecking-order heory. The rade-off heory, in posulaing ha higher growh firms uses less deb, is no suppored by our resul. Higher S&P 500 index reurn and NAREIT index reurn boh decrease he probabiliy of making deb issues bu he coefficiens are no significan. Baker and Wurgler (2000) sugges ha firms have he abiliy o ime he whole capial marke, and he equiy share in he oal equiy and deb issues could be used o predic fuure marke reurns. Higher equiy share would lead o lower marke reurns. Our resul shows, in conras, ha higher individual REIT reurns have significan effec on he securiy choice of REITs, bu general marke variables do no have significan effecs. I seems ha individual REIT sock reurn is more imporan han he overall marke condiions in he iming behavior of REITs. Higher sock reurn variance (SVAR) has negaive effec on he probabiliy of a deb issue, bu he effec is no significan. The sign of effec is consisen wih he rade-off heory, as he rade-off heory implies ha firms wih more volaile cash flows face higher expeced coss of financial disress and herefore use less deb. This is inconsisen wih he predicion 13 According o NAREIT( he levels of leverage for REITs are moderae in recen years, consisen wih our findings. 26

28 of he pecking-order heory ha firms wih volaile socks face more adverse selecion and use more deb. 27

29 Table 11: Probi Regression of Securiy Choice Dependen Variable: SECURITY CHOICE (zero if equiy is issued in one quarer, and one if deb is issued) Mehod: ML - Binary Probi (Quadraic hill climbing) Variable Coefficien Sd. Error z-saisic Prob. consan *** ASSET *** DEBT E DC *** INT *** LEV ** M/B NAREIT S&P STK_AP ** SVAR TS * DIVERSIFIED HEALTH_CARE *** INDUSTRIAL_OFFICE LODGING_RESORTS RESIDENTIAL *** SELF_STORAGE SPECIALTY ** Obs wih Dep=0 310 Toal obs 521 Obs wih Dep=1 211 ***: significan a 1% confidence level; **: significan a 5% confidence level; *: significan a 10% confidence level ASSET: log value of oal asses of REIT i a end of previous quarer; DEBT book value of oal long-erm deb of REIT i a end of previous quarer; LEV: Leverage raio of REIT i a end of previous quarer; DC: Deb capaciy dummy variable ha shows wheher he deb capaciy is reached when he new equiy/deb issue is made; INT: Long-erm (10 year) ineres rae a end of previous quarer; MB : Marke-o-book raio a end of previous quarer; NAREIT: NAREIT equiy REIT index reurn a end of previous quarer; S&P: S&P 500 composie index reurn a end of previous quarer; STK_AP: REIT sock price appreciaion in he previous quarer; SVAR: Variance of he sock reurns in he pas year; TS: Term-spread beween long-erm ineres rae and shor-erm ineres rae (3-monh T-bill rae) a end of previous quarer. Table 12 summarizes he above resuls o faciliae comparison. The analysis of he securiy choice of REITs in his secion shows ha boh equiy and deb marke condiions are imporan deerminans of he capial srucure changes of REITs. Despie he imporance 28