REIT-Based Pure-Play Portfolios: The Case of Property Types and Geographic Locations. Eui-Hoe Kim. Bachelor in Architecture, 1997

Transcription

1 REIT-Based Pure-Play Portfolos: The Case of Property Types ad Geographc ocatos by Eu-Hoe Km Bachelor Archtecture, 997 Seoul Natoal Uversty, Korea Submtted to the Departmet of Archtecture Partal Fulfllmet of the Requremets for the Degree of aster of Scece Real Estate Developmet at the assachusetts Isttute of Techology September, Eu-Hoe Km All rghts reserved The author hereby grats to IT permsso to reproduce ad to dstrbute publcly paper ad electroc copes of ths thess documet whole or part. Sgature of Author Departmet of Urba Studes ad Plag August 6, 24 Certfed by Hery O. Pollakowsk Vstg Scholar, Ceter for Real Estate Thess Supervsor Accepted by Davd Gelter Charma, Iterdepartmetal Degree Program Real Estate Developmet

2 REIT-Based Pure-Play Portfolos: The Case of Property Types ad Geographc ocatos by Eu-Hoe Km Submtted to the Departmet of Urba Studes ad Plag O August 6, 24 Partal Fulfllmet of the Requremet for the Degree of aster of Scece Real Estate Developmet ABSTRACT Over the last 2 years the REIT dustry has made remarkable progress terms of the magtude of market captalzato ad maagemet performace. However, t s dffcult for vestors to precsely target ther vestmets because publcly traded REITs dversfy ther assets across property types ad geographc locatos. Ths research has eplored the cocept ad methodology of a REIT-based pure-play portfolo. Ths portfolo s defed as a optmal combato of log ad short postos publcly traded REITs, such that the portfolo ca replcate the vestmet returs to a specfed target sector wthout ay drect eposure to other sectors whle mmzg the dosycratc rsk compoet of the portfolo. Ths thess costructed pure-play portfolos across four property types ad four geographc locatos usg seve years of structural formato o publcly traded REITs. For comparso wth the prvately held property market, NCREIF sub-dces whch represet also four property types ad four geographc locatos are used. I addto to the costructo of pure-play portfolos, the performace dces of REIT-based pure-play portfolos ad prvate property-based NCREIF sub-dces are preseted. ` Thess Supervsor: Ttle: Hery O. Pollakowsk Research Assocate, Ceter for Real Estate 2

3 ACKNOWEDGEENTS I would lke to thak my advsor Hery Pollakowsk for hs costat advce ad support I would lke to thak Davd Gelter for hs commets ad suggestos I would lke to thak atthew Ostrower at orga Staley for hs support ad sght I would lke to thak Paul Reeder at SN facal ad Jeff Fsher for ther support ast but ot least, I would lke to thak a ad Tae for beg there for me 3

4 TABE OF CONTENTS Chapter : Itroducto Abstract Ackowledgemets Table of Cotets Overvew Defto of a Pure-Play Portfolo Objectves of a Pure-Play Portfolo Chapter 2: ethodology - Costructo of the Pure-Play Portfolo 2-. Basc Assumptos Basc odel Varace mzato Applcato of Two ethods ethod : atr Iverse ethod ethod 2: Ecel Solver Tool Comparso of Results Other Cosderatos Adjustg for Debt The Data Chapter 3: Applcato of Pure-play Portfolo 3-. Property-Type Pure-play Portfolos Geeral Characterstcs Hstorcal Ide of Pure-Play Retur Comparso wth NCREIF Returs Comparso wth NAREIT Property Returs agged Correlato Geographc ocato Pure-play Portfolos Geeral Characterstcs Hstorcal Ide of Pure-Play Retur vs. NCREIF Cotemporaeous Correlato ad agged Correlato Cocluso Refereces APPENDIX-A APPENDIX-B

5 CHAPTER : Itroducto Overvew Over the last 2 years the REIT dustry has made remarkable progress terms of the magtude of market captalzato ad maagemet performace. REITs provde a way to realze the ecoomc beefts of uderlyg real estate assets, obta stable, cosstet come ad log-term growth whle creasg portfolo dversfcato. Therefore, the correlato of REIT returs to the returs of other stocks ad bods has decled sgfcatly. However, may vestors ad aalysts are terested obtag pure-play retur ad hstorcal performace whch are eposed to oly oe target sector such as a property type ad a geographc locato. It s dffcult for vestors to precsely target ther vestmets because publcly traded REITs hold large umber of dvdual propertes whch vary by property types or locatos. Although the REIT dustry has specalzed ts vestmets wth regard to property type for the last years, t s mpossble to acheve the portfolo whch s eactly eposed to oly oe target sector by vestg a group of REITs whch represets the targeted property type. The costructo of a pure-play portfolo s a mechasm to address ths ssue a systematc way. Ths thess demostrates ad apples the procedure of dervg ulevered returs to the uderlyg propertes held by the REITs, broke out by property type ad geographc locato. I addto to costructo of REIT-based pure-play portfolos, a comparso of performace wth prvately owed propertes s performed. I ths case, the NCREIF sub de was used as a proy of performace of prvately held real estate assets. As the REIT-based pure-play portfolo s based o the publc stock market ad uderlyg property market whle the NCREIF de s based o the prvate property market, the comparso of performace betwee two markets ca help detfy REIT-level maagemet performace. We costructed pure-play portfolos for property types ad geographc locatos for 7 years,

6 Defto of a Pure-Play Portfolo A REIT-based pure-play portfolo s defed as a optmal combato of log ad short postos publcly traded REITs, such that the portfolo ca replcate the vestmet returs to a specfed target sector wthout ay drect eposure to other sectors. Here, the target sector ca be defed terms of ay attrbutes of property assets such as usage, locato, ad sze. For eample, f we wat to vest to REITs whch are eposed oly to offce propertes the east rego, we ca costruct the pure-play portfolo whose performace s related to the target sector wthout eposure to other attrbutes usg a combato of log ad short postos REITs. Here, short postos the portfolo are used to elmate eposure to o-target sectors. It s mportat to ote that there ca be may combatos whch ca meet the codto of eposure to a target sector. For eample, suppose that we ca costruct a pure-play portfolo whose target s a offce property by a 5% log posto to REIT A, ad a -5% short posto to REIT B. We could also costruct aother pure-play portfolo whose target sector s the same by a 3% log posto ad a -% short posto to REIT A ad B respectvely. Therefore, we have to choose the optmal portfolos amog the possble combatos. The optmal pureplay portfolo s the oe whch mmzes radom volatlty, or dosycratc rsk, whle matag pure eposure to the target. Idosycratc rsk s defed as a retur compoet that s uque to each REIT ad property type. That s, a retur of a REIT oe property type s made up of two compoets, the dosycratc compoet ad the pure-play compoet whch s same for ay vestmet the gve property type. For stace, assume that REIT A makes a % retur a offce property ad REIT B makes a 9% a offce property. If a aggregate-level pure-play offce retur s %, REIT A has a +% dosycratc compoet ts retur ad REIT B has a -% dosycratc retur. However, as we do ot kow the aggregate pure-play retur of offce, say %, advace, we eed to etract the pure-play retur compoet by dversfyg away the dosycratc compoet. From the statstcal pot of vew, the more assets ad propertes we hold a offce property, the more dversfcato effects we ca have ad the more accurate a pure-play retur we ca acheve. Note that the pure-play compoet of our REIT-based portfolo ca reflect ot 6

7 oly the passve retur of uderlyg assets, but also the REIT-level maagemet effect of each property type. Objectves of a Pure-Play Portfolo There are four ma objectves for costructo of REIT-based pure-play portfolos: speculato, hedgg, estmates of beta, ad costructo of hstorcal de. From a vestmet pot of vew, vestors wat to costruct portfolos whch are eposed to a target sector or hedged agast oe sector. Sce the assets of REITs are already dversfed across property sectors ad geographc locatos, t s dffcult for vestors to meet ther eeds perfectly by smply vestg oe or a group of REITs whch represet ther target sectors. Oce pure-play portfolos are costructed, we ca estmate sectoral betas usg the correlato betwee targeted sectors ad market. These sectoral betas ca be used whe estmatg the cost of captal captal budgetg. For eample, f we obta the beta of REITs west rego as 2 agast the total REIT market, the opportuty cost of captal of vestg west rego ca be acheved from the rsk-free rate ad epected market retur usg Captal Asset Prcg odel (CAP). The last objectve s to costruct hstorcal de usg pure-play returs across tme. The hstorcal de ca help evaluate the performace of specfed REIT sectors. I addto, we ca detfy REIT-specfc ad target-specfc maagemet compoets by comparg across several sectors that de wth other dces of assets whch are already costructed. 7

8 CHAPTER 2: ethodology - Costructo of a Pure-Play Portfolo 2-. Basc Assumptos Basc odel We ca model the retur of oe REIT whch vests oly oe property type: retal property. I that case, we ca break dow the retur compoet to the true retur to that property type ad dosycratc retur whch ca be thought of as radom ose. Ths dosycratc retur of the REIT ca be also regarded as the results of REIT-level asset maagemet compoets ad facal leverage effects. Ths model s show equato () below: ~ r = ~ r + e~ () R R Where the r s a true retur ad the e s a dosycratc retur. Now, we ca apply above relatos to a set of REITs, whose assets are made up of just two property types: offce propertes ad retal propertes. Aga, the retur compoets ca be broke dow to a pure-play retur ad a REIT-specfc dosycratc retur. The dosycratc retur compoets are dfferet accordg to dfferet compaes ad property types sce the dosycratc returs of large amout of assets should be smaller tha those of small amout of assets. O the cotrary, the pure-play true retur compoets are cosdered to be same across the same property type by defto. For eample, REIT A whch holds $ mllo assets offce ad $2 mllo retal should have larger dosycratc compoets retal property. By the same toke, REIT B whch holds $2 mllo assets offce ad $ mllo retal should have larger dosycratc rsks a offce property. However, the true retur compoets of same propertes are same regardless of the absolute amout of assets. Therefore, the retur to REIT s ~ r = ( ~ r + e~ ) + ( ~ r + e~ ) (2) R R R O O O 8

9 r~ = Retur to REIT. r~ = Pure-play retal property retur. R r~ = Pure-Play offce property retur. O R O = = Fracto of REIT vested retal property. Fracto of REIT vested offce property. e~ = Idosycratc retur to REIT s retal property. R e~ = Idosycratc retur to REIT s offce property. O Where R + O should be oe because each REIT holds oly offce ad retal propertes. Aga, t s mportat to ote that the pure-play porto of the retur s the same for ay vestmets the gve property-type sector. The dosycratc compoets are ucorrelated wth other radom varables ad the epected value s zero. Now, sce the objectve of a pure-play portfolo s to costruct a portfolo from the formato o asset fractos of may REITs, we ca thk of vestg a umber of compaes. The retur of our portfolo would be ~ r = w ~ r (3) P = where the weght w s REIT s proporto of the total value of our portfolo. I other words, w s the weght of REIT our pure-play portfolo. From equatos (2) ad (3), the portfolo retur ca be the equato below: ~ r P = ~ r w + ~ r w + ( w e~ + w e~ ) (4) O = O R = R = O O R R Note that above equato we splt the retur compoets of a gve property-type to hadle the processes more effcetly. Now as a et step, we have to costruct the pure-play portfolo, whch s eposed oly to the specfed target sector, offce or retal ths case, whle mmzg 9

10 the dosycratc rsks at the same tme. For eample, the offce pure-play portfolo ca be acheved by costrag several terms above equato. Frst, the sum of weghts to offce vestmet should be oe whle the sum of weghts to retal vestmet should be zero. Secod, we should come up wth the oe optmal soluto whch ca mmze the dosycratc varace sce there ca be several solutos that ca meet the frst codtos. Sce the varace of pureplay retur compoets, that s Var(r O ) ths case, s same regardless of dfferet weghts, the soluto whch mmzes the varace of a portfolo ca also mmze the varace of a dosycratc compoet. The frst costrat ca be epressed as below: = w O = ad w R = (5) = As metoed above, the frst part of the equato (5) makes sure that the portfolo s eposed oly to offce property ad the secod part elmates ay eposure to the retal property. Table shows smple case that eplas the basc model of a pure-play portfolo. If we cosder oly three REITs, each of whch holds oly two property sectors: offce ad property, we ca costruct a pure-play portfolo usg the equato (5). Note that ths eample we ca come up wth fte solutos whch ca meet weght costrats sce we have oly three ukow varables, w, w 2, ad w 3, whle there are two equatos related to these varables from the weght costrats. I geeral, whe the umber of compaes eceeds the umber of sectors, there are may ways to costruct a pure-play portfolo. To obta the optmal soluto amog the possble umerous aswers, we should apply the optmzato procedure whch ca yeld the mmum varace soluto. Table : Eample, Pure-play portfolo wth two property types ad three REITs Share of Frm Assets Offce ( O ) Retal ( R ) Portfolo Ivestmet (w ) Eposure to Offce (w O ) Eposure to Retal (w R ) REIT 8.% 2.%.% 8.% 2.% REIT 2 2.% 8.% -.% -2.% -8.% REIT 3 4.% 6.%.% 4.% 6.% Total.%.%.%

11 Varace mzato As eplaed above, there ca be umerous solutos of a pure-play portfolo sce usually the umber of frms that costtute a portfolo s bgger tha that of property types. For eample, Table 2 computes the dosycratc compoets of a portfolo. Whe we vest %, -%, ad % to REIT, REIT 2, ad REIT 3 respectvely, the total dosycratc eposure s 22.4%. However, f we chage the weghts to 9%, -62%, ad 43% whle meetg the offce ad retal weght costrats, we ca come up wth 5.6% total dosycratc eposure Table 3. Ths result makes sese as the dosycratc rsk of retal propertes s larger tha that of offce propertes, the absolute weght value of REIT 2 whch has a hgh fracto of the retal property should be smaller tha before. Table 2: Eample, Pure-play portfolo wth two property types ad three REITs Share of Frm Assets Offce Retal ( O ) ( R ) Idosycratc Std.Dev Offce Retal (σ o ) (σ R ) Portfolo Ivestmet (w ) Idosycratc eposure Eposure to Offce (w O ) Eposure to Retal (w R ) REIT 8.% 2.%.% 2.%.% 8.9% 8.% 2.% REIT 2 2.% 8.%.% 2.% -.% 6.% -2.% -8.% REIT 3 4.% 6.%.% 2.%.% 2.6% 4.% 6.% Total.% 22.4% *.%.% * Total dosycratc eposure = 3 2 owσ o ) + ( ( R R w σ ) 2 Table 3: Eample, Pure-play portfolo wth two property types ad three REITs Share of Frm Assets Offce Retal ( O ) ( R ) Idosycratc Std.Dev Offce Retal (σ o ) (σ R ) Portfolo Ivestmet (w ) Idosycratc eposure Eposure to Offce (w O ) Eposure to Retal (w R ) REIT 8.% 2.%.% 2.% 8.9%.6% 95.% 23.8% REIT 2 2.% 8.%.% 2.% -62.%.% -2.4% -49.7% REIT 3 4.% 6.%.% 2.% 43.2% 5.5% 7.3% 25.9% Total.% 5.6%.%.%

12 To address ths potetal several soluto ssue, we should fd the optmal soluto whch ca mmze the varace of dosycratc compoets, whch tur ca mmze the varace of a portfolo. There are two ways to acheve a mmum varace portfolo: matr verse method ad ecel solver fucto method Applcato of Two ethods ethod : atr verse method usg agragea Relaato Frst, by applyg the weght costrats of the equato (5) to the equato (4), we ca smplfy the portfolo retur equato as below: ~ r P = ~ r + ( w e~ + w e~ ) (6) O = O O R R Note that the dosycratc retur part remas the above equato. The varace of the portfolo would be ( ~ ( ~ 2 2 ) ) [ ( ~ 2 2 Var r = Var r + w ) ( ~ OVar e + w RVar e )] (7) P O = O R To get the mmum varace, agragea relaato method s used. agragea relaato s a techque whch was veted by agrage 797. Ths techque has bee very useful cojucto wth Brach ad Boud method as to boudary solutos. Usually, ths s a obvous tool for gettg lower bouds. Gve the orgal problem, wth some fed varables, ths techque relaes some of the costrats by sertg argrage multpler ad performs subgradet optmzato, whch tur wll gve a lower boud. From ths relaato assumpto, ew fucto of s created as below 2

13 = Var( ~ ro ) [ w OVar( e ~ O ) w RVar( e ~ R )] a w R h w O (8) = = = where a ad h are agrage multplers. Note that the value of the ew equato (8) s the same as the equato (7) sce we assumed that w = ad w = the equato (5). To = get the mmum of, the frst dervatves accordg to dfferet weghts, δ/ δwj (for j =,..,), should be equal to zero. Therefore, we ca get the followg dfferet equatos as below: O = R 2 ( ~ 2 [2w ) 2 ( ~ OjVar e w RjVar e )] + a( ) + h( ) = (for j =,..,) (9) j Oj + j Rj Rj Oj Now we have + 2 lear equatos ( equatos (9) ad two equato (5)) wth + 2 ukow varables ( weghts ad two a ad h). All we have to do s solve the + 2 lear equatos by usg basc algebra. However, we eed to geeralze the + 2 equatos because the umber of REITs cosdered aalyss teds to be qute hgh ad the umber of property sectors or the umber of other target categores s dfferet accordg to the stuato. Suppose that there are t categores pure-play portfolos ad you wat to costruct a portfolo property category. H would be defed as below: H t 2 = 2 k Var( e ~ ) () k = k If we substtute the equato () to the equato (9) ad defe t more varables as the equato (8), we ca acheve geeralzed matr system of pure-play equatos as the equato () below. 3

14 = O t t t t t t t t t a a h w w w w H H H H () Therefore, our portfolo weghts of a target property sector ca be acheved as below: = O t t t t t t t t t H H H H a a h w w w w (2) Due to the characterstcs of matr calculatos ad deftos of varables, we ca derve t target sets of portfolo weghts by oly seeg the upper-rght t part of verse matr. If we defe the compoets of t matr as I t (for t=,.., t ad =,.., ), we ca smplfy the fal soluto of pure-play portfolos as below: = t t t t t t t t I I I I I I I I I I I I w w w w w w w w w w w w SET t SET SET (3) 4

15 To solve the matr verse, we used GAUSS whch s a specalzed matr calculato program due to the huge sze of the matr. Estmates of Idosycratc Varace All data that s ecessary to costruct lear equatos ca be obtaed drectly from the structural formato of each REIT ecept for the dosycratc varace, Var e ~ ). Therefore, we eed to estmate the dosycratc varace of each REIT from the structural formato. I fact, precse value about the dvdual REIT dosycratc varaces across sectors mght be dffcult to obta. However, the pure-play portfolo weghts are a fucto of the relatve dosycratc varaces stead of the absolute values. I other words, relatve dosycratc amouts are eough to obta the optmal weghts. To show ths, dvde both sdes of the equato (7) by a costat, Y. ( k ~ ( ~ Var( r ) ) ( ~ ) ( ~ P Var ro 2 2 Var eo 2 2 Var er ) = + [ w O + w R ] (4) Y Y Y Y = If the same agragea relaato procedure s appled as before to the above equato (4), we ca see that the same portfolo weghts ca meet the codtos of a pure-play portfolo as below. 2 ( ~ 2 [2w ) 2 ( ~ OjVar e w RjVar e )] + ay ( ) + hy ( ) = (5) j Oj + j Rj Rj Oj The oly dfferece betwee the equato (9) ad (5) s agragea multplers, ay ad hy, whch do ot have ay effects o the optmal soluto. From above relatoshp, we could prove that the relatve dosycratc varace s eough to obta the optmal pure-play weghts. The et step s to estmate relatve dosycratc varaces. We assume that dosycratc varaces are versely proportoal to the dollar amouts of each REIT s holdgs each property type because the more assets they hold, the more dosycratc varaces ca be 5

16 dversfed away. For eample, f we thk of REIT ad j whch hold property types o ad r, the relatoshp betwee dosycratc varaces across REITs would be Var( e ~ Var( ~ e b aj ) A j = ) A aj b (6) where A s the dollar value of REIT s assets. Therefore, we ca defe the relatve dosycratc varaces of REIT aother way by sertg a costat, Z, as below Z Z Var ( e~ b ) = ad Var( e ~ aj ) = (7) A A b aj Usg the above equato, we ca estmate the relatve dosycratc varace of each REIT from ts structural formato, whch ca fsh the costructos of the lear equatos. ethod 2: Ecel Solver Tool Ecel solver tool ca detfy the optmal soluto by tral ad error methods gve specfed costrats. Keep md that our objectve s to fd the optmal weghts whch lead to the mmum dosycratc varace. From above assumptos the equato (7), we ca obta total relatve dosycratc varace of REIT, whch holds t property sectors as below t t t Z 2 2 Var( e~ = = = ) w t Var( e ~ t ) w t Zw t (8) t= t= A t t= A t Therefore, the total portfolo-level amout of dosycratc varace would be t 2 2 Var( e~ = P ) Z w t (9) = t= A t 6

17 Now, we defed the whole compoets ecessary to ru Solver tool. The put of Solver tool s made up of three parts: target cells, chagg cells, ad costrats. Target Cells: ake sure that the total dosycratc varace, equato (9), should be mmum amout. Chagg Cells: ake sure that the weght cells, w, should be chagg cells. Costrats Iput several costrats defed above. Whe a offce property s a target, w, w O =, ad w R = = It s mportat to ote that the optmal portfolo weghts whch lead to the mmum dosycratc varace has othg to do wth the magtude of a costat Z. Comparso of results We appled the two methods to the same stuato a effort to check the accuracy of the two methodologes. Table 4 shows the ma assumpto, where there are 5 REITs, each of whch has dfferet amout of assets but holds oly two property sectors. Table 4: a assumpto the comparso of the two methodologes Share of Frm Assets Asset (A ) Offce ( O ) Retal ( R ) Total Relatve Idosycratc Var. Offce (Z/(A O )) Retal (Z/(A R )) REIT 5 % 9% % REIT % 75% % REIT 3 3 5% 5% % REIT 4 2 6% 4% % REIT 5 9% % %.. 2. Total H 7

18 Table 5: Comparso of the two methodologes, atr Iverse ad Solver Tool atr Iverse () Solver Tool (2) Dfferece (2)-() Offce Retal Offce Retal Offce Retal REIT REIT REIT REIT REIT Total As we ca see above Table 5, the resultg sets of optmal weghts from two methods have eglgble dffereces. I followg aalyss of pure-play portfolos, we maly coducted the atr Iverse method sce there s hgher volume of operatos the Ecel tool method Other Cosderatos Adjustg for Debt Utl ow, we just assumed that REITs hold % equty ther assets. However, we have to kow ulevered property level returs of REITs case they use debt amog ther assets. To apply above soluto to REITs whch hold debt ther assets, we eed to redefe several varables usg leverage ratos of REITs. everage ratos are obtaed from balace sheet formato across REITs ad years. (See Apped A) The Data As we ca see above equato (3), oly oe matr verse procedure ca lead to the total pure-play portfolo weghts as log as we are able to get the approprate put formato of the matr equato (). I other words, oe of the most crtcal parts pure-play aalyss s data 8

19 gatherg. The data we eed to fll out the matr cosst of three parts: balace sheet formato, retur data, ad asset breakdow data. Frst, balace sheet data s eeded to detfy the amout of dosycratc varace terms across REITs ad to ulever REIT returs usg leverage ratos. Secod, retur data s eeded to obta the total portfolo retur from pure-play weghts across REITs. These two kds of formato are easy to obta from facal data base such as CRSP (Ceter for Research Securty Prces). Fally, we eed formato o the fracto of assets across specfed sectors such as property types ad locatos. Ths fal formato s the most challegg part to obta ad, tur, the most crtcal part the pure-play aalyss. It s mpossble to get perfectly precse asset breakdow because the balace sheet formato s related to the book value whle the asset breakdow should be based o ther market values. oreover, the marked to market real estate value across tme ad REITs s prvate formato, whch s dffcult to obta from all REITs drectly. Istead, we used publcly avalable data from NAREIT (The Natoal Assocato of REITs) ad the commercal data servce frm, SN facal. Sce ths data s based o the lease area, we coverted area formato to market values by multplyg prce per SF data by property type ad rego from NCREIF (Natoal Coucl of Real Estate Ivestmet Fducares). As eplaed above, very lttle formato s requred to costruct a pure-play portfolo. We do ot eed the statstcal relatoshp betwee property returs ad REIT returs. All we eed s the structural formato o REITs ad ther balace sheet formato to estmate the relatve magtude of the dosycratc rsk. I other words, the procedures of costructg pure-play portfolos are very straghtforward as log as the accuracy of data s guarateed. I ths research, we followed a drect path to obta the structural formato of REITs. ore accurate ad market value-based formato o REITs holdg ca lead to more relable results. 9

20 CHAPTER 3: Applcato of Pure-play Portfolo We appled above pure-play methodologes to costruct REIT-based pure-play portfolos for major property types (apartmets, offce, dustral, ad retal) ad geographc locatos (east, west, south, ad mdwest) for the perod For the aalyss, we used area-based property holdg data from SN facal ad orga Staley. We also used NCREIF market value data to obta value-based property holdgs. For the geographc locato, regoal breakdow of NCREIF was appled sce our market values of propertes ca be orgazed accordg to NCREIF regos. Property holdgs ad balace sheet data are o a aual bass; we costructed aual pure-play portfolos across property types ad geographc locatos. We started from publcly traded REITs across the perod, whch ca be derved from NAREIT. Amog those REITs, about 5% of REITs were fltered out for the accuracy of aalyss ad the relevace of formato. The ma cosderatos the flterg process are as below. Equty REITs: Equty REITs ow ad operate come-producg real estate but ortgage REITs led moey drectly to real estate owers, or drectly through acqusto of loas or mortgage-backed securtes. Hybrd REITs are compaes that both ow propertes ad make loas to owers ad operators. Eve though there s techcal way to adjust REIT vestmets cosderg the percetage of mortgage ther assets, these mortgage REITs ca troduce error to the results because t s dffcult to get the eact formato o ther mortgage holdg values ad the accompayg returs. a-property REITs: We eclude REITs whch hold more tha % of ther property holdgs property sectors such as health care, hotel, golf course, ad self storage. It s ot possble to fer the values of those mscellaeous sectors from NAREIF formato. I addto, the other sectors would geerate more short postos to get the pure-play portfolos of ma sectors, whch, tur, would magfy the dosycratc rsk the portfolo. 2

21 Adjustmet: The remag REITs after two tests are assumed to hold oly major property sectors, apartmet, offce, dustral, retal. Therefore, f oe REIT hold 95% asset major sectors, the remag 5% asset s dstrbuted accordg to the proportos of asset major sectors. We also ecluded several REITs o whch we caot get structural formato. Ths would ot make a major dfferece sce the umber of REITs whose data s uavalable s very small compared to total umber. I addto, the data of most major REITs whch hold a large magtude of assets s avalable. The lsts of REITs pure-plays for geographc locatos also followed the same lsts for property sectors of gve year. Table 6 shows the summary of REITs cluded pure-play portfolos for 7 years. We ca see the large crease the average asset of REITs compared to the umber of REITs the portfolos for 7 years. It s also mportat to ote that the average leverage rato, average asset dvded by average debt, s 2.6. Sce we have to acheve ulevered property level REITs returs, ths hgh leverage ratos make the ulevered REITs returs largely attrbutable to corporate bod de. The detaled eplaato of REITs leverage wll be made Apped A. Table 6 : Summary of Pure-Play Portfolos Year Number of REITs Total Asset (llo $) Total Debt (llo $) Av. Asset (llo $) Av. Debt (llo $) Av. everage Rato ,383 38, ,2 63,965, ,874 3,339,987, ,76 27,675 2,247, ,59 42,96 2,354, ,934 48,4 2,599, ,94 74,349 2,783, Average 89 83,49 4,45 2,6,

22 3-. Property-Type Pure-play Portfolos Geeral Characterstcs Table P-P7 Apped B show the pure-play portfolos whose targets are major property sectors, apartmet, offce, dustral, ad dustral, for the perod Fourth ad ffth colums show total asset ad leverage rato (total asset dvded by total equty) o a book-value bass. The other parts epla the asset breakdow to property sectors ad the fal portfolo weghts. The postve sgs weghts mea log postos ad egatve sgs mea short postos ther vestmets. As we ca see, we should vest about 65% bods to obta property level REIT returs. Especally, retal pure-play portfolos eed hgher weghts bods sce the REITs who specalze retal sectors use hgher leverage ther assets. The Ibbotso corporate-bod total retur de was used to represet bod returs the portfolos sce we assumed that the geeral corporate bod de ca represet the overall bod market regardless of dfferet dustres. The REITs dustry has made a lot of progress over the past two decades terms of the umber of REITs ad specalzato by property types. Specfcally, after md 9s a large umber of REITs teded to have specalzed vestmet targets such as apartmet ad offce. I other words, may recet REITs are already dvdual pure play vehcles by themselves. We ca see these treds from ther fractos of assets across property types. Ecept several REITs whch dversfed well across property types, most REITs hold ther specalzed vestmet sectors. These mprovemets REITs dustry make t possble to costruct more effectve pure-play portfolos wth small amout of short postos sce large short postos act lke leverage wth the portfolo, creasg dosycratc rsk of the portfolo. It s mportat to ote that durg the perod REIT dustry has bee stablzed terms of specalzato ad structure of vestmets. For stace, f we look at the pure-play portfolo of apartmet 23, the sum of short postos s oly.37% wth the lowest posto oe REIT.34%. oreover, 96% of log postos are vested resdetal REITs whch are classfed by NAREIT. I other words, ths result meas that most REITs already hold pure- 22

23 play portfolo by themselves wth regard to property types, so costructg pure-play portfolos makes o bg dfferece compared to vestmets the specalzed REITs accordg to the target property types. I fact, f we compare pure-play portfolo de wth NAREIF property de adjusted wth bod returs, we could come up wth almost same de. Ths result wll be dscussed et secto. For comparso, we ca cotrast the pure-play portfolo 23 wth 987 whch was costructed by Davd Gelter ad Bra Kluger 998. I 987, there s small umber of REITs the dustry ad the REITs were less specalzed wth regard to property types compared to the preset. Therefore, we eed to have very large amout of short posto to elmate o-target property sectors, whch makes a great deal of ose the retur compoets ad creases dosycratc rsk. For eample, the 987 apartmet portfolo demaded 95% of the portfolo be vested just four REITs, ad the total of all the requred short postos equals 67% of the etre portfolo value. Aother ssue we eed to cosder s the frequecy of rebalacg of portfolos. Before md 9s, there was a bg chage the REIT dustry wth regard to the umber of REITs ad specalzato by property types. Therefore, we eed to rebalace our pure-play portfolo at least every year. However, durg the perod of our aalyss, t s less ecessary to rebalace the portfolos as frequetly because of the stablzato of REITs dustry. Nevertheless, we rebalaced our portfolos o a aual bass sce we performed pure-play procedures to ot oly property sectors but also geographcal locatos, whch are less stablzed tha property sectors. I addto, the bg crease total assets ca have a mpact o the pureplay weghts. Hstorcal Ide of Pure-Play Retur Oe of the ma objectves of costructg pure-play portfolo s to compute the hstorcal returs ad compare those returs wth comparable returs. The aalyss of vestmet performace by property type ca be very mportat for several perspectves. From a fudametal perspectve, 23

24 dsaggregatg by property type ca allow the rchess the study of the ecoomc relatoshp betwee uderlyg property markets ad captal markets. From a vestmet aalyss perspectve, property type has always bee cosdered a major dmeso for dversfcato of real estate portfolos. NAREIT property de whch has bee costructed from 993 by classfyg each REIT as represetg a gve property type has the same motvato. I addto to the above cosderatos, the correlatos betwee REIT pure-play returs wth those of total markets ca help obta the beta compoets wth regard to overall market stuatos. For stace, the relatve volatlty of pure-play portfolos across property types compared to the total REIT market ca represet the rsk compoet of a gve property type. That rsk compoets ca be traslated to market beta, whch have a mpact o the requred cost of captal of vestmet. From the pure-play weghts that we derved table P-P7 Apped B ad the past retur compoets cludg corporate bod de, we ca compute hstorcal returs of REIT-based pure-play portfolos for four property sectors. As eplaed above, the Ibbotso corporate-bod total retur de was used to represet bod returs. Table 7 shows the hstorcal returs ad Ide of pure-play portfolos for the perod The Ide value of pure-play returs was obtaed by computg compoud retur of oe dollar value vested 996. We ca see that, dustral ad retal sectors outperformed apartmet ad offce sectors for the past 7 years. 24

25 Table 7: Hstorcal Ide of Pure-Play Quarterly Returs vs. NCREIF usmoothed Quarterly Returs, 997~23 Pure-Play Ide Returs NCREIF Ide Returs (Usmoothed) Year Q APT Quarterly Total Retur (%) Cumulatve Total Value ($) Quarterly Total Retur (%) Cumulatve Total Value ($) offce Idus tral Retal APT offce Idustr al Retal APT offce Idustr al Retal APT Offce %.7% -.2% -.8% % -.5% 3.%.3% % 3 7.% 4 2.4% 4.8% 4.% 5.3% % 4.6% 4.3% 3.6% %.2% 5.4% % 7.2% 5.9% 3.3% % 5.9% 4.% % 9.5% 4.2% 4.% % -.6% -.3% 2.2% % 5.4% 2.7% 3.9% % 3 -.5% 4 -.7% -.9%.2% 2.4% % 4.5% 6.7% 3.7% %.4%.5% % 2.% 2.5% 2.4% % -2.3%.% % 5.2% 2.6% 3.5% % -2.2% -4.2% -4.4% %.7%.4% 3.9% % 3-2.5% 4.3% 3.4% 3.6% -.4% % 3.% 3.7% 2.% % -4.% -2.7% % 3.2% 3.3% 2.7% % 2.3% -.7% % 2.7% 2.% 3.2% % 2.6% 2.3% 2.9% % 2.4% 2.9%.7% % 3 6.% 4 6.4% 5.3% 6.8% 2.2% % 4.2% 4.3% 3.3% % 6.4% 3.3% % 2.9% 3.7% 2.% % 4.% 5.4% % 4.% 4.9%.9% %.6%.4% 6.% % -.% -.2% 2.3% % 3 3.% 4.6% 4.9% 4.3% 4.3% %.7% 3.8%.6% %.% 2.% %.3%.9% 2.8% % 3.4% 4.4% % -.9% -.7%.2% % 2.5% 4.4% 2.9% % 2.3% 2.7% 2.4% % 3.9% 4.6% 3.7% 7.3% 6.3% %.2%.% 3.8% % 2.9% 6.% %.8% 2.6% 3.6% %.8% 2.% % -.5% 2.% 5.4% %.9% 2.4% 3.6% % 3.% 2.6% 6.% % 3 2.9% 4.% 8.6% 6.% 7.2% %.3%.5% 4.2% % 2.6%.5% % 2.% 3.4%.4% % 3.4% 2.7% %.7% 2.9% 9.6% Idus tral Retal 25

26 Comparso wth NCREIF returs The pure-play de s based o REIT ad Bod returs, ad therefore represets the ecoomcs of publc securtes ad bod market. I other words, REIT performace ca be attrbutable to uderlyg property market, stock market, ad REIT-level maagemet power such as asset maagemet ad facal leverage. O the cotrary, NCREIF de s based o the prvately held propertes of sttutoal vestors. NCREIF de s the most wdely used bechmark of apprasal-based total returs to commercal property of sttutoal vestors. Therefore, the performace of NCREIF de ca be attrbutable to the ecoomc of the uderlyg property market although there are some smoothg ad laggg effects caused by the apprasal process ad the de costructo procedure. Recet studes have suggested that there are terestg dffereces ad smlartes betwee REIT returs de ad apprasal-based returs de. Specfcally, several recet studes have suggested that REIT returs ad apprasal based returs are fudametally related, but wth REIT returs leadg apprasal returs by oe to three years. From the several perspectves metoed above, t s very terestg to compare REIT-based de wth apprasal-based NCREIF de. The rght eght colums table 7 show NCREIF quarterly returs ad compoud retur values vested 996. The NCREIF returs have bee usmoothed usg the procedures descrbed Fsher, Gelter ad Webb (994). Ths usmoothg techque s appled to correct for the effect of apprasal smoothg ad temporal aggregato. NCREIF returs cosst of two compoets, come returs ad apprecato returs, ad there are two ways to defe the correspodg come compoets: NOI based ad Cash flow based. Accordg to chael, Davd, Wllard, Douglas (995), f the bulk of the REIT share prce apprecato reflects property developmet ad acqustos that are largely eterally faced, the comparable form of the NCREIF Ide apprecato returs should be the offcal NOI-based defto. If, o the other had, the bulk of the REIT share prce apprecato reflects same-store property market value growth ad route captal mprovemet epedtures that are terally faced (essetally, from the cash flow geerated by the REITs property holdgs), the the cash flow based defto of apprecato 26

27 retur the NCREIF Ide s best to use for comparablty to REIT prce apprecato. We used cash flow based returs sce we assumed that method would more accurately portray the actual prce path of propertes over tme gve the realty of captal mprovemets to the propertes. Fgure shows the values of $ vested to pure-play portfolos ad NAREIF de at the ed of 996 for each property type whch were computed Table 7. The overall movemets of the de value are almost smlar, but NCREIF returs outperformed REIT-based pure-play returs by small amout. I partcular, REIT portfolos suffered very lower returs the perod of every sector because of the geeral poor performace of REIT market. After that perod, REIT-based pure-play returs slghtly outperformed NCREIF returs, decreasg the gap betwee two dces. From the shape of de le ad the summary statstcs eplaed et secto, we ca see that the returs of REIT portfolo ted to be more volatle tha NCREIF returs, eve though the REIT-based returs are ulevered ad the NCREIF returs are usmoothed. Fgure : Value of $ vested at the ed of 996, REIT-based Pure-Play vs. NCREIF 2.5 Apartme ts 2. Value($) Quarter Pure-Play NCREIF 27

28 2.5 Idustral 2. Value($) Quarter Pure-Play NCREIF 2.5 Offce 2. Value($) Quarter Pure-Play NCREIF We ca also see that the best performace property sectors durg the past 7 years are dustral ad retal sectors both REIT pure-play de ad NCREIF de. 28

29 3. Retal 2.5 Value($) Quarter Pure-Play NCREIF Although the orders of performace do ot perfectly match betwee two dces, the geeral treds of performace are very smlar wth each other. We ca fer that the dfferece absolute de value ca be caused by the dfferet characterstcs of two market, publc REIT market ad prvate property market. However, the ecoomcs of uderlyg propertes both markets ca make the relatve values almost smlar. For eample, f apartmet propertes geerate hgher returs tha other sectors property market, the REITs whch specalze a apartmet sector ted to have hgher performace tha other REITs. Table 9 shows the summary statstcs for the quarterly returs of REIT-based pure-play portfolo ad usmoothed NCREIF de. As eplaed above, pure-play portfolos have hgher volatlty ad lower average returs tha NCREIF returs. The correlatos betwee pure-play portfolos ad NCREIF returs for the same property sectors are aroud 25%. We ca fer that the potetal laggg stuato betwee two markets ad the dfferece market ecoomcs ca lead to these low correlatos. O the other had, the correlatos betwee dfferet property sectors the same market show very close movemets ecept for the retal sector NCREIF returs. 29

30 Table 8: Basc Statstcs for REIT-based pure-play portfolo, NCREIF usmoothed Retur, ad Other Ide based o quarterly returs, 997~23 ea eda a.. APT Pure-Play Quarterly Retur (%) NCREIF Usmoothed Retur (%) Adjusted NAREIT property INDEX (%) Offce Idus tral Retal APT Offce Idus tral Retal APT Offce 2.3% 2.2% 2.7% 2.6% 2.7% 2.6% 2.9% 3.2% 2.3% 2.3% 2.6% 2.6% 2.3% 2.%.9% 2.% 2.8% 2.8% 2.6% 2.3% 2.8% 3.2%.6% 2.% 3.% 2.6% 2.8% 2.% 7.%.7%.2% 7.2% 5.3% 9.5% 6.7% 9.6% 7.5% 2.8% 8.6% 7.3% 2.3% 9.% -2.5% -4.8% -4.2% -4.4%.% -.9% -.7%.4% -2.5% -3.9% -3.7% -4.3% -7.3% -3.6% Idus Tral Retal S&P 5 Corp. Bod Ide Stadard devato 2.6% 3.5% 3.3% 2.9%.4% 2.5%.6%.8% 2.6% 3.6% 3.% 2.8%.3% 3.% CORREATION Pure-Play APT.% Offce 89.6%.% Idustral 8.5% 82.6%.% Retal 63.% 6.8% 66.9%.% NCREIF APT 2.8% 22.% 29.8% 2.%.% Offce 22.% 23.5% 25.3% 8.6% 63.9%.% Idustral 3.2% 26.8% 28.% -.2% 58.7% 7.5%.% Retal -7.5% 6.9%.6% 6.5%.6% 8.% -.7%.% NAREIT Property APT 99.% 89.% 82.4% 64.9% 3.5% 25.5% 3.4% -4.% % Offce 86.7% 97.6% 8.7% 59.8% 2.% 28.% 27.3% 6.6% 87% % Idustral 83.% Retal 66.8% S&P 5 3.9% Corporate Bod Ide 36.4% 83.7% 97.5% 69.3% 2.9% 8.7% 9.8%.9% 83% 8% % 64.% 7.2% 99.4% 2.%.2% 2.2% 8.7% 69% 62% 74% % 8.5% -.% -6.% 9.% 22.2% 3.2% 22.3% 6% 2% -2% -6% % 3.6% 39.% 79.% 37.7% 8.5% 6.% 7.6% 4% 34% 38% 78% -36% % 3

31 Comparso wth NAREIT property returs Over the past years, REIT dustry epereced cosderable specalzato to property sectors. Therefore, t has bee very easy to classfy oe REIT as represetg a gve property sector. As of Dec. 23, there were oly 6 REITs classfed as dversfed out of 73 NAREIT member compaes. The whole remag REITs have ther ow specalty the property sectors such as apartmet, offce, ad retal. NAREIT costructed NAREIT sub-de of property from 993 accordg to ther classfcatos. Property de s calculated usg portfolos whose weghts are proportoal to ther market captalzato wth ther specalzed property sectors. We cluded NAREIT property de our quarterly retur aalyss to compare the results wth pure-play portfolo returs. To make apples to apples comparso wth ulevered pure-play returs, the de was adjusted usg the weghts of debt vestmet the pure-play portfolos. For stace, f 5% debt vestmet s eeded the pure-play portfolo of the apartmet sector 2, the adjusted NAREIT property de of apartmet a gve year s acheved by applyg the same 5% weght debt vestmet. By dog so, we ca obta the comparable retur compoets depedet of bod returs. Table 8 also cludes several basc statstcs of adjusted NAREIT property de ad Fgure 2 shows the quarterly returs of three dfferet dces: Pure-play portfolo, Usmoothed NCREIF de, ad NAREIF property de. As we ca see, the mmum correlato betwee pure-play portfolo ad adjusted NAREIT de s 97.5%. Sce the REIT dustry already specalzed ts vestmet across property sectors, the vestmet strategy of focusg oe specalzed property sector does ot make a bg dfferece wth pure-play portfolos. We ca epect these results from above pure-play computatos sce, for eample, pure-play portfolo for apartmet 23 calls for 96% log postos apartmet REITs classfed by NAREIT. Fgure 2 shows that the pure-play ad NAREIT returs are almost smlar to each other, sometmes perfectly overlap. 3

32 Fgure 2 : Quarterly returs for Pure-Play, NCREIF, ad NAREIT for property sectors, 997~23.% Apartmets 8.% 6.% Value($) 4.% 2.%.% -2.% % Quarter Pure-Play NCREIF NAREIT Offce 4.% 2.%.% 8.% Value($) 6.% 4.% 2.%.% -2.% -4.% % Quarter Pure-Play NCREIF NAREIT 32

33 Idustral -6.% -4.% -2.%.% 2.% 4.% 6.% 8.%.% 2.% Quarter Value($) Pure-Play NCREIF NAREIT Retal -6.% -4.% -2.%.% 2.% 4.% 6.% 8.%.% 2.% Quarter Value($) Pure-Play NCREIF NAREIT 33

34 agged Correlato Wth regard to the statstcal relatoshp betwee REIT-based pure-play portfolo ad NAREIF sub-de, the correlatos show Table 8 are cotemporaeous correlatos usg quarterly returs wth sectors. However, accordg to the prevous study, quarterly cotemporaeous correlatos are subject to obfuscato by ose ad traset effects. Therefore, more fudametal relatoshp betwee two markets ca be detfed usg aual retur across tme. I ths aalyss, the results of the prevous pure-play study by Gelter ad Kluger (998) are used sce 7 year data s ot eough to get relable statstcal results. Therefore, we computed lagged correlatos betwee pure-play portfolo ad NCREIF sub-de usg 7 year (987-23) aual retur data. Prevous research has show that REIT-based returs ted to lead prvate-property returs by several years. Table 9 shows aual-frequecy lagged correlatos betwee pure-play portfolo ad NAREIF sub-dces wth each property type. Wth o lag tme, offce ad dustral sectors show postve correlatos whle apartmet ad retal sectors have egatve relato. By creasg lag up to 4 years, we ca see more substatal postve correlatos betwee two markets. I partcular, all four property types dsplay postve correlato at two-year lag. Therefore, these results ca partly support the prevous research that the REIT market teds to lead property market by some years. Table 9: agged Correlato betwee REIT-based pure-play portfolos ad NCREIF, aual frequecy total returs, 987~23 ag(year) Apartmet Offce Idustral Retal -76% 43% 48% -% -34% 57% 47% 46% 2 44% 4% 53% 25% 3 3% 45% 64% -3% 4 5% 37% 46% -22% 34

35 3-2. Geographc ocato Pure-play Portfolos Geeral Characterstcs After applyg the procedures of pure-play portfolos ad hstorcal de to property usages, we appled the same procedures to geographc locatos, whch are classfed as West, East, South, ad dwest by NCREIF (See fgure 3). For the effectve comparso data avalablty, the same REITs from property-type portfolos are pcked up for a gve year. Fgure 3: Geographc ocatos by NCREIF ad Table G-G7 Apped B show the pure-play portfolos whose targets are locatos, West, East, South, ad dwest, for the perod Every data table s orgazed the same way as the property-type case ecept that the locato classfcato s used stead of the property type. As metoed above, most REITs have specalzed vestmet targets across property types. As for locatos, o the other had, we ca see that the assets of may REITs are more dversfed tha property types. Therefore, to cacel out utargeted sectors, we should use larger amout of short postos wth the portfolos. For eample, f we see the pure-play portfolo 23, the sum of short postos of south s 29.96%, whch s farly hgh compared to.37% of the apartmet-targeted portfolo 23. I addto, to get the same portfolo, we should have log postos 5 REITs out of 96 but, case of the apartmet portfolo 23, oly 27 REITs get log postos. I oe sese, large umber of REITs log postos ca be thought to be helpful the dversfcato effect of dosycratc rsks. However, ths advatage teds to be offset by small amout of total assets log posto ad large amout of assets short posto. Therefore, pure-play results of locatos ca have more ose the retur compoets ad accompayg dosycratc rsks. 35

36 It s also mportat to ote that eve though the assets of REITs are more dversfed across geographc locatos tha property usages, most part of log postos are held by several bg REITs whch have large assets gve regos. Ths result s the same as the case of property type pure-plays. Hstorcal Ide of Pure-Play Retur vs. NCREIF Aga, we computed the hstorcal returs of pure-play portfolos ad compared those returs wth comparable returs, NCREIF sub-dces. ke the case of property sectors, we acheved hstorcal returs of REIT-based pure-play portfolos for four geographc locatos from the pure-play weghts table G-G7 Apped B ad the past retur formato o REITs ad corporate bod de. Table shows the quarterly returs of pure-play portfolos ad NCREIF for the perod Here, NCREIF de s also adjusted usg usmoothg techque appled to the prevous property typed case. Fgure 4 shows the comparso of performace for four dfferet regos, whch s terpreted as the amout of $ vested 996. We ca see that the West rego outperformed the other regos both REIT ad NCREIF de for the past 7 years. However, the orders of performace from 2 d to 4 th are a lttle dfferet. For stace, REIT-based pure-play results are sorted as West, dwest, East, ad South by descedg order but NCREIF results are West, East, South, ad dwest. Ths ca result from REIT-specfc maagemet compoets whch are assumed to be dfferet for every sector, dfferet ecoomcs of stock market ad uderlyg property market, ad the possble data ose from the costructo of pure-play portfolos. If we see Fgure 4, the overall movemets of de for the two markets are almost smlar, but NCREIF returs outperformed REIT-based pure-play returs by small amout ecept dwest rego. For the same reaso of the property sector case, REIT portfolos suffered low returs or losses the perod of 997~