The Ineres Rae Sensiiviy of Value and Growh Socks - Evidence from Lised Real Esae This Version: January 25, 2017 Please reques he mos recen version from he auhors Chrisian Weis, Universiy of Regensburg René-Ojas Wolering, Universiy of Regensburg Seffen Sebasian, Universiy of Regensburg Absrac This paper analyzes he reurn sensiiviy of value and growh socks o changes of five ineres rae proxies. The analysis is based on monhly daa over he 2000 o 2014 period for a global sample of 487 lised real esae companies in 24 counries. This rich seing offers subsanial heerogeneiy in ineres raes across ime and counries. We find ha value socks are more sensiive o changes in he shor-erm rae han growh socks. This is consisen wih he heory ha invesors wih a shor invesmen horizon rade-off he high iniial yield of value socks agains a lower risk shor-erm rae. In conras, growh socks are more sensiive o changes in he long-erm rae, which is consisen wih he fuure cash flows of growh socks being discouned a a higher rae. We also find ha value socks are more sensiive o changes in he credi yield. Since credi coss have a direc impac on a firm s cos of capial, his resul is consisen wih risk-based heories of he value premium, which argue ha value socks are riskier, because hey end o have higher leverage and a larger defaul probabiliy. 1
1 Inroducion There is a subsanial body of research examining he varying performance characerisics of value socks and growh socks. By definiion, value socks are socks wih a low raio of price o fundamenal value, while growh socks are characerized by a high price relaive o heir fundamenal value. Numerous sudies show ha value socks on average ouperform growh socks (he so-called value premium), boh for he U.S. (Rosenberg e al., 1985; Fama and French, 1992) and inernaional sock markes (Fama and French, 2012; Asness e al., 2013). There are wo key explanaory approaches for he value premium: Firs, risk-based explanaions (e.g. Davis e al 2000, Zhang 2005, Liew and Vassalou 2000) wih he assumpion of fundamenals, e.g. leverage, size, are causing he average ouperformance of value socks. Second, behavioral based explanaions which imply he reurn anomaly is due o subopimal invesor behavior (e.g. Lakonishok e al., 1994; De Bond and Thaler, 1985). In essence, he risk-based explanaions pu emphasis on idiosyncraic risk. An alernaive furher explanaion aemp for he value premium are macroeconomic facors linked wih sysemaic risk, e.g. business cycles or moneary policy (e.g. Jensen and Mercer 2002; Hahn and Lee 2006). Lewellen (1999) argues ha in asse pricing models like he CAPM (Sharpe (1964) and Linner (1965),) or he ICAPM (Meron 1973) marke reurn does no compleely capure he relevan risk in he economy, and addiional facors are required o explain expeced reurns. To address his issue, Hahn and Lee (2006) exend he hree-facor model of Fama and French (1993) by wo addiional macroeconomic variables. The defaul spread and he erm spread proxy for credi marke and he moneary policy condiions. More recenly, Lioui and Maio (2014) show ha value socks have higher ineres rae risk han growh socks, suggesing ha he value premium can be explained by changes of he moneary policy. In his paper, we sysemaically analyze wheher and o wha exen, he performance of value and growh socks can be explained by five macroeconomic facors, i.e. differen proxies of ineres raes and yield spreads. The five facors include changes of he shor-erm ineres raes (STIR), long-erm ineres raes (LTIR), he erm spread (TERM), he corporae bond yield (CBY), and he defaul spread (DEF). The corresponding research quesion is: Do he reurns of value and growh socks reac differenly o changes of various ineres rae proxies? Why are lised real esae companies paricularly qualified o analyze he ineres rae risk of value and growh socks? The commonaliy among previous research is ha hey separae value and growh socks according o heir book-o-marke raios of equiy. Thus, wheher explicily or implicily, he book value of equiy is used as he proxy for a firm's fundamenal or inrinsic value. Mos academics agree ha a firm's inrinsic value is deermined primarily by he presen value of is fuure cash flows, which is no necessarily refleced by balance shee daa. In his sudy we use a more reliable indicaor of inrinsic value, which allows us o beer disinguish beween value and growh socks. In paricular, we use a global panel of 487 lised real esae companies (REITs and REOCs) in 24 counries over he 2000-2014 period. 2
Owing o heir peculiar characerisics, lised real esae companies are paricularly well-suied o sudy he impac of ineres rae changes. In paricular, here are hree obvious channels hrough which ineres raes may impac he sock marke reurns of lised real esae companies: 1) ineres rae changes impac he relaive araciveness of equiies compared o oher asse classes such as fixed income or he money marke, 2) Ineres raes impac he prices of he underlying properies of he lised real esae companies, and 3) ineres raes have a direc impac on he operaing performance, by influencing a firm s coss of deb. Combined wih he abiliy o reliably disinguish beween value and growh socks, his provides an ideal research seing o learn more abou he relaionship beween he various ineres raes and sock marke reurns. Our objecive is 1) o examine he ineres rae sensiiviy of value and growh socks, by using he NAV as he proxy for inrinsic value, and 2) o idenify differen paerns of sensiiviy for various proxies for ineres raes and yield spreads of value and growh socks, boh on a global basis. Our empirical approach is based on a monhly soring procedure. A he end of each monh, we rank all socks according o heir deviaions from inrinsic value, as measured by he NAV spread. We hen form hree porfolios whose reurns are observed over he following monh, wih he focus being on he value porfolio, which is defined as he quinile of socks wih he highes discoun o NAV. In order o examine he ineres rae sensiiviy, we conrol for ineracion effecs beween he value, middle and growh porfolio and changes of he respecive ineres rae proxy. Secondly, we conrol for risk-adjused reurns and include he ineracion erms ino four-facor models (Carhar, 1997). We find ha value socks are more sensiive o changes of he shor-erm ineres rae, he corporae bond yield, and he defaul spread. In conras, growh socks are more sensiive o changes of long-erm ineres raes and he erm spread. To he bes of our knowledge, his is he firs paper o examine he diverging ineres rae sensiiviies of value and growh socks in he conex of real esae. Furhermore, his is he firs paper o address ineres rae sensiiviies in he conex of NAVs in a global seing. The remainder of his paper is organized as follows. Secion 2 reviews he relaed lieraure, and inroduces our hypoheses. The daa is described in Secion 3. Secion 4 provides mehodology and Secion 5 he empirical resuls. Secion 6 provides he discussion of resuls and Secion 7 concludes our findings. 3
2 Relaed Lieraure and Hypoheses 2.1 Value Socks and Risk The raionale of he efficien marke hypoheses (EMH) of Fama (1970) is ha financial markes "a any ime 'fully reflec' all available informaion" (Fama, 1970) including such informaion as he inrinsic value of a lised company. Shiller (1981) conradics he EMH finding ha a subsanial proporion of sock volailiy is unexplained by changes of fundamenal informaion (e.g. fuure dividends). The capial asse pricing model (CAPM) of Sharpe (1964) and Linner (1965) fails o describe such reurn anomalies. These anomalies include i.a. ha he marke porfolio does no enirely explain he relevan risk in he economy o expeced reurns (Lewellen 1999) such as overreacions o new financial informaion (De Bond and Thaler 1985). Anoher reurn anomaly goes back o he work of Rosenberg e al. (1985) and Fama and French (2012), who find ha socks wih high book-o-marke raios of equiy have higher reurns han hose wih low raios (he value premium). Fama and French (1992) address his shorcoming by exending he CAPM by he wo addiional risk facors size and book-omarke. They provide evidence ha he hree-facor model has increasing explanaory power and explains risk in expeced reurns more precisely. Regarding he value premium, lieraure exhibis wo key explanaory approaches: Firs, riskbased explanaions (e.g. Davis e al 2000, Zhang 2005, Liew and Vassalou 2000) wih he assumpion ha unsysemaic sock-specific fundamenals (e.g. leverage, size) are causing he average ouperformance of value socks. The explanaion aemp refers o unsysemaic risk facors, which are non-diversifiable. Second, behavioral based explanaions, which imply he reurn anomaly is due o subopimal invesor behavior (e.g. Lakonishok e al., 1994; De Bond and Thaler, 1985). A furher explanaory approach includes risk-based explanaions regarding sysemaic risk: Macroeconomic facors. The raionale behind his approach is ha value socks are paricularly prone o macroeconomic facors and hus produce a risk premium. Lewellen (1999) argues ha value socks are paricularly sensiive o changing macroeconomic facors owing o he "disress facor" suggesed by Fama and French (1993). Jensen and Mercer (2002) provide evidence ha he moneary policy is an imporan addiional facor in explaining he risk premia of he hreefacor model. Hahn and Lee (2006) exend he hree-facor model of Fama and French (1993) by wo addiional macroeconomic variables, based on he proposiion ha he long-esablished facors marke, size and book-o-marke do no fully proxy sysemaic risk and business cycle flucuaions. The wo addiional facors are he defaul spread and he erm spread. These yield spreads are commonly used as proxies for credi marke and moneary policy condiions. Hahn & Lee (2006) provide evidence ha value socks have higher (posiive) loadings on posiive changes of he erm spread han on growh socks. Oher sudies provide evidence ha value socks are relaed o oher macroeconomic sae variables: E.g. consumpion growh (Kang e al., 2011) or marke wide flucuaions in expeced cash flows (Da and Warachka, 2009). 4
2.2 Ineres Rae Sensiiviy of Sock Reurns This secion will give a brief review of relevan sudies in he conex of he ineres rae sensiiviy of sock reurns. The analysis of he ineres rae sensiiviy of sock reurns has been subjec of numerous sudies in he pas. Sone (1974) as well as Lloyd and Shick (1977) were he firs analyzing he ineres sensiiviy of sock reurns employing a wo-index version of he CAPM (marke and ineres rae erms). Fama and Schwer (1977) demonsrae ha monhly changes of shor-erm ineres raes have a negaive coefficien for fuure reurns of commons socks. Several oher sudies follow a similar mehodological approach, concenraing on financial insiuions. These sudies include iner alia Chance and Lane (1980), Lynge and Zumwal (1980), Flannery and James (1984) or Bae (1990). Elyasiani and Mansur (1998) follow a ime series approach employing a GARCH-M model o analyze he ineres rae sensiiviy of bank sock reurns. 2.3 Ineres Rae Sensiiviy of Lised Real Esae Companies Beside financial insiuions, a subsanial amoun of sudies documened he ineres rae sensiiviy of lised real esae companies (REITs and REOCs). Chen and Tzang (1988) as well as Allen e al (2000) find ha US-REITs are sensiive o changes of long-erm ineres raes and shor-erm ineres raes in pars of he 1980's and 1990's. Consisen wih hese findings, Devaney (2001) repors a highly significan and negaive coefficien for monhly changes of long-erm ineres raes in explaining he excess reurns of US-REITs beween 1978-1998. According o Devaney (2001), morgage REITs (MREITs) have a higher ineres rae sensiiviy han equiy REITs (EREITs). He e al (2003) repor similar resuls, i.e. ha MREITs are sensiive o changes o all of he seven incorporaed ineres rae proxies, while EREITs are only sensiive o changes of long-erm raes and corporae bond yields. To he conrary, Liang e al (2009) find no significan ineres rae risk facor for equiy REITs. As wih He e al (2003), Swanson e al (2002) and use a defaul and erm spread as ineres rae proxies. Their empirical resuls reveal ha REIT reurns are more sensiive o changes of he erm spread han o he defaul spread. In conras o He e al (2003), hey do no find diverging resuls for MREITs and EREITs. The majoriy of he reviewed sudies so far, are limied o U.S. daa. The paper of Akimov e al (2015) is one he few sudies analyzing global lised real esae markes. However, hey are using index level daa insead of more precise panel daa. Akimov e al (2015) demonsrae he imporance of ineres rae risk for lised real esae companies. In line wih he majoriy of previous research, hey find ha shor-erm and long-erm ineres raes are significan risk facors in explaining he reurns of lised real esae. Lizieri e al (1997) confirm he resuls of previous research as hey find an asymmeric effec of he sensiiviy of propery company share prices o ineres rae changes in he U.S. and U.K.. Amending previous research, hey hypohesize ha lised real esae companies are affeced by ineres rae changes on wo furher levels han merely he sock marke. 1. The "underlying direc [real esae] marke" level which is represened by ne asse value (NAV), appraised on a discouned cash flow basis. As ineres raes rise, he capial values of individual properies are depressed. 2. The corporae level of real esae companies is characerized by high leverage and decreasing profis as coss of borrowing increase when ineres raes rise. 5
To sum up, mos of he sudies have in common ha heir resuls hold rue for 1) REITs, 2) seleced coninenal markes like he U.S., 3) index level daa and 4) oudaed sample periods. We couner hese drawbacks wih a rich panel daa se focusing on REITs and REOCs in 24 counries wih a conemporary sample period (2000-2014). The ineres rae proxies employed in previous sudies can be caegorized ino hree main caegories: 1) Shor- and long-erm ineres raes represened by -bill raes and governmen bond yields wih diverse mauriies (e.g. 10 o 15 years), 2) Corporae bond yields, and 3) yield spreads (e.g. defaul and erm spread). The sudies have in common ha he selecion of an ineres raes proxy is in mos cases inconsisen. Following Akimov e al (2015) he raionale behind he proxy selecion is in some way random and he proxies canno be incorporaed ino a model simulaneously. To address his issue, we consider he enire se of ineres rae proxies in our sudy. Moreover, we make use of he defaul and erm spread as his allows o simulaneously esing he effec of more han one ineres rae proxy in a single model. 2.4 Ineres Rae Sensiiviy of Value and Growh Socks Thus far, only few papers disinguish beween he ineres rae sensiiviy of value or growh socks. Subsanial seleced sudies include Hahn & Lee (2006), Lioui and Maio (2014) and Jensen and Mercer (2002). Their approaches and findings will be discussed in he following and shape he basis o formulae our hypoheses regarding he sensiiviy o changes of five ineres rae proxies. Shor-erm Ineres Raes In a recen sudy, Lioui and Maio (2014) employ a macroeconomic asse pricing model and find ha value socks have a sronger ineres rae risk han growh socks. They conclude ha ineres rae risk is a key facor in explaining he value premium. In heir empirical analysis, hey find ha value socks load negaively on he moneary facor, represened by he shorerm ineres rae 1 and he effecive federal funds rae as ineres rae proxies. Lioui and Maio (2014) hypohesize ha value socks are more sensiive o unexpeced decreases of shor-erm ineres raes. They propose ha value socks face coninuing underperformance for years, which is likely o induce negaive shocks in heir cash flows making hem "financially consrained hrough ime". According o Bernanke and Gerler (1995) companies under disress are especially sensiive since increasing ineres raes direcly reduce cash flows as deb expenses rise. We argue ha anoher key subjec in he conex of he reurn sensiiviy of differen ineres rae proxies is he concep of relaive araciveness amongs asse classes. Invesors, who are willing o buy shor-erm bonds, migh pursue a shor-erm invesmen horizon. Due o heir larger price-o-earnings raios, value socks have higher dividend yields. When shor-erm ineres raes fall, shor-erm invesors migh reallocae heir funds o value socks since hey generae higher (dividend) yield income in he shor run. We hus formulae our firs hypohesis regarding he sensiiviy o changes of shor-erm ineres raes as follows: 1 I.e. 3-monh T-bill rae 6
Hypohesis 1: The risk-adjused reurns of value socks are more sensiive o changes of shor erm ineres raes han he risk-adjused reurns of growh socks. Long-erm Ineres Raes and he Term Spread According o Campell and Viceria (2001), long-erm bonds are held by risk-averse invesors wih a long-erm invesmen horizon seeking sable cash flows and a erm premium over shorerm bonds. REITs have long been praised as a bond-like invesmen, due o heir high cash flow sabiliy. The research quesion which we seek o answer in his paper is he following: Are value REITs or growh REITs more sensiive o changes in he long erm rae? Changes in long-erm ineres raes end o be accompanied by changes in fuure expecaions. In paricular, growh socks are valued based on fuure cash flow expecaions. Increasing longerm ineres raes resul in higher discoun raes (Thorbeke 1997). Thus, fuure cash flows are discouned a higher raes, which over-proporionally affecs he marke values of growh socks. Hence, he reurns of growh socks should be more sensiive o changes in he long erm ineres rae, han he reurns of value socks. We formulae our second hypohesis accordingly: Hypohesis 2: The risk-adjused reurns of growh socks are more sensiive o changes of long erm ineres raes han growh socks. Similarly, a widening erm spread, i.e. he difference beween long-erm and shor-erm ineres raes, increases he relaive araciveness of value socks over growh socks. Hence, growh socks should also be more sensiive o changes of he erm spread han growh socks. Corporae Bond Yields and he Defaul Spread Corporae bonds represen one imporan form of deb financing for real esae companies. He e al. (2003) find ha changes of high-yield corporae bonds have he sronges explanaory power in explaining he reurns of U.S. REITs in conras o oher ineres rae proxies. Increasing corporae bond yields cause an increase of he cos of deb and hus have a negaive impac on he corporae performance (corporae level). Hahn and Lee (2006) argue ha value socks end o be higher leveraged han growh socks. Thus, increasing corporae bond yields should lead o negaive reurns as he cos of capial increases (a similar argumen is made by Bernanke and Gerler, 1995). Thus, we formulae our hird hypohesis as follows: Hypohesis 3: The risk-adjused reurns of value socks are more sensiive o changes in corporae bond yields han growh socsk. Relaed o he corporae bond yield is he defaul spread, which is defined as he difference beween he corporae bond yield and he long erm ineres rae. Fama and French (1989) argue ha he defaul spread is an indicaor for long-erm business condiions and associaed wih high expeced reurns near business cycle buss, and low expeced reurns near booms. Hence, value socks should also be more sensiive o changes in he defaul spread han growh socks. 7
3 Daa and Descripive Saisics 3.1 Sample Descripion Our sample is based on he FTSE EPRA/NAREIT Global Real Esae Index, which is comprised of lised companies wih "relevan real esae aciviies." Four ground rules regarding he consiuen underlying REOCs and REITs ensure sufficien index qualiy: 1) a minimum free-floa marke capializaion, 2) minimum liquidiy requiremens, 3) a minimum share of EBITDA (>75%) from relevan real esae aciviies 2, 4) publicaion of audied annual accouning reporing in English. 3 The sample period for he analysis is 2000:03 o 2014:05. To avoid survivorship bias, we consider hisoric changes of he index consiuen composiion in every monh of he period. Our final sample consiss of 487 socks from 24 counries including 345 REITs and 142 REOCs. The advanages of panel daa are iner alia increasing degrees of freedom, weakening of mulicollineariy, consrucion of more realisic behavioral models and obaining more precise esimaes of micro relaions (Hsiao 2014). 3.2 Consrucion of value and growh sock porfolios In order o consruc he value and growh sock porfolios we sor socks according o heir price deviaion from NAV. In his regard, he NAV per share (or he book value of equiy) is calculaed by dividing Daasream's "common equiy" by "number of shares." The discoun o NAV is calculaed based on he "unadjused share price" as repored by Daasream. As socks may also rade a a premium o NAV, we name our soring crieria NAV spread:, =,, (1) The major shorcoming of consrucing he global value porfolio on he (absolue) NAV spread is ha he global value porfolio can be overly exposed o counry risk. For example, if a counry is rading a depressed levels relaive o oher counries, he global value porfolio may sill include growh socks of he discoun counry. Thus, he inerpreaion of he resuls may be ambiguous. To avoid his shorcoming, we sor socks according o he relaive NAV discoun of sock i wih respec o he average NAV discoun of counry j in a given monh :,, =,,, (2) We sor he sample based on monh-end daa and consruc hree ranking porfolios. Then we observe he oal reurns of he porfolios as repored by Daasream over he following monh. The quinile wih he highes discoun o NAV forms he value porfolio (P1), he middle porfolio (P2) and he quinile of socks wih he highes NAV premiums he growh porfolio (P3). All porfolios are equally weighed. We do no consider value-weighed reurns as our sample size is relaively small, and value-weighing would pu non-essenial emphasis on he performance of individual socks. To ensure ha he resuls are no biased by exchange rae flucuaions, all reurns are denominaed in local currencies. Noe, ha in conras o he majoriy exising asse pricing sudies, we follow a monhly soring procedure, based on 2 Which is defined as "he ownership, rading and developmen of income-producing real esae 3 hp://www.epra.com/research-and-indices/indices/ 8
Daasream's "Earnings per share repor dae (EPS)." We can hus ensure ha financial reporing daa are acually published as new porfolios are formed. For example, if he annual repor for calendar year 2014 is published in April 2015, Daasream will repor a new book value of equiy from December 2014 onward, bu we can shif his informaion by four monhs using he "Earnings per share repor dae." Financial reporing frequency is generally semiannual and may even be quarerly. Thus, NAVs may only change semiannually, bu we observe monhly changes in he book-o-marke raios due o share price flucuaions. 3.3 Ineres Rae Proxies Our panel analysis approach allows o consider ineres rae sensiiviies on individual sock level. Accordingly, he five ineres rae proxies are derived for each of he 24 counries in every monh of our panel in he 2000:03 o 2014:05 period. Wih regard o he selecion of appropriae proxies we follow previous research on ineres rae sensiiviies (e.g. He e al 2003, Hahn and Lee 2006 or Allen e al 2000, Jensen and Mercer 2002). STIR is represened by he 1-year deposi rae in each individual counry, LTIR by he 10-year governmen bond yield, CBY by he redempion yield of qualiy (invesmen grade) corporae bonds; MPR is represened by he base ineres rae of a counry's associaed cenral bank. Following Hahn and Lee 2006 and He e al 2003, he defaul spread (DEF) and erm spread (TERM) of counry j in monh are derived as follows:, =,, (3), =,, (4) The sources of he ineres rae proxies are Daasream, Morningsar and publicly accessible daabases like FRED (Federal Reserve Economic Daa) of he S. Louis FED or he Saisical Daa Warehouse of he European Cenral Bank. 3.4 Summary Saisics Table 1 conains some summary saisics on reurns and (relaive) NAV Spreads for our global sample over he 2000:03 o 2014:05 period. The able includes subpanels for he saisics of he hree porfolios value, middle and growh (Panel A-C). Panel D includes he summary saisics for he oal sample and he five ineres rae proxies. On average monhly reurn of value socks (1.44%) is noably higher han he average reurn of growh socks (0.80%) indicaing a value premium. However, he sandard deviaion reveals ha value socks are riskier han growh socks, which is in line wih previous research (e.g. Rosenberg e al. 1985). On average, he oal sample performed on average by 1.07% per monh (13.60% p.a.). The oal sample raded a an average discoun o relaive NAV of -0.03. The summary saisics of he five ineres rae proxies are in line wih economic inuiion. On average, long-erm ineres raes are higher han shor-erm raes by 0.08% per monh. Alhough, long-erm raes have he leas risk as measured by monhly volailiy. Corporae bonds ouperform boh, he shor and he long-erm ineres rae, however he corporae bond yield is also associaed wih he highes risk. Table 2 conains he conemporaneous correlaion coefficiens of reurns, relaive NAV Spreads and he five ineres rae proxies. 9
4 Mehodology: Modelling he Ineres Rae Sensiiviy of Value and Growh Socks To deermine he ineres rae sensiiviy of he reurns of value and growh socks, we run he following regression model for he hree porfolios, which are consruced according heir relaive NAV spread. In order o conrol for differen behavior of ineres rae changes on he hree porfolios we follow Jensen and Mercer (2002) and include hree ineracion erms: R RF ( D _ Value 6 IR i 1 * IR [ RM ) ( D _ Mid 7 2 RF * IR ] SMB 3 HML ) ( D _ Growh 8 4 WML 5 * IR ) (5) where R RF is he oal reurn of he global value, middle, or growh porfolio in monh in i i excess of he one-monh risk-free rae. IRi in monh, STIR, LTIR, CBY, DEF, or TER. risk-free rae; momenum facor. SMB i, is he size facor; is he firs difference of he respecive ineres rae RM RF, i is he marke reurn in excess of he HML i, is he book-o-marke facor and WML i,, he D _ Value, D _ Mid, D _ Growh represen dummy variables aking he value 1 if a sock is associaed o in he respecive porfolio in monh. ( D _ Value* IR i, ) is he ineracion erm for he value porfolio and he respecive ineres rae proxy. We obain he four risk facors from Kenneh French's websie. 4 French's daa library provides regional facors in USD for "Asia Pacific ex Japan," "Europe," "Japan," and "Norh America," so we conver he regional USD reurns ino local currency reurns for he respecive counries. RM, SMB, HML and WML are no limied o he subsecor of lised real esae. We do so o reflec he original raionale of he Carhar four-facor model, which implies ha he risk facors are markewide and are no indusry-specific proxies for no diversifiable facor risk. As we follow an inernaional approach, i seems sraighforward o use global RM, SMB, HML and WML facors. To es Hypoheses 1-3 we also direcly conrol differences in regarding he ineres rae sensiiviy of value and growh socks by reducing he enire sample o value and growh socks and performing he following panel regression model: R RF ( D _ Value 6 IR i 1 * IR ) [ RM 2 RF ] SMB 3 HML 4 WML 5 (6) The sign and significance of he coefficien 6 in equaion (6) indicaes wheher value socks are more or less sensiive han growh sock o changes of he five ineres rae proxies. We use panel regressions wih fixed effecs o empirically es our hypoheses. 4 hp://mba.uck.darmouh.edu/pages/faculy/ken.french/daalibrary.hml 10
5 Empirical Resuls Tables 3 o 7 conain he regression resuls for our five ineres rae proxies (STIR, LTIR, CBY, DEF, and TERM), which are used o es our Hypoheses 1 o 3. The ables are srucured as follows: Model (1) is he base model, which esimaes he general impac of he respecive ineres rae proxy. The following hree models exend he base model by ineracion erms for he value (model 2), middle (model 3), and growh porfolio (model 4). Model (5) simulaneously includes ineracion erms for all hree porfolios. Finally, model (6), direcly es beween differences in he ineres rae sensiiviy of value and growh socks, by excluding he middle porfolio. Hence, our empirical evidence is based on he ineracion erm beween he respecive ineres rae proxy and he value indicaor variable. Table 3 conains he resuls for shor-erm ineres raes (STIR). Model (1) shows, as expeced, ha he reurns of real esae socks are negaively relaed o changes in he shor erm ineres rae in general. In model (2) he coefficien for he value porfolio ineracion erm is negaive and significan a he 1% level. This resul indicaes ha value socks are more sensiive and negaively relaed o changes of STIR han socks being in he middle and growh porfolio. Afer including he hree porfolio ineracion erms and he referred dummy variables, he resuls of he aggregae model (5) reveals ha value socks are associaed wih a more negaive coefficien (-5.38) han growh socks (-3.37). To which exen are value socks more sensiive o an increase in STIR? The regression resuls in model (6) are based on a reduced sample, which merely consiss of socks in he value and growh porfolio. Thus, he coefficien for he ineracion erm of he value porfolio reveals he reurn difference beween value and growh. For he ineracion erm beween value and STIR he coefficien is -2.24 and significan a he 1% level. Tha is, in he even of an increase of STIR by 100 basis poins, he decrease of reurn for value socks is on average by -2.24 pps larger han for growh socks (ceeris paribus). In summary, he Table 3 resuls are consisen wih Hypohesis 2, i.e. he risk-adjused reurns of value socks are more sensiive o changes of he shor erm raes han growh socks. Table 4 conains he regression resuls for long-erm ineres raes (LTIR). The relaed Hypohesis 2 saes ha he risk-adjused reurns of growh socks are more sensiive o longerm ineres raes changes, han hose of value socks. The regression resuls shown in Table 4 differ considerably from hose in Table 3, which is consisen wih hypoheses 1 and 2, which predic diverging ineres rae sensiiviies for value and growh socks depending on he chosen ineres rae. In model (2) he coefficien for he value porfolio ineracion erm is posiive and significan a he 1% level. In conras, models (3) and (4) reveal ha he middle and growh porfolio are more sensiive o changes in he long erm rae, i.e. when he long erm rae increases, he reurns of hese socks end o fall more han hose of value socks. The resuls shown in model (6) are consisen wih hypohesis 2. The ineracion erm beween value and LTIR is posiive (3.04) and significan a he 1% level. Tha is, in he even of an increase of LTIR by 100 basis poins, he decrease of reurn for value socks is on average by 3.04 pps smaller han for growh socks (ceeris paribus). 11
Table 5 repors he resuls for changes of he erm spread (TERM). Overall, he resuls are in line wih he Table 4 resuls. Value socks are associaed wih a posiive coefficien (2.07) while he coefficien for growh socks is negaive (-1.57). This resul is in line wih Hahn and Lee (2006) who repor a (posiive) loading for value socks o changes of he erm spread. Model (6) shows ha he coefficien of he ineracion erm beween value and TERM is posiive (3.54) and significan a he 1% level. Tha is, in he even of an increase of TERM by 100 basis poins, he decrease of reurn for value socks is on average by 3.45 pps smaller han for growh socks (ceeris paribus). Table 6 conains he regression resuls for he corporae bond yield. The comparison of he marginal ineres rae sensiiviies in models (2) o (4) suggess ha value socks suffer he mos when he corporae bond yield increases. This resul is suppored by model (6). The ineracion erm of value and CBY in model (6), reveals ha he difference in reurn sensiiviies beween value and growh is -3.54 and significan a he 1% level. Tha is, in he even of an increase of CBY by 100 basis poins, he decrease of reurn for value socks is on average by -3.54 pps larger han for growh socks (ceeris paribus). This finding is consisen wih hypohesis 3 and may be explained by he fac ha value socks end o be higher leveraged han growh socks and hus more prone o increasing cos for bond financing. Table 7 conains he resuls for defaul spread (DEF) which are similar o CBY. However, resuls of he model (6) regression reveal ha he reurn difference for changes of DEF is even larger (-4.19) and significan a he 1% level han for CBY. Hahn and Lee (2006) argue ha an increasing defaul spread (DEF) is commonly inerpreed as an indicaor for "he marke's expecaion of worsening credi marke condiions". Thus, he resuls confirm our Hypohesis 3 ha increasing corporae bond yields and defaul spread cause an increase of he cos of deb. Thus, he increase has a sronger negaive impac on he corporae performance (corporae level) and as a resul he reurns of value socks. 6 Conclusion The aim of his sudy was o examine he diverging ineres rae sensiiviies of value and growh socks. Using a global sample of real esae socks and five ineres rae proxies, we provide new insighs ino he relaionship beween ineres rae changes and he reurns of socks wih fundamenally differen characerisics. In paricular, he following resuls sand ou: Firs, value socks are more sensiive o changes of shor erm ineres raes. Due o heir low raio of price-o-fundamenal value, value socks promise higher iniial yields han growh socks. When shor erm ineres raes rise, income-oriened invesors end o remove heir funds from risky asses and reinves in he meanwhile higher-yielding risk-free rae. Second, growh socks are more sensiive o changes in he long erm rae. This is consisen wih he fuure cash flows of growh REITs being discouned a a higher rae. In conras, he more fron-loaded cash flows of value REITs are less srongly affeced by higher discoun raes. 12
Third, value socks are more sensiive o changes in he corporae bond yield. Credi coss have a direc impac on a firm s cos of capial. Since value socks end o use more leverage, hey are also more han proporionally affeced by higher bond raes compared o growh sock. Furhermore, our resuls suppor he "macroeconomic risk sory", which saes he value premium anomaly is relaed o value socks having larger ineres rae risk han growh socks (Lioui and Maio, 2014). 13
7 References Akimov, A., Sevenson, S., Zagonov, M., 2015. Public Real Esae and he Term Srucure of Ineres Raes: A Cross-Counry Sudy. The Journal of Real Esae Finance and Economics 51, 503 540. doi:10.1007/s11146-014-9492-x Allen, M.T., Madura, J., Springer, T.M., 2000. REIT Characerisics and he Sensiiviy of REIT Reurns. The Journal of Real Esae Finance and Economics 21, 141 152. Asness, C.S., Moskowiz, T.J., Pedersen, L.H., 2013. Value and Momenum Everywhere. The Journal of Finance 68, 929 985. doi:10.1111/jofi.12021 Bae, S.C., 1990. Ineres Rae Changes and Common Sock Reurns of Financial Insiuions: Revisied. Journal of Financial Research 13, 71 79. doi:10.1111/j.1475-6803.1990.b00537.x Barkham, R., Ward, C., 2009. Invesor Senimen and Noise Traders: Discoun o Ne Asse Value in Lised Propery Companies in he U.K. Journal of Real Esae Research. Bernanke, B.S., Gerler, M., 1995. Inside he black box: he credi channel of moneary policy ransmission. Naional bureau of economic research. Brounen, D., Laak, M., 2005. Undersanding he Discoun: Evidence from European Propery Shares. Journal of Real Esae Porfolio Managemen 11, 241 251. Campbell, J.Y., Viceira, L.M., 2001. Who Should Buy Long-Term Bonds? The American Economic Review 91, 99 127. Carhar, M.M., 1997. On persisence in muual fund performance. The Journal of finance 52, 57 82. Chance, D.M., Lane, W.R., 1980. A Re-Examinaion of Ineres Rae Sensiiviy in he Common Socks of Financial Insiuions. Journal of Financial Research 3, 49 55. doi:10.1111/j.1475-6803.1980.b00036.x Chen, K., Tzang, D., 2009. Ineres-rae sensiiviy of real esae invesmen russ. Journal of Real Esae Research. Da, Z., Warachka, M.C., 2009. Cashflow risk, sysemaic earnings revisions, and he crosssecion of sock reurns. Journal of Financial Economics 94, 448 468. doi:10.1016/j.jfineco.2008.12.008 Davis, J.L., Fama, E.F., French, K.R., 2000. Characerisics, Covariances, and Average Reurns: 1929 o 1997. The Journal of Finance 55, 389 406. doi:10.1111/0022-1082.00209 De BONDT, W.F.M., Thaler, R., 1985. Does he Sock Marke Overreac? The Journal of Finance 40, 793 805. doi:10.1111/j.1540-6261.1985.b05004.x Devaney, M., 2001. Time varying risk premia for real esae invesmen russ: A GARCH-M model. The Quarerly Review of Economics and Finance 41, 335 346. Elyasian E., Mansur, I., 1998. Sensiiviy of he bank sock reurns disribuion o changes in he level and volailiy of ineres rae: A GARCH-M model. Journal of Banking & Finance 22, 535 563. doi:10.1016/s0378-4266(98)00003-x Fama, E.F., 1970. Efficien Capial Markes: A Review of Theory and Empirical Work. The Journal of Finance 25, 383. doi:10.2307/2325486 Fama, E.F., French, K.R., 2012. Size, value, and momenum in inernaional sock reurns. Journal of Financial Economics 105, 457 472. doi:10.1016/j.jfineco.2012.05.011 Fama, E.F., French, K.R., 1993. Common risk facors in he reurns on socks and bonds. Journal of Financial Economics 33, 3 56. Fama, E.F., French, K.R., 1992. The Cross-Secion of Expeced Sock Reurns. The Journal of Finance 47, 427 465. doi:10.1111/j.1540-6261.1992.b04398.x Fama, E.F., French, K.R., 1989. Business condiions and expeced reurns on socks and bonds. Journal of financial economics 25, 23 49. Fama, E.F., Schwer, G.W., 1977. Asse reurns and inflaion. Journal of Financial Economics 5, 115 146. doi:10.1016/0304-405x(77)90014-9 14
Flannery, M.J., James, C.M., 1984. The effec of ineres rae changes on he common sock reurns of financial insiuions. The Journal of Finance 39, 1141 1153. Hahn, J., Lee, H., 2006. Yield spreads as alernaive risk facors for size and book-o-marke. Journal of Financial and Quaniaive Analysis 41, 245 269. He, L.T., Webb, J.R., Myer, F.N., ohers, 2003. Ineres rae sensiiviies of REIT reurns. Inernaional Real Esae Review 6, 1 21. Hsiao, C., 2014. Analysis of Panel Daa. Cambridge Universiy Press. Jensen, G.R., Mercer, J.M., 2002. Moneary Policy and he Cross-Secion of Expeced Sock Reurns. Journal of Financial Research 25, 125 139. Kang, J., Kim, T.S., Lee, C., Min, B.-K., 2011. Macroeconomic risk and he cross-secion of sock reurns. Journal of Banking & Finance 35, 3158 3173. doi:10.1016/j.jbankfin.2011.04.012 Lakonishok, J., Shleifer, A., Vishny, R.W., 1994. Conrarian Invesmen, Exrapolaion, and Risk. The Journal of Finance 49, 1541 1578. doi:10.1111/j.1540-6261.1994.b04772.x Lee, C.M.C., Myers, J., Swaminahan, B., 1999. Wha is he Inrinsic Value of he Dow? The Journal of Finance 54, 1693 1741. doi:10.1111/0022-1082.00164 Lewellen, J., 1999. The ime-series relaions among expeced reurn, risk, and book-omarke. Journal of Financial Economics 54, 5 43. Liang, Y., Prudenial, W., Webb, J., 2009. Ineremporal changes in he riskiness of REITs. Journal of Real Esae Research. Liew, J., Vassalou, M., 2000. Can book-o-marke, size and momenum be risk facors ha predic economic growh? Journal of Financial Economics 57, 221 245. Linner, J., 1965. The Valuaion of Risk Asses and he Selecion of Risky Invesmens in Sock Porfolios and Capial Budges. The Review of Economics and Saisics 47, 13 37. doi:10.2307/1924119 Liou A., Maio, P., 2014. Ineres Rae Risk and he Cross Secion of Sock Reurns. Journal of Financial and Quaniaive Analysis 49, 483 511. doi:10.1017/s0022109014000131 Lizier C., Sachell, S., 1997. Propery company performance and real ineres raes: a regime-swiching approach. Journal of Propery Research 14, 85 97. doi:10.1080/095999197368654 Lizier C., Sachell, S., Worzala, E., ohers, 2009. Real ineres regimes and real esae performance: a comparison of UK and US markes. Journal of Real Esae Research. Lloyd, W.P., Shick, R.A., 1977. A Tes of Sone s Two-Index Model of Reurns. The Journal of Financial and Quaniaive Analysis 12, 363. doi:10.2307/2330540 Lynge, M.J., Zumwal, J.K., 1980. An Empirical Sudy of he Ineres Rae Sensiiviy of Commercial Bank Reurns: A Muli-Index Approach. The Journal of Financial and Quaniaive Analysis 15, 731. doi:10.2307/2330406 Meron, R.C., 1973. An Ineremporal Capial Asse Pricing Model. Economerica 41, 867. doi:10.2307/1913811 Rosenberg, B., Reid, K., Lansein, R., 1985. Persuasive evidence of marke inefficiency. Porfolio Managemen 11, 9 16. doi:10.3905/jpm.1985.409007 Sharpe, W.F., 1964. Capial asse prices: A heory of marke equilibrium under condiions of risk*. The journal of finance 19, 425 442. Shiller, R.J., 1981. Do Sock Prices Move Too Much o be Jusified by Subsequen Changes in Dividends? The American Economic Review 71, 421 436. Sone, B.K., 1974. Sysemaic Ineres-Rae Risk in a Two-Index Model of Reurns. The Journal of Financial and Quaniaive Analysis 9, 709. doi:10.2307/2329656 Swanson, Z., Theis, J., Casey, K.M., 2002. REIT risk premium sensiiviy and ineres raes. The Journal of Real Esae Finance and Economics 24, 319 330. 15
Thorbecke, W., 1997. On sock marke reurns and moneary policy. The Journal of Finance 52, 635 654. Zhang, L., 2005. The Value Premium. The Journal of Finance 60, 67 103. doi:10.1111/j.1540-6261.2005.00725.x 16
Tables Table 1: Summary Saisics of Value, Middle and Growh Porfolios Mean Sd. Deviaion Min Max Panel A: Value Porfolio Toal Reurn 1.44 11.74-79.80 236.42 Rel. NAV Spread -2.80 7.69-72.36 4.73 Panel B: Middle Porfolio Toal Reurn 1.04 9.68-97.90 343.07 Rel. NAV Spread -0.40 1.75-54.80 11.46 Panel C: Growh Porfolio Toal Reurn 0.80 8.84-60.50 65.75 Rel. NAV Spread 3.65 40.69-54.72 1773.61 Panel D: Toal Sample Toal Reurn 1.07 9.96-97.90 343.07 Rel. NAV Spread -0.03 18.96-72.36 1773.61 STIR 0.21 0.15-0.00 0.74 LTIR 0.29 0.11 0.04 1.29 CBY 0.41 0.19 0.04 1.97 DEF 0.13 0.17-1.08 1.73 TERM 0.08 0.10-0.25 1.15 This able conains he summary saisics of oal reurns, relaive NAV spreads and ineres rae proxies for he global sample of lised real esae socks in he 2000:03 o 2014:05 period. All saisics are in monhly frequency and %. Panel A conains he daa for he sample of value socks; Panel B he sample of he middle porfolio and Panel C he sample of growh socks. 17
Table 2: Correlaion of reurns, relaive NAV spreads and ineres rae proxies TR Rel. NAVS STIR LTIR CBY DEF TERM Panel A: Conemporaneous correlaions TR 1.00 Rel. NAVS 0.00 1.00 Spread_ STIR -0.08 *** -0.00 1.00 LTIR -0.03 *** -0.00 0.78 *** 1.00 CBY -0.06 *** -0.00 0.43 *** 0.45 *** 1.00 DEF -0.05 *** -0.00-0.01-0.11 *** 0.84 *** 1.00 TERM 0.09 *** 0.00-0.69 *** -0.09 *** -0.16 *** -0.12 *** 1.00 Panel B: Lagged correlaions Toal Reurn_-1 0.04 *** 0.01-0.06 *** 0.00-0.09 *** -0.10 *** 0.10 *** Rel. NAV 0.00 0.84 *** -0.00-0.00-0.00-0.00-0.00 Spread_-1 STIR_-1-0.08 *** 0.00 0.99 *** 0.77 *** 0.44 *** 0.02 *** -0.69 *** LTIR_-1-0.04 *** 0.00 0.79 *** 0.99 *** 0.45 *** -0.10 *** -0.11 *** CBY_-1-0.00 0.00 0.40 *** 0.44 *** 0.97 *** 0.81 *** -0.13 *** DEF_-1 0.02 *** 0.00-0.04 *** -0.12 *** 0.80 *** 0.96 *** -0.08 *** TERM_-1 0.09 *** 0.00-0.67 *** -0.10 *** -0.18 *** -0.15 *** 0.97 *** 18
Table 3: Shor-erm ineres rae (STIR) sensiiviy of value socks and growh socks (1) (2) (3) (4) (5) (6) d_stir_ -0.59 ** (-3.13) -0.18 (-0.85) -0.98 *** (-3.40) -0.73 *** (-3.48) 3.33 * (2.05) 0.15 (0.34) d_stir*d.value_ -1.87 *** (-4.13) d_stir*d.mid_ 0.67 (1.78) d_stir*d.growh_ 0.69 (1.46) D.Value(P1)_ 0.01 *** (4.41) D.Mid(P2)_ 0.00 * (2.08) D.Growh(P3)_ -0.01 *** (-7.19) -5.38 ** (-3.22) -3.64 * (-2.22) -3.37 * (-2.01) -0.00 (-0.32) -0.01 (-1.76) -0.02 *** (-3.95) -2.24 *** (-3.75) 0.01 *** (4.47) RM_ 0.90 *** (89.62) 0.90 *** (89.74) 0.90 *** (89.65) 0.90 *** (89.68) 0.90 *** (89.75) 0.99 *** (58.99) SMB_ -0.13 *** (-7.90) -0.13 *** (-7.91) -0.12 *** (-7.89) -0.13 *** (-8.03) -0.13 *** (-7.92) -0.03 (-1.17) HML_ 0.34 *** (21.06) 0.34 *** (20.91) 0.34 *** (21.11) 0.34 *** (21.13) 0.34 *** (20.79) 0.51 *** (19.14) WML_, -0.26 *** (-24.56) -0.26 *** (-24.39) -0.26 *** (-24.49) -0.26 *** (-24.61) -0.26 *** (-24.44) -0.30 *** (-18.04) Consan 0.00 *** (8.72) 0.00 *** (4.79) 0.00 ** (3.18) 0.01 *** (11.24) 0.01 ** (2.95) -0.00 * (-2.40) Observaions 35221 35221 35221 35221 35221 14520 Adjused R 2 0.231 0.232 0.231 0.232 0.233 0.238 This able conains he regression resuls in erms of he reurn sensiiviy of value and growh socks o monhly changes of shor-erm ineres raes (STIR). The dependen variable is he monhly oal reurn in excess of he risk-free rae of 487 global lised real esae socks in he 2000:03 o 2014:05 period. P1 represens he value porfolio, P2 he middle porfolio and P3 he growh porfolio consruced according o NAV spread in he previous monh. The ineres rae sensiiviy of value and growh socks is measured by ineracing he monhly changes of STIR wih he respecive dummy variable for each porfolio. Models (1) o (5) are esimaed based on he full sample while model (6) is esimaed based on a sample reduced o P1 and P3 in order o conrol for he direc relaionship beween value and growh socks. RM, SMB, HML and WML represen he four-facor-model conrol variables. The models are 19
Table 4: Long-erm ineres rae (LTIR) sensiiviy of value socks and growh socks (1) (2) (3) (4) (5) (6) d_ltir_ -2.72 *** (-12.96) -3.48 *** (-14.53) -1.88 *** (-6.30) -2.39 *** (-10.19) 2.45 (1.54) -4.03 *** (-8.97) d_ltir*d.value_ 3.11 *** (6.67) d_ltir*d.mid_ -1.62 *** (-4.00) d_ltir*d.growh_ -1.53 ** D.Value(P1)_ 0.01 *** (4.86) D.Mid(P2)_ 0.00 (1.87) (-3.08) D.Growh(P3)_ -0.01 *** (-7.35) -2.83 (-1.71) -5.94 *** (-3.66) -6.37 *** (-3.85) 0.00 (0.21) -0.01 (-1.40) -0.02 *** (-3.65) 3.04 *** (4.98) 0.01 *** (4.84) RM_ 0.93 *** (90.54) 0.93 *** (90.65) 0.93 *** (90.60) 0.93 *** (90.59) 0.93 *** (90.67) 1.01 *** (59.23) SMB_ -0.10 *** (-6.19) -0.10 *** (-6.49) -0.10 *** (-6.24) -0.10 *** (-6.33) -0.10 *** (-6.51) -0.01 (-0.27) HML_ 0.33 *** (20.66) 0.33 *** (20.56) 0.33 *** (20.63) 0.33 *** (20.73) 0.33 *** (20.43) 0.51 *** (19.31) WML_, -0.27 *** (-25.62) -0.27 *** (-25.41) -0.27 *** (-25.63) -0.27 *** (-25.62) -0.27 *** (-25.55) -0.31 *** (-18.74) Consan 0.00 *** 0.00 *** 0.00 ** 0.01 *** 0.01 * -0.00 ** (7.93) (3.84) (2.91) (10.67) (2.48) (-3.00) Observaions 35221 35221 35221 35221 35221 14520 Adjused R 2 0.234 0.236 0.235 0.236 0.237 0.241 This able conains he regression resuls in erms of he reurn sensiiviy of value and growh socks o monhly changes of long-erm ineres raes (LTIR). The dependen variable is he monhly oal reurn in excess of he riskfree rae of 487 global lised real esae socks in he 2000:03 o 2014:05 period. P1 represens he value porfolio, P2 he middle porfolio and P3 he growh porfolio consruced according o NAV spread in he previous monh. The ineres rae sensiiviy of value and growh socks is measured by ineracing he monhly changes of LTIR wih he respecive dummy variable for each porfolio. Models (1) o (5) are esimaed based on he full sample while model (6) is esimaed based on a sample reduced o P1 and P3 in order o conrol for he direc relaionship beween value and growh socks. RM, SMB, HML and WML represen he four-facor-model conrol variables. The models are esimaed using panel regressions wih effecs. saisics are in parenheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 20
Table 5: Term Spread (TERM) sensiiviy of value socks and growh socks (1) (2) (3) (4) (5) (6) d_term_ -1.18 *** (-7.29) -1.86 *** (-10.22) -0.49 * (-2.01) -0.87 *** (-4.77) -0.71 (-0.53) -2.52 *** (-7.18) d_term*d.value_ 3.22 *** (8.28) d_term*d.mid_ -1.24 *** (-3.87) d_term*d.growh_ -1.42 *** D.Value(P1)_ 0.01 *** (4.62) D.Mid(P2)_ 0.00 (1.95) (-3.66) D.Growh(P3)_ -0.01 *** (-7.19) 2.07 (1.49) -1.00 (-0.73) -1.57 (-1.13) 0.00 (0.08) -0.01 (-1.45) -0.02 *** (-3.66) 3.45 *** (6.91) 0.01 *** (4.58) RM_ 0.91 *** (89.97) 0.91 *** (90.23) 0.91 *** (90.03) 0.91 *** (90.05) 0.91 *** (90.23) 0.99 *** (59.15) SMB_ -0.11 *** (-6.77) -0.11 *** (-7.07) -0.11 *** (-6.85) -0.11 *** (-6.92) -0.12 *** (-7.27) -0.02 (-0.69) HML_ 0.35 *** (21.59) 0.34 *** (21.47) 0.35 *** (21.65) 0.35 *** (21.65) 0.35 *** (21.61) 0.52 *** (19.54) WML_, -0.27 *** (-25.44) -0.27 *** (-24.89) -0.27 *** (-25.29) -0.27 *** (-25.44) -0.27 *** (-24.90) -0.31 *** (-18.31) Consan 0.00 *** 0.00 *** 0.00 *** 0.01 *** 0.01 ** -0.00 * (8.82) (4.75) (3.34) (11.34) (2.63) (-2.42) Observaions 35221 35221 35221 35221 35221 14520 Adjused R 2 0.232 0.234 0.232 0.233 0.235 0.239 This able conains he regression resuls in erms of he reurn sensiiviy of value and growh socks o monhly changes of he Term Spread (TERM). The dependen variable is he monhly oal reurn in excess of he risk-free rae of 487 global lised real esae socks in he 2000:03 o 2014:05 period. P1 represens he value porfolio, P2 he middle porfolio and P3 he growh porfolio consruced according o NAV spread in he previous monh. The ineres rae sensiiviy of value and growh socks is measured by ineracing he monhly changes of TERM wih he respecive dummy variable for each porfolio. Models (1) o (5) are esimaed based on he full sample while model (6) is esimaed based on a sample reduced o P1 and P3 in order o conrol for he direc relaionship beween value and growh socks. RM, SMB, HML and WML represen he four-facor-model conrol variables. The models are esimaed using panel regressions wih effecs. saisics are in parenheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 21
Table 6: Corporae Bond Yield (CBY) sensiiviy of value socks and growh socks (1) (2) (3) (4) (5) (6) d_cby_ -1.81 *** (-23.99) -1.62 *** (-20.88) -1.68 *** (-12.44) -2.04 *** (-24.33) 0.51 (0.88) -0.76 *** (-4.62) d_cby*d.value_ -2.79 *** (-9.74) d_cby*d.mid_ -0.18 (-1.16) d_cby*d.growh_ 1.10 *** D.Value(P1)_ 0.01 *** (3.50) D.Mid(P2)_ 0.00 * (2.23) (6.20) D.Growh(P3)_ -0.01 *** (-7.21) -4.92 *** (-7.74) -2.38 *** (-4.11) -1.46 * (-2.46) -0.00 (-0.67) -0.01 (-1.82) -0.02 *** (-4.06) -3.54 *** (-10.70) 0.01 *** (3.93) RM_ 0.84 *** (81.68) 0.84 *** (81.82) 0.84 *** (81.69) 0.84 *** (81.83) 0.84 *** (81.81) 0.94 *** (55.36) SMB_ -0.10 *** (-6.63) -0.11 *** (-6.82) -0.10 *** (-6.57) -0.10 *** (-6.68) -0.11 *** (-6.87) -0.01 (-0.45) HML_ 0.37 *** (23.26) 0.36 *** (22.86) 0.37 *** (23.30) 0.37 *** (23.42) 0.36 *** (22.89) 0.52 *** (19.69) WML_, -0.25 *** (-23.26) -0.24 *** (-22.55) -0.25 *** (-23.31) -0.25 *** (-23.29) -0.24 *** (-22.63) -0.28 *** (-16.40) Consan 0.00 *** 0.00 *** 0.00 ** 0.01 *** 0.01 ** -0.00 * (8.41) (4.80) (2.90) (10.98) (3.03) (-2.22) Observaions 35221 35221 35221 35221 35221 14520 Adjused R 2 0.243 0.246 0.243 0.245 0.248 0.249 This able conains he regression resuls in erms of he reurn sensiiviy of value and growh socks o monhly changes of corporae bond yields (CBY). The dependen variable is he monhly oal reurn in excess of he riskfree rae of 487 global lised real esae socks in he 2000:03 o 2014:05 period. P1 represens he value porfolio, P2 he middle porfolio and P3 he growh porfolio consruced according o NAV spread in he previous monh. The ineres rae sensiiviy of value and growh socks is measured by ineracing he monhly changes of CBY wih he respecive dummy variable for each porfolio. Models (1) o (5) are esimaed based on he full sample while model (6) is esimaed based on a sample reduced o P1 and P3 in order o conrol for he direc relaionship beween value and growh socks. RM, SMB, HML and WML represen he four-facor-model conrol variables. The models are esimaed using panel regressions wih effecs. saisics are in parenheses, * p < 0.10, ** p < 0.05, *** p < 0.01. 22