Property Type, Size and REIT Value

Transcription

1 THE JOURNAL OF REAL ESTATE RESEARCH Property Type, Sze and REIT Value Denns R. Capozza* Sohan Lee** Abstract. Ths study documents the wde devaton of securtzed real estate assets n equty REITs from the value of the underlyng commercal propertes. A procedure for estmatng the net asset value of REITs s developed and the estmates are used to nvestgate the sources of premums/dscounts from net asset value n a large sample of equty REITs. To avod measurement error bas, two-way analyss of varance s used to test for dfferences among sze and property-type categores. The results ndcate that retal REITs trade at sgnfcant premums relatve to the average REIT whle warehouse/ndustral REITs trade at dscounts and small REITs trade at sgnfcant dscounts whle large REITs trade at premums. The dscounts and premums from net asset value do not translate nto hgher cash flow yelds. Introducton The most dramatc change n the real estate ndustry n the last few years s the rapd ncrease n the securtzaton of real estate through real estate nvestment trusts (REITs) Many factors have contrbuted to the explosve growth of REIT ntal and secondary publc offerngs. Often cted factors nclude changes n the captal requrements for commercal lenders that have made mortgage loans more costly, and changes n tax laws that reduced the tax-favored status of other ownershp forms lke real estate lmted partnershps. 1 It s often argued that REITs trade at dscounts to the net asset value of the underlyng propertes. Despte the mportance for valuaton and management ssues, the only evdence avalable s ndrect (Corgel, McIntosh and Ott, 1995; Goebel and Ma, 1993; Shllng, Srmans and Wansley, 1986). 2 Ths paper provdes the frst drect evdence on premums and dscounts by developng a procedure for estmatng net asset values for REITs and applyng the procedure to a large sample of equty REITs from Descrptve statstcs on the sample are reported n three ways. Frst s a summary of the sze, property types, ncome, expenses, and dversfcaton of the equty REITs for each year from The sample conssts of seventy-fve REITs but not for every year. The sample grows from thrty-three to seventy-two REITs durng ths perod. The data s then segmented by property type and sze class. Each REIT s classfed as an apartment, warehouse, retal or offce REIT f more than 50% of the propertes owned are of one type. The sze segmentaton uses quartles derved from total assets. In each segmentaton we nvestgate dfferences n expenses, cash flow yeld, dversfcaton, and captal structure. *Fnance Department, School of Busness Admnstraton, Unversty of Mchgan, Ann Arbor, Mchgan **Korean Insttute of Fnance, 80-6 Soosong-Dong, Chongro-Gu, Seoul , Korea. Date Revsed June 1995; Accepted July

2 364 THE JOURNAL OF REAL ESTATE RESEARCH To prevew the conclusons, the stock market valuatons of REITs as measured by premums above the values of the underlyng propertes declned durng the perod. Wall Street (.e., the stock markets) placed hgher values on retal REITs and lower values on warehouse REITs relatve to Man Street (the local property markets). In the sze quartles, small REITs trade at much larger dscounts than large REITs. The dfferences are hghly statstcally sgnfcant. Whle dscounts from net asset value are sgnfcantly below average for small REITs, cash flow yelds are not sgnfcantly hgher than those of larger frms. That s, there s no evdence that dscounts from net asset value are beng translated nto hgher cash flow yelds. Ths suggests that when portfolos of propertes are securtzed nto equty REITs, consderable value can be added or substracted from the underlyng propertes. The next secton descrbes the data and our procedure for estmatng net asset values. The thrd secton characterzes the sample. The fourth and ffth sectons analyze the REITs by property-type and sze categores. The fnal secton concludes. The Data Crtera for Incluson n the Sample The Natonal Assocaton of Real Estate Investment Trusts (NAREIT) provdes a lst of all publcly traded REITs each year n the NAREIT Sourcebook. To focus on commercal property, all mortgage, hotel, restaurant, and hosptal REITs are excluded. REITs that do not trade on NYSE, AMEX and NASDAQ or for whch property nformaton s not avalable are also excluded from our database. Wth these exclusons the sample conssts of seventy-fve REITs. The ncluded REITs are lsted n Exhbt 1. Sample Sze The seventy-fve equty REITs appear n the sample for at least one year. Four hundred and sxteen total observatons are avalable to study. Thrty-two of the REITs appear n all eght years. Each REIT s classfed nto a property type when more than 50% of the property held s of one type (apartment, warehouse, retal, offce). If no one property type s more than 50%, the REIT s classfed as Dversfed. Retal REITs are the most common wth thrteen to twenty n each year. Apartment REITs are the least common wth only three to sx n each year. There s an ncrease n warehouse REITs n 1991 because eghteen Publc Storage Equtes partnershps converted to REIT status. Sources of the Data The data for each REIT was compled from 10-K reports, annual reports to shareholders, proxy statements, and the CRSP daly return fle. Metro- and regonal-level property nformaton was obtaned from the Natonal Real Estate Index (NREI), Market Hstory Reports. 3 Varables n the Database The database ncludes balance sheet, ncome statement and property varables from the 10-K reports. The property data are classfed by regon usng the eght economc regons VOLUME 10, NUMBER 4, 1995

3 PROPERTY TYPE, SIZE AND REIT VALUE 365 Exhbt 1 Included REITs *B.R.E. Propertes Inc. Berkshre Realty Co. Inc. *Bradley Real Estate Trust Burnham Pacfc Propertes Inc. *Calforna Real Estate Invt Tr. Cedar Income Fund Ltd. Cedar Income Fund 2 Ltd. Chcago Dock and Canal Trust *Clevetrust Realty Investors *Contnental Mortgage & Eqty Tr. Copley Property Inc. Cousns Propertes Inc. Dal REIT Inc. Duke Realty Investments Inc. *E.Q.K. Realty Investors 1 *Eastgroup Propertes *Federal Realty Investment Trust *Frst Unon Real Est. E.Q.&M.G. Invts Grubb & Ells Realty Inc. Trust *H.R.E. Propertes *I.C.M. Property Investors Inc. *I.R.T. Property Co. Income Opportunty Realty Trust Koger Equty Inc. Landsng Pacfc Fund Lnpro Specfed Pptys *M.G.I. Propertes Inc. *M.S.A. Realty Corp. *Merdan Pont Realty Tr. 83 *Merdan Pont Realty Tr. 84 Merdan Pont Realty Tr. IV Merdan Pont Realty Tr. VI Merdan Pont Realty Tr. VII Merdan Pont Realty Tr. VIII *Merry Land & Investment Inc. Monmouth Real Estate Invt Corp *New Plan Rlty Trust *Nooney Realty Trust Inc. *One Lberty Propertes Inc. P.S. Busness Parks Inc. Partners Preferred Yeld Inc. Partners Preferred Yeld II Partners Preferred Yeld III *Pennsylvana Real Est. Invt Tr. *Property Trust Amer *Prudental Realty Trust Publc Storage Propertes VI Publc Storage Propertes VII Publc Storage Propertes VIII Publc Storage Propertes IX Inc. Publc Storage Propertes X Inc. Publc Storage Propertes XI Inc. Publc Storage Propertes XII Publc Storage Propertes XIV Publc Storage Propertes XV Inc. Publc Storage Propertes XVI Publc Storage Propertes XVII Publc Storage Propertes XVIII Publc Storage Propertes XIX Publc Storage Propertes XX *Real Estate Investment Trust Ca. Realty South Investors Inc. *Santa Anta Rlty Enterprses Szeler Property Investors Inc. *Trammell Crow Real Estate Invs. *Transcontnental Rlty Invstrs *U.S.P. Real Estate Investment Trust *Unted Domnon Realty Tr. Inc. Vanguard Real Estate Fund I Vanguard Real Estate Fund II Vnland Property Trust *Washngton Real Est. Invt Tr. *Wengarten Realty Investors *Western Investment Real Est. Tr. Wetterau Propertes Inc. Ths table lsts all the REITs n the data sample. Each REIT appears n at least one year. Starred REITS appear n all years. as defned n Hartzell, Shulman and Wurtzebach (1987). Exhbt 2 provdes descrptve statstcs on the varables. Calculated varables are explaned below. Net Operatng Income Net operatng ncome (NOI) or property ncome for each REIT s defned to be ncome from propertes before nterest, deprecaton, and overhead expenses (G&A) and s calculated by takng rental ncome mnus property expenses (property taxes, property management expenses, property operatng expenses, etc.). To refne ths measure, adjustments were made to reflect purchases and sales durng the year and jont ventures where the REIT owns less than 100%.

4 366 THE JOURNAL OF REAL ESTATE RESEARCH Exhbt 2 Summary Statstcs Varable Mean Max. Mn. Std Dev. Total assets ($ mllon) Property assets ($ mllon) Prop. assets/market value of prop. (%) Total book assets/total market assets (%) Occupancy rate (%) Weght cap. rate (%) Net ncome ($ thousand) 3,963 49,446 (58,609) 9,209 General & admnstratve ($ thousand) 1,390 15, ,478 Cash flow per share ($) G&A/total assets (%) Cash flow yeld (%) Leverage rato (%) Herfndahl ndex propertes (%) Herfndahl ndex regon (%) Value wt. premum (dscount) (%) Ths table provdes summary statstcs on selected varables from the equty REIT database. Total assets and property assets are book values. Total market assets are measured by (estmated market value of propertes + other assets). The leverage rato s defned as total labltes / (total labltes + market value of the equty). Net Asset Value Our procedure for estmatng net asset value modfes standard apprasal methods by usng a weghted captalzaton rate approach. Property-level captalzaton rates are used to derve a weghted captalzaton rate for the property portfolo. Ths weghted captalzaton rate s then appled to total property net operatng ncome to obtan an estmate of the value of the property portfolo n each REIT. Specfcally, assume that value addtvty holds so that BASE V = V, (1) where V s the value of the portfolo and V s the value of property. Usng the defnton of a captalzaton rate gves BASE V = NOI, (2) PCR where PCR s the portfolo cap rate, and NOI s the portfolo net operatng ncome. Therefore PCR NOI NOI = = BASE V, (3) V VOLUME 10, NUMBER 4, 1995

5 PROPERTY TYPE, SIZE AND REIT VALUE 367 and BASE 1 PCR = NOI, NOI 1 CR (4) whch can be approxmated by 4, NOI BASE PCR (5) NOI CR where CR s the cap rate on the th property. Ths can be rewrtten as BASE PCR = wcr, (6) NOI where w = are the weghts. That s, the portfolo cap rate s a weghted average of the NOI cap rates on the ndvdual propertes. 5 Snce net operatng ncome by property s rarely avalable for use n determnng the weghts, the choce of weghts s problematc. One alternatve would be to use the square footage as weghts. The dsadvantage of usng square footage s that t gnores the wde varaton n prce per foot by locaton and property type. To mprove on ths, weghts determned by metropoltan area averages of property values are used. The property values are roughly proportonal to net operatng ncome by defnton. That s, w V P SF = =, BASE V P SF (7) where P s the average prce per square foot for property of the same type n the same cty and SF s the square footage of the th property. To clarfy, prces per square foot, as compled by NREI, are used only to help determne the weghts to place on the cap rates (see equaton 7). Snce the prces per square foot are metro area specfc, error arses when the propertes are above or below average n qualty for that area and property type. If a property s below average n qualty so that the actual prce per square foot s below the NREI prce/sf then we overweght the cap rate of that property type and metro area n our calculaton of the cap rate for the frm. Each property s also assgned a property-type and locaton (metro area)-specfc captalzaton rate. These captalzaton rates are then weghted as ndcated above. The weghted average captalzaton rate s then appled to the portfolo s current net operatng ncome (property ncome) to derve a market value for the entre property portfolo usng equaton (2). Each REIT s net asset value (NAV) s computed as follows: MarketValue of the Propertes + Other Assets Total Labltes BASE NAV =. (8) Number of Shares Outstandng

6 368 THE JOURNAL OF REAL ESTATE RESEARCH Labltes wth below or above market nterest rates were adjusted to reflect market values. To further refne the estmates, addtonal adjustments were made to reflect occupancy rates and the qualty of the propertes when approprate. Snce the data on net asset values are estmates, they are subject to measurement error. As a result, care must be taken f ths varable s to be used n stuatons where measurement error bas can arse. One example would be usng the estmated net asset values as an ndependent varable n an OLS regresson equaton. 6 Ths paper uses twoway analyss of varance where groupng the data avods the bas problem. Value-Weghted Premum The premum of each REIT s measured by (stock prce net asset value)/net asset value. Value-weghted premum (VWPREM) s defned as follows: BASE VWPREMt = w premt, (9) where NAVt w = n t BASE NAV (00) 1 Prem t = (SP t NAV t )/NAV t, NAV t = net asset value of REIT at end of perod t, SP t = stock prce of REIT at end of perod t, and n t = the number of REITs wth avalable PREM t and NAV t data at the end of perod t. n t = 1 Concentraton Indces The measures of dversfcaton/concentraton are Hrschman-Herfndahl ndces that are commonly used n ndustral economcs to measure monopoly power. 7 Defne: Property Dversfcaton Index, BASE HHPROP = 2 S (10) Regonal Dversfcaton Index, BASE HHRGN = 2 (11) where S = the proporton of a REIT s portfolo nvested n property-type, and S j = the proporton of a REIT s portfolo nvested n regon j. The above ndces measure how concentrated the propertes n a REIT are,.e., f the REIT s hghly focused, the ndex s close to one, and f dversfed, the ndex s close to zero. j S j VOLUME 10, NUMBER 4, 1995

7 PROPERTY TYPE, SIZE AND REIT VALUE 369 Cash Flow Yeld Defne cash flow as: net ncome deprecaton and amortzaton gan on property sales extraordnary expenses. The cash flow yeld s measured by CF t /SP,t 1, where CF t =cash flow of REIT at end of perod t, and SP,t 1=stock prce of REIT at end of perod t 1. Characterstcs of the Sample The Value of Property n REITs Exhbt 3 descrbes the property assets held by equty REITs. Book value of property assets trpled from 1985 to The square footage of retal space held by REITs doubled whle the square footage of warehouse space quadrupled. The large ncrease n warehouse space s partly due to the converson to REIT status of the eghteen Publc Storage Lmted Partnershps n The REITs n the sample are much more concentrated n retal property than the U.S. stock (see Exhbt 3). Over half the property n REITs was retal n 1992 whle only about 25% was retal n the U.S. accordng to the RREEF natonal estmate. Warehouse/ndustral s underweghted n the REIT sample. The reasons for these under/overweghtngs are an nterestng open queston on whch some evdence s provded below. Average REIT Sze Exhbt 4 outlnes the average REIT sze by property type and sze quartle. The average REIT had $127 mllon of total assets durng the perod. By property type, retal, apartment, and dversfed REITs are above average n sze whle offce and warehouse Exhbt 3 Selected Characterstcs by Property Square Footage Property Value (000) ($ mllon) Total Total Number Assets Ware- Ware- of REITs ($ mllon) Offce house Retal Apt. Offce house Retal Apt ,945 9,714 12,216 35,805 16,216 1, , ,483 11,015 17,337 44,219 16,306 1, , ,385 13,780 25,004 50,967 16,429 2, , ,474 17,650 27,389 57,911 18,740 3, , ,112 19,192 27,970 62,832 19,973 3, ,630 1, ,284 20,812 31,963 67,240 23,366 2,999 1,106 7,264 1, ,744 22,296 53,480 69,067 32,357 3,142 1,850 7,006 1, ,300 23,401 51,921 75,142 45,751 2,916 1,668 7,104 1,936 These statstcs are based upon 75 REITs and 416 observatons. Total assets are book values.

8 370 THE JOURNAL OF REAL ESTATE RESEARCH Exhbt 4 Average Total Assets by Property Type and Sze Quartle Panel A: By Property Type Average Total Assets by REIT Property Type ($ mllon) Offce Warehouse Retal Apartment Dversfed REITs REITs REITs REITs REITs wt. avg Panel B: By Sze Quartle Average Total Assets by Sze Quartle ($ mllon) Frst Second Thrd Fourth Quartle Quartle Quartle Quartle wt. avg These statstcs are based upon 75 REITs and 416 observatons. Total assets are book values. REITs are below average. The average small REIT (frst quartle) has $29 mllon n assets. The large REITs (fourth quartle) average $279 mllon. Property Characterstcs Total property assets average $95 mllon per REIT over the perod. The average occupancy rate of the propertes, reflectng general market trends, falls from 92% n 1985 to 87% n 1991 before recoverng to 89% n The average cap rate rses from 8.8% to 9.4% also reflectng market trends. Concentraton /Dversfcaton An mportant ssue n the valuaton of REITs by Wall Street s the degree of dversfcaton. Our measures of dversfcaton are the Hrschman-Hrfndahl ndces VOLUME 10, NUMBER 4, 1995

9 PROPERTY TYPE, SIZE AND REIT VALUE 371 Exhbt 5 Value-Weghted Premums Value-Weghted (Avg.) Premum (Dscount) (%) (4.7) 1990 (35.9) 1991 (23.4) 1992 (9.7) Wt. Avg. (8.3) whch show conflctng trends. Whle concentraton by property type ncreased (61 to 70), concentraton by regon declned (60 to 46) durng the perod. Wall Street vs. Man Street Our data sample provdes the frst drect evdence on REIT valuatons n publc markets n Exhbt 5. Valuatons of REIT stocks fluctuate wdely from the value of the underlyng propertes durng the sample perod. On average REITs have traded at a dscount of 8% from the net asset values of the propertes; however the value-weghted premum by year vares from a hgh of 13% n 1986 to a low of 36% n Wall Street s wllngness to pay for securtzed property fell by over 40% n just four years before recoverng n The next two sectons test whether there are sgnfcant dfferences among the REIT property-type and sze categores. In partcular t s tested whether the premums to net asset value dffer sgnfcantly among types and f any dfferences n premums to net asset value spll over nto cash flow yelds. If the premum dfferences are justfed by the cost or other dfferences, cash flow yelds should not dffer sgnfcantly. REITs by Property Type Premum to Net Asset Value To try to deepen our understandng of why REIT premums vary, Exhbt 6 provdes cross-tabs on some key varables by REIT property-type category. Each panel provdes F- tests for sgnfcant varaton n the relevant varable by year and by property type. In addton, tests for whether any of the category averages are sgnfcantly dfferent from the overall sample mean are ndcated by astersks. In Panel A the premums and dscounts from net asset value are lsted. Warehouse REITs are dscounted the most heavly whle retal REITs are least dscounted. The dfferences from the sample averages are sgnfcant for these two categores. Panels B to E explore possble sources for these dfferences n the data on leverage, concentraton and overhead expenses.

10 372 THE JOURNAL OF REAL ESTATE RESEARCH Panel A: Value-Weghted Premum (%) Exhbt 6 REITs by Property Type (Avg.) Offce Warehouse Retal Apartment Dversfed (1.6) (23.6) (4.2) (7.4) (4.2) (9.9) (24.4) (1.5) (37.1) (1.7) (1.8) (12.5) 1990 (31.6) (44.5) (32.7) (63.1) (29.7) 1991 (34.1) (54.7) (8.1) (26.4) (30.0) 1992 (38.2) (55.5) (10.6) Avg. (7.5) (24.2)(**) (1.1)(*) (10.7) (8.7) ly effect F-value = 5.1 Prob. >F =.00 Property-type effect F-value = 2.0 Prob. >F =.12 Panel B: Leverage Rato (%) (Avg.) Offce Warehouse Retal Apartment Dversfed Avg. 31(**) (***) 32(**) ly effect F-value = 4.57 Prob. >F =.00 Property-type effect F-value = 9.82 Prob. >F =.00 Panel C: Property Concentraton Index (%) (Avg.) Offce Warehouse Retal Apartment Dversfed Avg. 69.2(***) 68.3(**) 71.6(***) (***) ly effect F-value =.65 Prob. >F =.71 Property-type effect F-value = Prob. >F =.00 VOLUME 10, NUMBER 4, 1995

11 PROPERTY TYPE, SIZE AND REIT VALUE 373 Panel D: Regonal Concentraton Index (%) (Avg.) Offce Warehouse Retal Apartment Dversfed Avg (***) (***) 52.3(**) ly effect F-value = 1.23 Prob. >F =.32 Property-type effect F-value = Prob. >F =.00 Panel E: General and Admnstratve Expenses/Total Assets (%) (Avg.) Offce Warehouse Retal Apartment Dversfed Avg (**).8(***) 1.6(***) ly effect F-value =.53 Prob. >F =.80 Property-type effect F-value = 6.04 Prob. >F =.00 Panel F: Cash Flow Yeld (%) (Avg.) Offce Warehouse Retal Apartment Dversfed Avg ly effect F-value = 9.99 Prob. >F =.00 Property-type effect F-value =.83 Prob. >F =.52 We classfy each REIT by property type when more than 50% of the property held s of one type. If no one property type s more than 50% of the REIT t s classfed as dversfed. Value-weghted premum s defned n the paper. Leverage rato s defned as total lablty/(total lablty+market value of the equty). The ndces are Hrschman-Herfndahl ndces. We mplement two-way analyss of varance (wthout nteracton terms) procedure to see f there exst yearly effects or property-type effects, and then test f each property-type s mean s dfferent from the overall mean. *, **, *** denotes two-taled sgnfcance at the 10%, 5%, 1% level, respectvely.

12 374 THE JOURNAL OF REAL ESTATE RESEARCH Leverage Panel B shows that apartment REITs are sgnfcantly more hghly levered than average whle dversfed and offce REITs are below average durng the sample perod. The F- statstcs for both the yearly effect and the property-type effect are sgnfcant. Concentraton Panels C ad D provde the concentraton ratos. As mght be expected dversfed REITs show low concentraton levels by property type. Offce, warehouse and retal REITs are sgnfcantly more concentrated by property type than the average REIT. The yearly dfferences are not sgnfcant. By regon the pattern s qute dfferent. Warehouse and dversfed REITs have low focus by regon whle apartment REITs are hghly focused. Agan the yearly dfferences are not sgnfcant. Expenses Panel E fnds that expense levels are hgh for dversfed REITs and low for retal and apartment REITs. The yearly dfferences are not sgnfcant. Cash Flow Yeld Whle some REIT types are more heavly dscounted from the value of the underlyng property, one mght expect that n effcent markets these dscounts would reflect lower earnngs potental. Cash flow s one measure of earnngs potental. Cash flow yelds, however, mght also reflect hgher rsk levels. The prevous panels show that apartment REITs are more hghly leveraged and that dversfed REITs have hgher expenses. Apartment REITs are also less dversfed and carry more local market rsk n addton to the hgher fnancal rsk. Therefore one mght expect to see hgher cash flow yelds on apartment REITs. Panel F dsplays the cash flow yelds by type. The dfferences among types are not sgnfcant. Therefore there s no evdence that NAV premums or rsk levels are affectng the cash flow yelds. However, the yearly varaton s sgnfcant. To summarze ths secton, retal REITs sell at a premum whle warehouse REITs trade at dscounts relatve to the average REIT and these dfferences are statstcally sgnfcant. Retal REITs are sgnfcantly more focused by property type and carry sgnfcantly less overhead expense. Warehouse REITs are also focused by property type but are sgnfcantly more dversfed by regon. Warehouse REITs also have above average expenses but not sgnfcantly so. None of the REIT types have sgnfcantly dfferent cash flow yelds. Sze Quartles It s well known that many equty market anomales are related to sze. For example Banz (1981) showed that returns from buyng very small frms are 20% hgher than for very large frms. Roll (1981) and Renganum (1981) present evdence that the small frm VOLUME 10, NUMBER 4, 1995

13 PROPERTY TYPE, SIZE AND REIT VALUE 375 effect s partly due to errors n estmatng the rsk (beta) of small frms; but the effect remans even when the estmaton problems are corrected. Stoll and Whaley (1983) argue that, gven the dfferences n transactons costs between small and large frms, a roundtrp transactons cost every three months s enough to elmnate the small frm effect. Kem (1983) provdes evdence that 25% of the sze effect occurs durng the frst fve tradng days n January. Ths secton nvestgates whether sze s related to other REIT characterstcs. Premum to Net Asset Value Panel A of Exhbt 7 llustrates the dramatc effect of sze on the premum to net asset value of the propertes. There s a monotonc ncrease n the premum n the larger sze quartles. Small REITs (frst quartle, average $29 mllon) are dscounted 33% more than large REITs (fourth quartle, average $279 mllon). Both the sze effect and the yearly effect are hghly sgnfcant. Leverage Panel B shows that large REITs (fourth quartle) are more hghly leveraged than the small REITs (frst quartle). The yearly effect s also sgnfcant. Concentraton Panel C: Small REITs (quartle 1) are sgnfcantly more hghly concentrated by property type. The next two quartles are less concentrated than the sample. There are no sgnfcant dfferences ether by sze or year n the regonal concentraton ndces (Panel D). Expenses Panel E: Small REITs are almost twce as costly to admnster as large REITs. The G&A rato for small REITs s 1.7% whle for large REITs t falls to.9%. Ths may account for some of the large dscount from net asset value for small REITs. The yearly dfferences are not sgnfcant. Cash Flow Yeld Panel F: Snce small REITs have lower fnancal rsk (less leverage) and less local market rsk (more dversfed by regon), one mght expect lower cash flow yelds. Instead, cash flow yelds for small REITs are hgher than for other REITs but not sgnfcantly so. The yearly dfferences, on the other hand, are sgnfcant. To summarze ths secton, small REITs are heavly dscounted (33%) relatve to large REITs. These small REITs are less levered, more focused by property type, and have much hgher overhead expenses ratos than large REITs. Thus, a consstent pattern n both the analyss by property type and by sze s that hghly dscounted categores are also categores wth hgh expense ratos.

14 376 THE JOURNAL OF REAL ESTATE RESEARCH Panel A: Value-Weghted Premum (%) Exhbt 7 REITs by Sze Quartles (Avg.) Frst Quartle Second Quartle Thrd Quartle Fourth Quartle (13.7) (4.7) (1.8) (1.3) (9.9) 3.3 (1.6) (14.9) (31.0) (7.1) (42.9) (42.7) (28.1) (32.1) 1991 (51.5) (34.4) (20.4) (3.6) 1992 (49.9) (26.4) Avg. (24.7)(***) (14.4)(**) (3.7) 8.4(***) ly effect F-value = 9.22 Prob. =.00 Sze effect F-value = Prob. >F =.00 Panel B: Leverage Rato (%) (Avg.) Frst Quartle Second Quartle Thrd Quartle Fourth Quartle Avg. 30(**) (**) ly effect F-value = 4.36 Prob. =.00 Sze effect F-value = 2.71 Prob. >F =.07 Panel C: Property Concentraton Index (%) (Avg.) Frst Quartle Second Quartle Thrd Quartle Fourth Quartle Avg. 72.8(***) 60.6(**) 59.2(**) 65.5 ly effect F-value = 1.08 Prob. >F =.41 Sze effect F-value = 6.41 Prob. >F =.00 VOLUME 10, NUMBER 4, 1995

15 PROPERTY TYPE, SIZE AND REIT VALUE 377 Panel D: Regonal Concentraton Index (%) (Avg.) Frst Quartle Second Quartle Thrd Quartle Fourth Quartle Avg ly effect F-value =.82 Prob. >F =.58 Sze effect F-value =.35 Prob. >F =.79 Panel E: General and Admnstratve Expenses/Total Assets (%) (Avg.) Frst Quartle Second Quartle Thrd Quartle Fourth Quartle Avg. 1.7(***) (***) ly effect F-value =.99 Prob. >F =.46 Sze effect F-value = Prob. >F =.00 Panel F: Cash Flow Yeld (%) (Avg.) Frst Quartle Second Quartle Thrd Quartle Fourth Quartle Avg ly effect F-value = 9.85 Prob. >F =.00 Sze effect F-value = 1.20 Prob. >F =.33 The sze segmentaton uses quartles derved from total assets. Value-weghted premum, Hrschman-Herfndahl ndex and cash flow yeld are defned n the paper. Leverage rato s defned as total lablty/(total lablty+market value of equty). We mplement two-way analyss of varance (wthout nteracton terms) procedure to see f there exst yearly effects or sze effects, and then test f each quartle s mean s dfferent from the overall mean. *, **, *** denotes two-taled sgnfcance at the 10%, 5%, 1% level, respectvely.

16 378 THE JOURNAL OF REAL ESTATE RESEARCH Conclusons Ths paper develops an equty REIT database and provdes descrptve statstcs on REIT property holdngs. It then categorzes REITs by property type and sze to explore sources of dfferences n valuaton. Leverage, dversfcaton and overhead expenses are nvestgated as possble causes of dscounts to net asset value. Other possble causes of dscounts such as external/nternal advsor, type of debt (fxed versus varable), or dvdend yeld have not been examned. It s shown that expenses as measured by the rato of G&A to total assets, remaned constant durng the perod; but as mght be expected, dversfed REITs and small REITs have above average expense ratos. Leverage rose durng the perod. Large REITs and apartment REITs supported more leverage than average. By property type, apartment REITs are the most concentrated by locaton. Small REITs are more focused by property type. The stock market valuatons as measured by premums above the values of the underlyng propertes declned durng the perod. Warehouse REITs and small REITs sell at sgnfcant dscounts from net asset value relatve to the average REIT. Retal REITs sell at premums relatve to the average REIT. Cash flow yelds, on the other hand, are not sgnfcantly dfferent among REITs. Therefore there s no evdence that dfferences n premums to net asset value affect cash flow yelds. If we take cash flow to be a coarse measure of expected return to shareholders, 8 then the evdence n ths sample suggests that Wall Street s correctly processng the nformaton that leads to the dscounts/ premums. As ndcated above, other varables not ncluded n ths study could affect the results. These results may help to explan why retal property s overrepresented and warehouse/ndustral s underrepresented n REITs. Retal property, once securtzed, often sells at a premum whle the opposte s the case for warehouse/ndustral. The dscounts/premums do not affect cash flow yelds at statstcally sgnfcant levels. Therefore, securtzaton adds value to retal property but destroys value for warehouse/ndustral property. Ths suggests that ether Wall Street dsagrees wth the valuatons on Man Street or the synerges that arse n a retal portfolo are greater than those n an ndustral portfolo. Notes 1 The look through provson n the 1993 OBRA tax revson whch effectvely elmnates the fve or fewer rule for penson fund nvestors has also contrbuted to the growth of REITs. 2 The lterature n the excellent revew of REIT research by Corgel, McIntosh and Ott (1995) s volumnous but overlooks ths mportant ssue, undoubtedly because of lack of data. A frequent thread n the lterature s return and performance, e.g., Kuhle (1987), Kuhle, Walther and Wurtzebach (1986), Myer and Webb (1993, 1994). Unlke the closed-end fund lterature where dscounts/premums have been carefully studed, the ssue s overlooked wth REITs. 3 These data are explaned n detal n the Market Hstory Reports. Brefly, the NREI data, unlke the NCREIF data, are based on actual sales rather than apprasals. Cap rates are derved from pro forma net operatng ncome. Property transactons are standardzed to meet prespecfed property norms n order to ensure that transacton trends of comparable qualty property transactons are beng reported. The ndex reports average values for each property type analyzed. 4 The approxmaton works well for cap rates n the typcal 8% 10% range that s observed. VOLUME 10, NUMBER 4, 1995

17 PROPERTY TYPE, SIZE AND REIT VALUE Note that the property portfolo cap rate s not the same as the captalzaton for the common stock. The captalzaton rate for equalty can be qute dfferent, e.g., because of leverage or other factors. 6 Notce that usng net asset values as a dependent varable does not volate the assumptons of OLS and does not ntroduce a bas. 7 The ndex frst acqured the name of Orrs Herfndahl from work on energy n the 1950s and that of Albert Hrschman from work on foregn trade patterns. See Hrschman (1964). 8 Ths s a coarse measure because t gnores dfferences n expected growth rates. References Banz, R. W., The Relatonshp between Return and Market Value of Common Stock, Journal of Fnancal Economcs, 1981, 9, Corgel, J., W. McIntosh and S. Ott, Real Estate Investment Trusts: A Revew of the Fnancal Economcs Lterature, Journal of Real Estate Lterature, 1995, 3, Goebel, P. and C. Ma, Stock Returns and Busness Performance n REITs, workng paper, Texas Tech Unversty, Hartzell, D. J., D. G. Shulman and C. H. Wurtzebach, Refnng the Analyss of Regonal Dversfcaton for Income-Producng Real Estate, Journal of Real Estate Research, 1987, 2:1, Hrschman, A. O., The Paternty of an Index, Amercan Economc Revew, 1964, 54, 761. Kem, D. B., Sze Related Anomales and Stock Return Seasonalty: Further Emprcal Evdence, Journal of Fnancal Economcs, 1983, 12, Kuhle, J. L., Portfolo Dversfcaton and Return Benefts Common Stock vs. Real Estate Investment Trusts, Journal of Real Estate Research, 1987, 2:2, 1 9., C. H. Walther and C. H. Wurtzebach, The Fnancal Performance of REITs, Journal of Real Estate Research, 1986, 1:1, Myer, F. C. N. and J. R. Webb, Return Propertes of Equty REITs, Common Stocks, and Commercal Real Estate: A Comparson, Journal of Real Estate Research, 1993, 8:1, , Retal Stocks, Retal REITs, and Retal Real Estate, Journal of Real Estate Research, 1994, 9:1, Renganum, M. R., Msspecfcaton of Captal Asset Prcng: Emprcal Anomales Based on Earnngs Yelds and Market Values, Journal of Fnancal Economcs, 1981, 9, , The Anomalous Stock Market Behavor of Small Frms n January: Emprcal Tests for Tax-Loss Sellng Effect, Journal of Fnancal Economcs, 1983, 12:1, Roll, R., A Possble Explanaton of the Small Frm Effect, Journal of Fnance, 1981, 31, Shllng, J., C. F. Srmans and J. Wansley, Do REIT Shares Trade at a Dscount? Some Indrect Evdence, workng paper, Lousana State Unversty, Stoll, H. R. and R. E. Whaley, Transacton Costs and the Small Frm Effect, Journal of Fnancal Economcs, 1983, 12:1, Webb, J. R. and W. McIntosh, Real Estate Investment Acquston Rules for REITs: A Survey, Journal of Real Estate Research, 1986, 1:1, We thank S. Mchael Glberto, the revewers and the edtors for ther helpful comments. The usual dsclamer apples.