Secured Debt and Corporate Performance: Evidence From REITs

Secured Debt and Corporate Performance: Evdence From REITs Brent W. Ambrose The Pennsylvana State Unversty Shaun Bond Unversty of Cncnnat and Joseph Oo Natonal Unversty of Sngapore March 31, 2009 Contact Author: Brent W. Ambrose, Ph.D Jeffrey L. and Cndy M. Kng Faculty Fellow and Professor of Real Estate Drector, Insttute for Real Estate Research 368 Busness Buldng The Smeal College of Busness The Pennsylvana State Unversty Unversty Park, PA 16802 (814) 867-0066 (offce) (814) 865-6284 (fax) bwa10@psu.edu

Secured Debt and Corporate Performance: Evdence From REITs Abstract Agency theory n modern corporate fnance suggests the presence of a conflct of nterest between managers and shareholders. Pror theoretcal and emprcal research has dentfed leverage as an mportant mechansm that s lkely to mtgate agency costs. Although debt plays a central role n mtgatng corporate agency conflcts, relatvely few studes have examned the mplcatons that arse from the use of secured versus unsecured debt. Gven the dfferences that exst n the ncentves to engage n costly montorng actvtes between secured and unsecured debt holders, we explore the role of secured and unsecured debt as montorng devces. Specfcally, we answer the followng queston: Does the use of unsecured debt result n a performance dfference between frms that employ a hgher proporton of unsecured debt versus frms that utlze secured debt? We fnd robust evdence that an ncrease n the use of secured debt by REITs s assocated wth postve excess returns n the followng quarter.

Introducton The current crss n the credt markets rases a number of nterestng questons regardng the use of debt by corporatons. One aspect that has generated a number of conflctng theores n the academc lterature s the use of collateral n debt contract. A number of competng theores have developed surroundng the use of collateral (or secured) debt by frms. Collateral s often assocated wth models based on adverse selecton or moral hazard. In adverse selecton models, borrowers use collateral to sgnal qualty suggestng that less rsky frms are wllng to utlze secured debt (Bester, 1985; Chan and Kanatas, 1985; Besanko and Thakor, 1987a, b). However, moral hazard models assume that the use of collateral mproves the ncentves for borrowers to work hard to repay debt (Chan and Thakor, 1987; Boot and Thakor, 1994). Overall, these models of debt choce based on adverse selecton and moral hazard pont to secured debt as beng assocated wth better subsequent performance. In contras recent theoretcal models call nto queston the assumpton that less rsky borrowers prefer collateral. For example, the model developed n Boo Thakor, and Udell (1991) fnds that rsky borrowers wll utlze more collateral when borrower qualty s observable and moral hazard s present. Most recently, Inderst and Mueller (2007) present a model showng that observably rsker borrowers should pledge more collateral and tha holdng observable borrower rsk constan secured loans wll be more lkely to default ex post. Interestngly, Inderst and Mueller (2007) obtan ther result wthout appealng to assumptons regardng the presence of moral hazard or adverse selecton. Jmenez, Salas and Saurna (2006), on the other hand, observe a negatve 1

assocaton between collateral and borrower s rsk, whch they nterpret to be consstent wth theores that vew collateral as a soluton to adverse selecton problems. In ths paper, we emprcally test these competng theores regardng the use of secured debt. In order to drectly test whether secured debt allevates the problems assocated wth adverse selecton and moral hazard, we make use of the fundamental prncple n fnance that rsker assets must generate hgher expected returns. Specfcally, we answer the followng questons: Does the use of unsecured debt result n a performance dfference between frms that employ a hgher proporton of unsecured debt versus frms that utlze secured debt? In order to emprcally test these theores, we rely on the unque features of the captal structure of real estate frms. For real estate frms or real estate nvestment trusts, the choce of secured or unsecured debt s hghly materal. Unlke other corporatons, REITs possess tangble assets that are easy to collateralze. Other frms, partcularly those that do not own tangble assets, do not have assets to pledge as collateral n the frst place. Furthermore, REITs are tax transparen thus avodng the controversy over the relatonshp between corporate tax rates and utlzaton of secured debt. 1 Thus, we propose a novel test of the competng theores regardng the use of secured or unsecured debt usng changes n debt choce by REITs. To prevew our results, we fnd strong evdence for the relatonshp between changes n the use of secured debt and subsequent performance. At the one-month, oneyear and three-year holdng perod horzons, we fnd that an ncrease n the use of secured debt s assocated wth postve excess stock returns. For example, the results ndcate that a 1 percent ncrease n secured debt over the prevous year mples that the 1-year holdng 1 Accordng to DeAngelo and Masuls (1980), frms facng hgher tax rates should ssue more debt. Therefore, frms wth hgher tax rates should ssue the lowest prorty and hence most rsky debt clams n order to ncrease the value of ther tax sheld. 2

perod return s 5.2 bass ponts hgher than the 1-month holdng perod return of the REIT that dd not alter ts use of secured debt (holdng all else, ncludng total leverage, constant.) The observed postve relaton between frm utlzaton of secured debt and future stock performance s consstent wth adverse selecton and moral hazard models of secured debt. Examnng the determnants of changes n secured debt rato, we observe that large frms wth low leverage are more lkely to ncrease ther secured debt rato. On the bass that they are less rsky borrowers, the results suggest that adverse selecton models are more relevant n explanng the postve relaton between frm utlzaton and future stock performance. Thus, REITs ssue secured debt to sgnal qualty. Our paper s structured as follows. In the next secton, we outlne the prevous lterature that examnes the use of unsecured and secured deb and the factors that determne corporate debt choce. In Secton 3, we descrbe the data and emprcal method employed n the analyss. In Secton 4, we present the results of our analyss, and n Secton 5 we conclude. Lterature on Secured versus Unsecured Debt Although debt plays a central role n mtgatng corporate agency conflcts, relatvely few studes have examned the mplcatons that arse from the use of secured versus unsecured debt. Unsecured debt refers to general oblgaton bonds and secured debt refers to debt collateralzed by specfc assets. Investors n secured and unsecured debt look to the frm s cash flow for payment of nterest and prncpal. However, n bankruptcy, unsecured debt holders have a general clam on the frm s assets whereas 3

secured debt holders have specfc corporate assets that can be sold to recover any losses and these clams take precedence over the unsecured credtors clams on the frm s assets. As a resul unsecured credtors have greater ncentves to engage n montorng actvtes than secured credtors. The customary justfcaton for secured debt s that ssung collateral helps reduce borrowng costs through a reducton n the lender s admnstraton costs and ncreasng the costs assocated wth default (Barro, 1976; Benjamn, 1978). Frs collateral helps reduce admnstratve and enforcement costs because the lender holds ttle to the pledged asset and can thus quckly sell the asset to cover loan losses assocated wth borrower default. Secured debt also helps lower screenng costs for credtors because they do not have to concern themselves wth the frm's other assets snce ther nterests are protected by the pledged assets (Shah and Thakor, 1987). Second, snce pledged assets cannot be dsposed of easly, secured debt lowers total costs of borrowng by lmtng asset substtuton opportuntes. Consequently, fnancng new projects wth secured debt also helps allevate the undernvestment problem assocated wth rsky debt. As the cost of borrowng s lower for debt wth securty provsons, the frm can undertake projects that t would have otherwse foregone f usng normal debt (Stulz and Johnson, 1985; Berkovtch and Km, 1990). Thus, snce the nterest rate charged on secured debt s lower than that of unsecured deb frms should ssue as much secured debt as possble. Such ratonalzaton s, however, one sded because t gnores the costs nvolved n ssung secured debt (Scot 1977). Frs the costs of establshng secured debt contracts are more expensve than normal debt contracts because of addtonal reportng requrements (Smth and Warner, 1979a). Second, collateralzng a loan leads to a loss of 4

flexblty wth respect to use and sale of the pledged asset (Stulz and Johnson, 1985). Thrd, a moral hazard problem may arse due to the dvergence of nterest between the borrower and lender wth regard to mantanng the value of the pledged asset (Igawa and Kanatas, 1990). Fourth, whle ssung secured debt may allevate the under-nvestment problem, the lower cost of borrowng may create an ncentve towards excessve nvestment (Berkovtch and Km, 1990). Thus, the frm s decson to utlze collateralzed debt depends on the trade-off between the benefts and costs of securng the debt (Smth and Warner, 1979a). Naturally, secured loans wll only be ssued f the benefts of dong so exceed the costs. However, n an effcent captal marke the debt senorty decson s nconsequental to the value of a frm. Fama and Mller (1972), for example, contend that a frm cannot alter the total value of ts outstandng securtes by ssung or retrng any type of securty f the securtes are protected aganst fnancng actons that would reduce ther value wthout adequate compensaton. The noton of perfect nformaton mples that unsecured credtors, who necessarly take on greater rsk, wll demand hgher nterest rates to compensate for the smaller pool of assets avalable to satsfy ther clams. In equlbrum, the collateral pledged provdes no net benefts to the frm and total nterest costs should be unaffected by the exstence of collateral (Fama, 1978; Schwartz, 1981). In the search for a meanngful exstence of secured deb researchers have relaxed the rgd assumptons assocated wth perfect markets. For example, recent studes by Jmnez and Salas (2004) and Inderst and Mueller (2007) have examned the role of 5

collateral n an mperfectly compettve loan market. In ther model, collateral mtgates ncentve problems on the part of the lender. 2 Theoretcal models have assumed agency problems on the part of the borrower. Bester (1994), for nstance, llustrates a stuaton where a borrower s motvated to default (even though the underlyng project s a success) because of the chance of ganng debt forgveness. The motvaton to chea however, s reduced f collateral s posted. Hence, collateral n the Bester model protects credtors aganst cheatng by borrowers. Besdes resolvng the moral hazard problem, theoretcal models of secured debt have also assumed that collateral reduces potental adverse selecton problems n the presence of asymmetrc nformaton (Smth and Warner, 1979a; Chan and Thakor, 1987). The underlyng premse n these models s the nablty of the lenders to dstngush between good and bad borrowers. Consequently, an nterest rate that reflects the average qualty of borrowers n the market results n an under-prcng of low qualty frms but an overprcng of hgh qualty frms. As a resul hgh qualty frms do not have ncentves to enter the market leadng to an adverse selecton stuaton. Under such condtons, secured debt can play a role n sgnalng the real worth of a frm. In the model of Chan and Kanatas (1985), collateral functons as a credble sgnalng devce. In the case of a hgh qualty projec the borrower offers more collateral because of the low probablty of default. Hgh rsk projects, on the other hand, have a hgher chance of falure and consequently, the prospect of the collateral beng forfeted s also hgher. Thus, the 2 Holdng local market competton constan the Inderst and Mueller (2007) model predcts that observably rsker borrowers should pledge more collateral and tha holdng observable borrower rsk constan collateralzed loans are more lkely to default ex post. The above two predctons, however, do not follow from exstng models of collateral. 6

borrower has to balance the gans (n the form of lower nterest rates) aganst the prospect of losng the collateral n the event of default. Examnng the role of collateral from the lender s perspectve, the credt ratonng models of Bester (1985) and Besanko and Thakor (1987) prescrbe that lenders can sort borrowers nto rsk classes by desgnng credt contracts wth nversely related nterest rates and collateral requrements. As opposed to the conventonal wsdom that hgh-rsk frms have to ssue collateral n order to attract credtors, ther models prescrbe that hgh rsk borrowers wll opt for contracts wth hgh nterest rates and low collateral requrements. Low rsk borrowers, on the other hand, wll choose contracts wth low nterest rates and hgh collateral requrements. In summary, the use of collateral can be ratonalzed on the grounds that t helps to resolve moral hazard and adverse selecton problems. The adverse selecton models predct that low-rsk borrowers wll pledge more collateral. Lkewse, the moral-hazard models are also based on the premse that postng collateral mproves borrowers ncentves to work hard, thereby reducng ther lkelhood of default. In contras Boo Thakor and Udell (1991) fnd that hgh rsk borrowers may pledge more collateral and hence, collateralzed loans may be rsker ex post. Furthermore, Inderst and Muller (2007) suggest that f borrower qualty and effort are substtutes, low-qualty borrowers post collateral to commt to hgher effort. Although the use of secured debt has theoretcal support for resolvng moral hazard and adverse selecton problems, the relaton between frm utlzaton of secured debt and future stock performance s less clear. 7

Despte the wealth of theores on the economc relevance of secured deb only a few studes have examned emprcally the factors determnng corporate secured debt choce. For example, Leeth and Scott (1989) fnd that the probablty of usng a secured loan s postvely related to the lkelhood of defaul loan sze, loan maturty and marketablty of assets, and s also dependent on changes n the economc and legal envronment. Barclay and Smth (1995a) argue that a frm wth more growth opportuntes should have a greater proporton of ts long-term labltes n senor clams such as captalzed leases or secured debt. 3 However, they dd not fnd any sgnfcant relatonshp between growth opportuntes and secured debt. Chang et al (2007) observe a sgnfcantly postve relatonshp between a frm s nvestment opportuntes and ts stock prce response to announcements of secured debt ssues. They nterpret the results to support the nvestment opportuntes hypothess,.e. secured debt fnancng s more valuable for ssung frms wth hgh growth opportuntes. However, they do not compare the long-term performance of frms based on ther debt choce. To summarze, the lterature ponts to two competng hypotheses regardng the relaton between secured debt and frm stock performance. Frs models assocated wth adverse selectonsuggest that frm rsk should be negatvely assocated wth the use of secured debt. That s, less rsky frms are more wllng to use secured debt than rsky frms. In contras moral hazard and recent models focusng on borrower effort mply that hgher rsk borrowers wll utlze secured debt. Thus, these theores pont to contrastng emprcal hypotheses regardng the relaton between the use of secured debt and frm rsk. Although these models have conflctng vews on who uses secured deb 3 The agency costs assocated wth asset substtuton and undernvestment are lkely to be more serous for frms wth hgher growth optons n ther nvestment opportunty sets snce they have more flexblty n ther choce of future nvestments (Ttman and Wessels, 1988; Barclay and Smth, 1995a,b). 8

they are nevertheless consstent n predctng a postve relaton between change n secured debt and subsequent corporate performance. It follows that we should see a postve relaton between an ncrease n the usage of secured debt and the frm s subsequent stock performance f the adverse selecton and moral hazard theores of debt usage are correct. However, a postve relaton between frm qualty and the use of secured debt s consstent wth the collateral sgnalng models, whlst a negatve relaton between frm qualty and the use of secured debt s consstent wth the moral hazard theory of collateral. In the next secton, we outlne a smple emprcal test of these competng hypotheses. Data and Research Desgn We examne all equty REITs havng nformaton reported n the SNL database and securty prces n the Center of Research n Securty Prces (CRSP) database. After flterng for obvous errors or ncomplete nformaton, we collected fnancal nformaton and stock prces for 114 publcly traded REITs spread over 73 quarters (1990Q1 to 2008Q2). Because some key accountng data where not recorded n the data durng the early years of the sample, the sample used for estmaton s effectvely reduced to the perod 1992Q2 to 2007Q4. To solate the collateralzaton decson from leverage decson, we normalze the amount of secured debt (SecDebt) employed by a frm by ts total debt. We measure the annual change n the secured debt as: SecDebt SecDebt t SecDebt t 1 9

Table 1 presents the descrptve statstcs of SecDebt as well as several other frm attrbutes of the fnal sample (number of observatons, debt rato, secured debt rato, return on assets, and prce-ffo rato). Growth opportuntes are proxed by the prce-tofunds from operatons (FFO) wth a hgher prce multpler representng more growth opportuntes. Frm proftablty s measured by the return on assets (ROA). As evdent n Table 1, the sample sze changes dramatcally over the study perod. The number of REITs n the sample ncreases from a low of 21 n the frst quarter of 1992 to 108 n 2006 and 2007. Furthermore, secured debt s on a downward trend durng the sample 4. Ths contrasts wth the amount of leverage used by frms (debt rato), whch has generally been ncreasng over ths tme. Ths trend s more evdent n Fgure 1, whch shows the cross-sectonal mean of the secured debt rato and leverage rato on a quarterly bass over the sample perod. Fgure 1 shows the declnng average secured debt rato and the correspondng ncrease n leverage. Table 2 presents the mean and medan returns over 1-month, 3-month, 1-year and 3-year holdng perods for the full sample as well as sub-samples of the REITs, parttoned by the change n ther secured debt rato. 44.6% of the observatons experenced a reducton, 31.2% experenced an ncrease, and 24.2% dd not see any change n ther secured debt ratos over the prevous quarter. The overall pattern n Table 2 suggests that REITs n the frst category tend to pose below average returns, whlst REITs n the second category tend to regster above average returns. Ths observaton s consstent wth a postve relatonshp between change n secured debt rato and frm stock performance. 4 We do not nclude the 100% observaton n 2008 n makng ths clam, as nformaton s only avalable for two companes n ths perod. 10

To formally test the competng hypotheses concernng the relaton between secured debt use and frm stock performance, we estmate an OLS regresson to dentfy the determnants of stock performance wth the key varable of nterest beng the level of secured debt employed by the ndvdual frm. Specfcally, we estmate the followng equaton: r t j, l 3 6 9 11 l 1 Beta DUM UPREIT l prce _ FFO 7 Age 10 1 SecDebt 4 ROA 8 SelfMgt Recesson 11 5 2 debt rato SelfAdv ln( sze e ) (1.) where r t+j, s the contnuously compounded holdng perod return for REIT from quarter t to t+j (j=1-month, 3-months, 1-year, and 3-years), and SecDebt s the change n the secured debt rato n quarter t for REIT. In equaton (1), we control for frm sze by ncludng the natural log of REIT 's market captalzaton n quarter t ( ln( sze ) ). Chan and Kanatas (1985) suggest that smaller frms, wth less nformaton avalable to lenders, may use more secured debt to sgnal ther project qualty. Smth and Warner (1979) also argue that smaller frms use secured debt more frequently because they have a hgher probablty of lqudaton. Both of these models mply that the use of secured debt may be assocated wth frms havng growth opportuntes. Snce growth opportuntes are correlated wth future stock prce performance, we nclude the rato of REIT 's funds from operatons (FFO) to ts stock prce n quarter t (prce_ffo ) n order to control for dfferences n growth opportuntes. Snce future stock prce performance also reflects the frm s proftablty and use of leverage, we nclude as control varables ROA, REIT 's return on assets n quarter Debt_Rato, the debt asset rato for REIT n quarter RISK, the systematc rsk for REIT n quarter t and DUM l, a dummy varable capturng 11

REIT 's property sector. 5 In addton, we add three more dummy varables capturng the REIT s organzaton structure, namely UPREIT status (UPREIT), self-managed (SelfMgt) and self-advsed (SelfAdv).Fnally, to test for the mpact of a credt crss on the use secured deb we nclude a dummy varable Recesson to denote the 2001 recesson. Table 3 summarzes the varable names and defntons. Results We begn the analyss by examnng the wthn frm varaton n the secured debt rato (SecDebt). Fgure 2 dsplays the scatter plot of the average secured debt rato and the wthn frm standard devaton of the secured debt rato ftted wth a second order polynomal trend lne. The dea behnd ths presentaton s to see f varatons n the secured debt rato are common or f frms rarely adjust ths rato. It s clear from Fgure 2 that a small number of frms target a partcular secured debt rato and do not change ths rato (note the low volatlty around 0% and 100%). However, we do see a large number of frms that exhbt wde varatons n ths rato. The second order polynomal trend lne dsplays an nverted U-shape suggestng the hghest shfts n secured debt use occur at the 50 percent secured debt rato. We now turn to a formal analyss of the mpact of changes n the use of secured debt on frm stock performance. Table 4 reports the coeffcents from the estmaton of equaton (1). The results show weak evdence for the relatonshp between secured debt and subsequent performance at the one-month holdng perod horzon wth the coeffcent for change n the secured debt rato beng postve and sgnfcantly (at the 10 percent 5 The sectors covered nclude dversfed, healthcare, hotel, ndustral, manufactured homes, mult-famly, offce, regonal mall, retal other, storage space, shoppng center, and specalty retal. 12

level). The coeffcent for SecDebt mples that a 1 percent ncrease n the use of secured debt corresponds to a 5.2 bass pont ncrease n the 1-month holdng perod return. The results, however, show no evdence of sgnfcant relatonshp between secured debt and subsequent long-term performance. In partcular, the coeffcents for SecDebt are not statstcal sgnfcant for the three-month, 1-year and 3-year holdng perods. Table 4 shows that the debt rato s sgnfcantly postve over each horzon perod. Consstent wth the addtonal rsk assocated wth hgher leverage, we see that REITs wth hgher debt ratos have consstently hgher future holdng perod returns. We also see that the systematc rsk of the frm s weakly sgnfcant n the 1-month holdng perod return regresson. Not surprsng, accountng profts (ROA) and stock performance are related postvely and the relatonshp s statstcally sgnfcant n two of the cases. Frm age s also postvely related to stock performance. The lack of sgnfcance for the prce to FFO varable mples that stock returns are not correlated wth current perod valuatons. Consstent wth the lterature, frm sze s negatvely related wth future stock returns up to one year holdng perod. Beyond the one year holdng perod, frm sze and future stock returns are postvely related. The results also show that UPREITs and REITs that are externally managed tend to perform poorly. Robustness Test One concern wth the results reported n Table 4 s that the use of secured debt may be endogenous to factors assocated wth future stock performance. To control for ths possblty, we estmate a two-stage regresson model that ncorporate the potental 13

for REIT past performance to mpact ts shft n the use of secured debt. Specfcally, we estmate the followng frst-stage regresson of the change n the secured debt rato ( SecDebt t, ): SecDebt 4 7 ROA t t 1, SelfMgt 11 l 1 DUM 5 l Debt _ Rato 8 SelfAdv 1 X 2 ln( sze e 6 t 1, ) UPREIT (2.) where X represents a seres of year dummy varables to capture any tme-varyng effects over the study perod. In ths specfcaton, the change n the secured debt rato s regressed on the frm s sze and return on assets n quarter t-1. The resduals from the OLS estmaton of (2) are then used n the followng second-stage regresson: r t j, 3 6 9 l 11 l 1 Beta DUM UPREIT l prce_ FFO 7 Age 10 1 4 ROA SelfMgt SecDebt 8 Recesson 11 5 2 debtrato SelfAdv ln( sze e ) (3.) where SecDebt t, represents the change n secured debt rato resduals from the estmaton of equaton (2). In ths framework, SecDebt t, s the REIT s devaton from the expected change n the secured debt rato gven ts proftablty and frm characterstcs at each quarter. Thus, SecDebt t, corresponds to the unexpected shft n use of secured debt. Under the hypothess that secured debt allevates the undernvestment problem, then a postve ncrease n the use of secured debt above the general trend n secured debt usage should correspond to a postve future stock performance. 14

Table 5 reports the estmaton results for the frst stage OLS regresson of equaton (2). Interestngly, debt rato has a negatve and sgnfcant mpact on REITs usage of secured debt. Frm sze, on the hand, s related postvely to the use of secured debt. Both results are contrary to the argument that small frms wth hgh leverage tend to depend more on secured debt when rasng external debt or the moral hazard models of secured debt whch prescrbe a postve relaton between frm rsk and collateral usage. The results are, nevertheless, consstent wth adverse selecton models of secured debt that less rsky frms ssue more collateral to sgnal ther qualty. However, we do see that UPREITs are more lkely to ncrease ther use of secured debt and dfferences n the use of secured debt exst across property sectors as REITs that focused on shoppng centers and specalty propertes are less lkely to ncrease ther use of secured debt. Proftablty (ROA), however, does not have a sgnfcant mpact on the frm s decson to alter ts secured debt rato. Turnng to the second stage model, Table 6 reports the results for the estmaton of equaton (3). Ths model ncludes the resdual from the estmaton of equaton (2) as a measure the mpact of devatons from expected changes n the use of secured debt. Comparng the results wth Table 4 (equaton 1), we agan fnd that REITs wth hgher debt ratos have consstently hgher future holdng perod returns. Although REITs wth hgher current perod valuatons (prce-to-ffo) have lower future holdng perod returns, the coeffcents are not statstcally sgnfcant. The coeffcents for the other varables n the model also behave as reported n Table 4. Agan, we note that the estmated coeffcent for SecDebt t, s only statstcally sgnfcant for the 1 month holdng perod horzon. Thus, the reported results are robust despte the potental endogenety problem. 15

Concluson Agency theory n modern corporate fnance suggests the presence of a conflct of nterest between managers and shareholders n frms and pror research has dentfed leverage as an mportant mechansm that s lkely to mtgate these agency costs. Gven the dfferences that exst n the ncentves to engage n costly montorng actvtes between secured and unsecured debt holders, we explore the mpact of the use of secured and unsecured debt on future stock prce performance. Usng a sample of REITs, we fnd evdence for the relatonshp between the use of secured debt and subsequent stock prce performance. The estmated coeffcents from a regresson of future holdng perod returns on varables assocated wth the use of secured debt ndcate that a 1 percent ncrease n the use of secured debt corresponds to a sgnfcant 5.2 bass pont ncrease n the 1-month holdng perod return, after controllng for the use of leverage, sze, property segmen proftablty, and current quarter prcng rato. The observed postve relaton between frm utlzaton of secured debt and future stock performance s consstent wth adverse selecton and moral hazard models of secured debt. In adverse selecton models, borrowers use collateral to sgnal qualty suggestng that hgh-qualty (less rsky) frms are wllng to utlze more secured debt. Moral hazard models, n contras assume that the use of collateral mproves the ncentves for borrowers to work hard to repay debt. On the premse that borrower qualty and efforts are substtutes, low-qualty (more rsky) borrowers post collateral to commt to hgher effort n the presence of moral hazard. Although these model have conflctng 16

vews on who uses secured deb they are nevertheless consstent n predctng a postve relaton between change n secured debt and subsequent corporate performance. Examnng the determnants of changes n secured debt rato, we observe that large frms wth low leverage are more lkely to ncrease ther secured debt rato. On the bass that they are less rsky (hgh qualty) borrowers, the results suggest that adverse selecton models are more relevant n explanng the postve relaton between frm utlzaton and future stock performance. In summary, REITs ssue secured debt to sgnal qualty. 17

References Agrawal, A. and C.R. Knoeber (1996) "Frm Performance and Mechansms to Control Agency Problems between Managers and Shareholders", Journal of Fnancal and Quanttatve Analyss 31(3), 377-97. Barclay, M.J. and C.W. Smth (1995a) The Prorty Structure of Corporate Labltes, Journal of Fnance 50, 899 917. Barclay, M.J. and C.W. Smth (1995b) The Maturty Structure of Corporate Debt, Journal of Fnance 50, 609 631. Barro, R.J. (1976) "The Loan Marke Collateral, and Rates of Interest", Journal of Money, Credt and Bankng 8(4), 439-56. Benjamn, D.K. (1978) "The Use of Collateral to Enforce Debt Contracts", Economcs Inqury 16, 333-59. Berkovtch, E. and E.H. Km (1990) "Fnancal Contractng and Leverage Induced Overand Under-Investment Incentves", Journal of Fnance 45(3), 765-94. Besanko, D. and A.V. Thakor (1987) "Collateral and Ratonng: Sortng Equlbra n Monopolstc and Compettve Credt Markets", Internatonal Economc Revew 28(3), 671-89. Bester, H. (1985) "Screenng versus Ratonng n Credt Markets wth Imperfect Informaton", Amercan Economc Revew 57, 850-55. Bester, H. (1994) "The Role of Collateral n a Model of Debt Renegotaton", Journal of Money, Cred and Bankng 26(1), 72-86. Boo A., A.V. Thakor and G.F. Udell (1991) Secured Lendng and Default Rsk: Equlbrum Analyss, Polcy Implcatons, and Emprcal Results, Economc Journal 101, 458-472. Chan, Y.S. and Kanatas, G. (1985) Asymmetrc Valuatons and the Role of Collateral n Loan Agreements, Journal of Money, Credt and Bankng 17(1), 84-95. Chan, Y.S. and A.V. Thakor (1987) "Collateral and Compettve Equlbra wth Moral Hazard and Prvate Informaton", Journal of Fnance 42(2) 345-63. Chang, S.C., S.S. Chen, A. Hsng and C.W. Huang (2007) Investment Opportuntes, Free Cash Flow, and Stock Valuaton Effects of Secured Debt Offerngs, Revew Quanttatve Fnancal Accountng 28, 123-145. 18

Damond, D.W. (1991) Montorng and Reputaton: The Choce between Bank Loans and Drectly Placed Debt, Journal of Poltcal Economy 99(4), 689-721. DeAngelo H. and R.W. Masuls (1980) Optmal Captal Structure under Corporate and Personal Taxaton, Journal of Fnancal Economcs 8, 3-29. Fama, E. (1978) "The Effects of a Frm's Investment and Fnancng Decsons on the Welfare of ts Stockholders" Amercan Economc Revew 68, 272-84. Fama, E.F. and M. Mller (1972), The Theory of Fnance, New York: Dryden Press. Grossman, S.J. and O. Hart (1982) "Corporate Fnancal Structure and Manageral Incentves", n J. McCall (ed.) The Economcs of Informaton and Uncertanty, Chcago: Unversty of Chcago Press, 107-40. Harrs, M. and A. Ravv (1988) Corporate Control Contests and Captal Structure: An Emprcal Test, Manageral and Decson Economcs 15, 563-76). Har O. and J, Moore (1990) A Theory of Corporate Fnancal Structure Based on the Senorty of Clams, Workng Paper 560, MIT. Igawa, K. and G. Kanatas (1990) "Asymmetrc Informaton, Collateral, and Moral Hazard", Journal of Fnancal and Quanttatve Analyss 25(4), 469-90. Inders R. and H.M. Mueller (2007) A Lender-Based Theory of Collateral, Journal of Fnancal Economcs 84, 826-859. Jensen, M.C. (1986) Agency Costs of Free Cash Flow, Corporate Fnance and Takeovers, Amercan Economc Revew 76, 323-329. Jmenez, G., V. Salas and J. Saurna (2006) Determnants of Collateral, Journal of Fnancal Economcs 81, 255-281. Jung, K., Y.C. Km and R.M. Stulz (1996) "Tmng, Investment Opportuntes, Manageral Dscreton and the Securty Issue Decson", Journal of Fnancal Economcs 42, 159-85. Leeth, J.D. and J.A. Scott (1989) The Incdence of Secured Debt: Evdence from the Small Busness Communty Journal of Fnancal and Quanttatve Analyss 24, 379-94. Maloney, M.T., R.E. McCormck and M.L. Mtchell (1993) "Manageral Decson Makng and Captal Structure", Journal of Busness 66(2), 189-217. Rajan, R. and A. Wnton (1995) Covenants and Collateral as Incentves to Montor, Journal of Fnance 50, 1113-1146. 19

Schwartz, A. (1981) "Securty Interests and Bankruptcy Prortes: A Revew of Current Theores", Journal of Legal Studes 10, 1-37. Scot J.H. (1977) Bankruptcy, Secured Deb and Optmal Captal Structure, Journal of Fnance 32, 1-19. Shah, S. and Thakor, A.V. (1987) Optmal Captal Structure and Project Fnancng Journal of Economc Theory 42, 209-243. Smth, C.W. and Warner, J.B. (1979a) Bankruptcy, Secured Deb and Optmal Captal Structure, Journal of Fnance 34(1), 247-251. Smth, C.W. and Warners, J.B. (1979b) On Fnancal Contractng: An Analyss of Bond Covenants, Journal of Fnancal Economcs 7, 117-161. Stulz, R.M. (1990) Manageral Dscreton and Optmal Fnancng Polces, Journal of Fnancal Economcs 14, 501-21. Stulz, R.M. and H. Johnson (1985) An Analyss of Secured Debt, Journal of Fnancal Economcs 14, 501-521. Ttman, S. and R. Wessels R (1988) The Determnants of Captal Structure Choce, Journal of Fnance 43, 1 20. Zwebel, J. (1996) "Dynamc Captal Structure under Manageral Entrenchment", Amercan Economc Revew 86(5), 1197-215. 20

Table 1: Descrpton of Sample (1992-2007) Year Number of REITS Secured Debt Rato Debt - Asset Rato ROA Prce FFO Rato 1992 21 75.30 0.33 4.87 9.31 1993 28 75.17 0.38 4.25 14.62 1994 43 79.84 0.38 9.10 12.59 1995 58 70.11 0.42 4.77 10.56 1996 65 74.83 0.42 3.91 11.03 1997 68 70.36 0.41 4.47 12.26 1998 76 58.75 0.42 4.18 12.43 1999 81 58.80 0.48 3.59 8.57 2000 86 59.73 0.49 3.67 7.92 2001 84 62.35 0.50 3.91 8.56 2002 86 63.41 0.51 3.32 10.69 2003 86 62.01 0.53 3.04 10.11 2004 87 62.56 0.50 3.03 15.95 2005 107 67.19 0.51 3.38 14.14 2006 108 64.42 0.53 3.61 16.25 2007 108 62.24 0.54 3.26 17.39 21

Table 2: Relatonshp between Changes n Secured Debt Rato and Stock Performance SecDebt refers to change n Secured Debt Rato over the precedng quarter. Ret1m, Ret3m, Ret1y and Ret3y refer to 1-month, 3-months, 1-year, and 3-year holdng perod returns. ALL SecDebt < 0 SecDebt = 0 SecDebt > 0 Mean Medan Mean Medan Mean Medan Mean Medan Ret1m 0.94% 0.97% 0.83% 1.06% 1.02% 0.78% 1.03% 1.10% Ret3m 4.43% 4.35% 4.57% 4.71% 4.96% 3.92% 3.81% 4.22% Ret1y 16.85% 15.91% 15.82% 15.54% 19.38% 15.97% 16.36% 16.71% Ret3y 65.99% 59.20% 60.51% 55.39% 69.85% 62.32% 70.46% 63.83% 22

Table 3: Defntons of Explanatory Varables Independent Varables Secured Debt Rato SecDebt Frm Sze Leverage Growth opportuntes Proftablty Systematc Rsk Age Symbol SecDebt SecDebt Ln(sze ) Debt_Rato Prce_FFO ROA Beta IPO Age Defnton of Proxes Secured debt over total debt. Change n Secured Debt rato over the precedng quarter. Natural logarthm of the market captalzaton. Total debt as a percentage of total captalzaton. Common stock prce at the end of the perod as a multple of the annualzed FFO per share. Return on average asse whch s calculated as net ncome a percentage of average assets (annualzed). Systematc rsk of the ndvdual REIT, computed over a rollng wndow of 60 months. Age of the frm, calculated from the date of frm s ntal publc lstng date. Property Focus 11 l 1 DUM l 1=dversfed, 2=healthcare, 3=hotel, 4=ndustral, 5=manufactured homes, 6=mult-famly, 7=offce, 8=regonal mall, 9=retal other, 10=storage space, 11=shoppng center. Note that the default category s specalty propertes. Partnershp Management Advsor UPREIT Self-Managed Self-Advsed Bnary varable equals one f the ndvdual REIT s structured as an operatng partnershp. Bnary varable equals one f the ndvdual REIT s nternally managed. Bnary varable equals one f the ndvdual REIT does not engage external advsor. 23

Table 4: Determnants of REIT performance (Equaton 1) (Dependent varable s 1-month, 3-month, 1-year, and 3-year holdng perod returns. Standard errors n parentheses.) Varable 1 month 3 month 1 year 3 year Intercept -0.03596-0.00745-0.00745-0.22808 (0.02745) (0.04532) (0.04532) (0.17836) SecDebt t, 0.00052* 0.00027 0.00027-0.00068 (0.00028) (0.00046) (0.00046) (0.00180) ln( sze ) -0.00413*** -0.00407-0.00361 0.02739*** (0.00147) (0.00242) (0.00492) (0.00954) prce_ffo -0.000004-0.000001-0.000003-0.000005 (0.000008) (0.000001) (0.000003) (0.000005) ROA 0.00059 0.00354*** 0.00381*** 0.00231 (0.00037) (0.00060) (0.00122) (0.00237) IPO Tme 0.00004 0.00008* 0.00029*** 0.00005 (0.00003) (0.00005) (0.00009) (0.00018) Debt_Rato 0.00101*** 0.00128*** 0.00473*** 0.01168*** (0.00014) (0.00023) (0.00047) (0.00092) Beta 0.00960* 0.00462-0.01047-0.04494 (0.00519) (0.00857) (0.01739) (0.03372) Dversfed -0.00422-0.04926-0.13291* -0.34001** (0.02348) (0.03876) (0.07868) (0.15254) Health Care 0.01844-0.05582-0.10759-0.24882* (0.02302) (0.03801) (0.07715) (0.14958) Hotel 0.03642 0.01140 0.08556 0.23397 (0.02489) (0.04110) (0.08342) (0.16172) Industral 0.00169-0.02085-0.05356-0.21638 (0.0246) (0.04062) (0.08245) (0.15985) Manufactured 0.00141-0.05745-0.17079** -0.53526*** Homes (0.02443) (0.04033) (0.08187) (0.15872) Mult-famly 0.00572-0.04530-0.13243* -0.42111*** (0.02376) (0.03923) (0.07964) (0.15439) Offce 0.00241-0.02888-0.05684-0.14637 (0.02357) (0.03892) (0.07900) (0.15316) Regonal Mall 0.00592-0.04135-0.09465-0.18707 (0.02385) (0.03938) (0.07994) (0.15499) Retal Other 0.00543-0.04986-0.14316* -0.34950** (0.02322) (0.03833) (0.07781) (0.15086) Storage Space 0.0305-0.01840 0.02256 0.11925 (0.02629) (0.04341) (0.08812) (0.17084) Shoppng Mall 0.00228-0.03333-0.08672-0.25202 (0.02366) (0.03906) (0.07929) (0.15372) 24

Recesson 0.00189 0.03832*** 0.02720 0.24415*** (0.00619) (0.01022) (0.02073) (0.04020) UPREIT -0.01998-0.02111* -0.10397*** -0.24403*** (0.00662) (0.01093) (0.02218) (0.04299) Self-Managed 0.03895*** 0.07583*** 0.30866*** 0.82465*** (0.00937) (0.01547) (0.03140) (0.06087) Self-Advsed -0.00834-0.01705-0.04162-0.08404 (0.01048) (0.01730) (0.03512) (0.06809) Adjusted R 2 0.0429 0.0415 0.0905 0.1844 *, **, and *** represent statstcal sgnfcant at the 10%, 5% and 1% level, respectvely. 25

Table 5: Estmaton Results for 1 st Stage Regresson of Determnates of Changes n Secured Debt Rato. (Dependent varable s SecDebt,. Yearly fxed effects are ncluded n the regresson but t ther results not reported.) Varable Coeffcent Standard Error t-statstc Intercept 4.8680 1.5093 3.23*** ROA t-1, -0.0223 0.0176-1.27 ln( sze t 1, ) 0.1783 0.0903 1.98** Debt_Rato t-1-0.0363 0.0084-4.35*** UPREIT 0.6618 0.3417 1.94* Self-Managed 0.1060 0.4497 0.24 Self-Advsed 0.0521 0.5583 0.09 Dversfed -2.6785 0.9944-2.69*** Health Care -2.9569 1.0069-2.94*** Hotel -3.1174 1.0231-3.05*** Industral -3.5093 1.0547-3.33*** Manufactured Homes -2.4269 1.1044-2.20** Mult-famly -3.8677 1.0162-3.81*** Offce -3.3121 1.0008-3.31*** Regonal Mall -3.5019 1.0343-3.39*** Retal Other -3.7735 1.0360-3.64*** Storage Space -1.8676 1.1831-1.58 Shoppng Mall -4.4265 1.0023-4.42*** Adjusted R 2 0.0228 *, **, and *** represent statstcal sgnfcant at the 10%, 5% and 1% level, respectvely. 26

Table 6: Estmaton of the second-stage regresson of REIT performance (Dependent varable s 1-month, 3-month, 1-year, and 3-year holdng perod returns. Standard errors n parentheses.) Varable 1 month 3 month 1 year 3 year Intercept -0.03287-0.00605-0.08749-0.23426 (0.02746) (0.04529) (0.09192) (0.17806) SecDebt t, 0.00055** 0.00040-0.00041 0.00130 (0.00028) (0.00046) (0.00093) (0.00180) ln( sze ) -0.00411*** -0.00396-0.00340 0.02797*** (0.00147) (0.00243) (0.00492) (0.00954) prce_ffo -0.0000004-0.0000014-0.0000037-0.0000050 (0.0000008) (0.0000013) (0.0000026) (0.0000050) ROA 0.00058 0.00345*** 0.00365*** 0.00203 (0.00037) (0.00061) (0.00123) (0.00238) IPO Tme 0.00004 0.00007 0.00030*** 0.00005 (0.00003) (0.00005) (0.00009) (0.00018) Debt_Rato 0.00099*** 0.00126*** 0.00475*** 0.01169*** (0.00014) (0.00023) (0.00047) (0.00091) Beta 0.00965* 0.00441-0.01295-0.05183 (0.00519) (0.00857) (0.01739) (0.03368) Dversfed -0.00560-0.05032-0.13152* -0.34042** (0.02349) (0.03875) (0.07864) (0.15233) Health Care 0.01708-0.05631-0.10514-0.24702* (0.02303) (0.03799) (0.07709) (0.14933) Hotel 0.03515 0.01259 0.09488 0.25042 (0.02489) (0.04106) (0.08333) (0.16142) Industral 0.00004-0.02115-0.04947-0.21149 (0.02460) (0.04057) (0.08234) (0.15951) Manufactured 0.00020-0.05808-0.17018** -0.53786*** Homes (0.02445) (0.04033) (0.08185) (0.15855) Mult-famly 0.00386-0.04613-0.12992-0.42049*** (0.02377) (0.03920) (0.07955) (0.15410) Offce 0.00086-0.03038-0.05577-0.14906 (0.02357) (0.03888) (0.07890) (0.15285) Regonal Mall 0.00427-0.04200-0.09254-0.18690 (0.02386) (0.03936) (0.07987) (0.15473) Retal Other 0.00346-0.05080-0.14030* -0.34915** (0.02323) (0.03831) (0.07775) (0.15062) Storage Space 0.02953-0.01879 0.02264 0.11736 (0.02632) (0.04340) (0.08809) (0.17064) Shoppng Mall 0.00004-0.03429-0.08376-0.25121 27

(0.02365) (0.03901) (0.07918) (0.15338) Recesson 0.00162 0.03821*** 0.02741 0.24423*** (0.00619) (0.01021) (0.02072) (6.08000) UPREIT -0.01957*** -0.02136* -0.10485*** -0.24564*** (0.00662) (0.01092) (0.02216) (0.04293) Self-Managed 0.03924*** 0.07715( 0.31268*** 0.83309*** (0.00939) (0.01549) (0.03144) (0.06091) Self-Advsed -0.00848-0.01780-0.04342-0.08829 (0.01049) (0.01731) (0.03513) (0.06805) Adjusted R 2 0.0432 0.0412 0.0907 0.1861 *, **, and *** represent statstcal sgnfcant at the 10%, 5% and 1% level, respectvely. 28

Standard Devaton of Secured Debt Rato (%) Fgure 1 Cross-Sectonal Mean of Secured Debt Rato and Debt Asset Rato 90 80 70 Secured Debt Rato 60 50 40 30 Debt Rato 20 10 0 1992M03 1993M03 1994M03 1995M03 1996M03 1997M03 1998M03 1999M03 2000M03 2001M03 2002M03 2003M03 2004M03 2005M03 2006M03 2007M03 Tme Fgure 2 Scatter Plot of Wthn Frm Secured Debt Rato and Volatlty of the Secured Debt Rato 50 45 40 35 30 25 20 15 10 5 0 0 20 40 60 80 100 120 Average Frm Secured Debt Rato 29