An Exploration of the Relationship between Size and Risk in a Downside Risk Framework Applied to Malaysian Property Shares

Transcription

1 12 th Pacc Rm Real Estate Socety Conerence January 2006 An Exploraton o the Relatonshp between Sze and Rsk n a Downsde Rsk Framework Appled to Malaysan Property Shares PACIFIC RIM REAL ESTATE SOCIETY (PRRES CONFERENCE AUCKLAND NEW ZEALAND Chy Ln Lee, Jon Robnson and Rchard Reed c.lee27@pgrad.unmelb.edu.au; jrwr@unmelb.edu.au and r.reed@unmelb.edu.au Faculty o Archtecture, Buldng and Plannng The Unversty o Melbourne Melbourne 3010, Australa Abstract The relatonshp between sze and rsk (systematc and unsystematc rsk has receved consderable attenton n recent lterature. However, these studes employ varance as the rsk measure, whch the approprateness or usng ths rsk measure s always questoned by researchers and practtoners due to ts underlyng strct assumptons. Thereore, there s crucal to adopt an alternatve rsk measure or ascertanng the relatonshps. The am o the study s to examne the relatonshps between sze and systematc downsde rsk and unsystematc downsde rsk n lne wth the theoretcal sound o ths rsk measure. The emprcal evdences reveal that the sze s strongly correlated wth unsystematc downsde rsk. Whle, there s a weak nverse relatonshp between sze and systematc downsde rsk. Keywords: systematc downsde rsk, unsystematc downsde rsk, sze, property shares 1. Introducton Snce the ntroducton o Modern Portolo Theory by Markowtz (1952, dverscaton has become the man ssue or nvestors, partcularly nsttutonal nvestors, n makng nvestment decsons. Consequently, Sharpe (1963 proposed a lnear model that provdes a better understandng or nvestors on rsk and dverscaton n whch the total rsk o asset s decomposed nto systematc rsk and unsystematc rsk. Systematc rsk s an undversable rsk, whch s attrbuted wth a common actor. On the other hand, unsystematc rsk s a dversable rsk, whch s also known as specc rsk and t can be totally elmnated va a well dversed portolo (Brown & Matysak, 2000, Hargtay & Yu, Recently, there have been many attempts to demonstrate the relatonshp between sze and rsk. For nstance, Conover et al. (1998 ound that derences between large rms and small rms n terms o return and rsk are statstcally sgncant. Ltt et al. (1999 1

2 12 th Pacc Rm Real Estate Socety Conerence January 2006 provded a plausble explanaton or ths scenaro, whch revealed that sze has a moderate negatve correlaton wth unsystematc rsk. Ths s consstent wth ndngs by Byrne & Lee (2003, Chaudhry et al. (2004 and Malkel & Xu (1997, who they ound a negatve relatonshp between sze and specc rsk. Conversely, Gyuorko & Nellng (1996 ound a postve and statstcally sgncant relatonshp between sze and systematc rsk. Byrne & Lee (2003 also ound a postve and statstcally sgncant relatonshp between sze and systematc rsk. Interestngly, ths relatonshp became weak and negatve when nvestment characterstcs were controlled. Smlarly, Ltt et al. (1999 also ound a weak negatve correlaton between beta and sze. Thereore, there s no consensus or a relatonshp between sze and systematc rsk. One o the possble explanatons s that these studes employ varance as the rsk measure. Varance s currently the most wdely used measure o portolo rsk and t has receved the consderable attenton rom researchers and practtoners (Evans, But the use o varance as a rsk measure s constraned by several strct assumptons: nvestors have a constant quadratc utlty uncton and asset return dstrbutons are normally dstrbuted. Many nance and real estate lterature have rejected both assumptons (Lee et al., A survey conducted by Mao (1970 reported that nvestors are more concerned about the probablty o return beng lower than a target rate o return. In other words, nvestors dslke downsde volatlty and the man concern or most nvestors s downsde rsk, whch s the lkelhood o returns allng below a target rate o return (Byrne & Lee, Markowtz (1959 also recognsed the mportance o ths argument and suggested the use o sem-varance, or lower partal moment, whch s more appealng than varance. As a result, downsde rsk appears to be a more robust and sensble rsk measure than varance; t s suggested that t used n portolo analyss (Lee et al., The approprateness o usng downsde rsk was also demonstrated by Hogan & Warren (1974 and Bawa & Lnderberg (1977. They generalsed the downsde rsk (lower partal moment nto the Captal Asset Prcng Model (CAPM and developed a mean-lower partal moment captal asset prcng model (D-CAPM. Nantell & Prce (1979 and Prce et al., (1982 examned the derence between the systematc rsks that are derved n a downsde rsk ramework and a mean varance ramework. Ther ndngs depcted that systematc rsk n a downsde rsk ramework ders rom systematc rsk n a mean varance ramework the return dstrbutons are n lognormal orm. A growng body o research has also demonstrated the superorty o systematc downsde rsk than tradtonal systematc rsk. These studes also suggested the use o downsde beta (systematc downsde rsk as an alternatve or tradtonal beta n portolo management (Bhardwaj & Brooks, 1993, Cheng, 2005, Chang et al., 2004, Conover et al., 2000, Estrada, 2000, 2002, Harlow & Rao, However, there s lmted lterature about the relatonshp between sze and downsde rsk. Devaney & Lee (2005 probably conducted the rst study that examned the relatonshp 2

3 12 th Pacc Rm Real Estate Socety Conerence January 2006 between real estate portolo sze and downsde rsk. Ther study replaced varance by downsde rsk due to the hypothess that the rsk or a und manager s the rsk o underperormance o the benchmark rather than the volatlty o returns o the portolo. They used Monte Carlo smulaton and the returns rom 1728 propertes over the perod n the Investment Property Databank (IPD database. The results ndcated that the ncrease n portolo sze reduced the portolo s downsde rsk (total rsk. Importantly, to date, there s no study o the mplcatons o sze n reducng systematc rsk and unsystematc rsk n a downsde rsk ramework. The am o ths study s to examne the relatonshp between sze and rsk (systematc and unsystematc rsks n a downsde rsk ramework. In Sectons 2 and 3, the concept o systematc downsde rsk and unsystematc downsde rsk are provded. Secton 4 dscusses the data and then ntroduces the methodology used n ths study. The next secton emprcally tests the relatonshp between sze and systematc rsk and unsystematc rsk n a downsde rsk ramework. Secton 6 summarses the ndngs and provdes a concluson. 2. The Concept o Systematc Downsde Rsk In the Mean Varance Analyss ramework, Sharpe (1964 proposed that the expected return o asset n CAPM s estmated as: COV ( R, Rm E R = R + ( E( Rm R Var( R ( m (1 where COV ( R, Rm E( R = Covarance o returns on the market portolo wth returns on securty = Expected return on asset E( Rm = Expected return on the market portolo Var( Rm = Varance o returns on the market portolo R = Rsk ree rate o return usually smpled and presented as ollow: ( E( R R E( R = R + β m (2 where β s the beta, whch s computed by usng ( R R / Var( R COV. 1, m m In the downsde rsk ramework, semvarance s used as the rsk measure. Bawa & Lnderberg (1977, Harlow & Rao (1989 and Hogan & Warren (1974 generalsed 1 See Elton & Gruber (1995 and Estrada (2002 or the detals o beta computaton. 3

4 12 th Pacc Rm Real Estate Socety Conerence January 2006 downsde rsk nto CAPM and proposed that the expected return o the asset n Mean Lower Partal CAPM (or D-CAPM s estmated rom the ollowng: CVS R ( R, R m E R = R + ( E( Rm R SV ( R ( R m (3 where R ( R Rm CVS, SV ( R Rm CVS R, R SV R R ( ( R m m = Cosemvarance o returns on the market portolo wth returns on securty = Semvarance o returns below R on the market portolo D β = Downsde Beta ( The expected return o the asset n Equatons (1 and (3 s computed n exactly n the same way n both rameworks except or the estmaton o beta. As such, beta n the downsde rsk ramework s dened as CVS R, R / SV ( R and Equaton (3 can be smpled as ollows: ( m R ( m R m D E R = R + β ( E( R R (4 where D β s the downsde beta. However, Estrada (2002 hghlghted the lmtatons o the exstng estmaton o downsde beta, whch Bawa & Lnderberg (1977, Harlow & Rao (1989 and Hogan & Warren (1974 dened as co-semvarance as ollows: {( R R. Mn[ ( R R,0 ]} CVSR ( R, Rm = E M (5 Clearly, Equaton (5 s n asymmetry orm. Thus, the cosemvarance between two assets and M s derent rom the cosemvarance between asset M and (Estrada, 2000, 2002, Lee et al., 2005, Nawrock, As a result, Estrada (2002 proposed a replacement o the CAPM beta by the rato below (n symmetrc orm n order to obvate the lmtatons o the Equaton (5: β D { Mn[ ( R µ ] Mn[ ( RM µ M,0]} E Mn[ ( R µ,0] 2 E = (6 { } M M where R s return o asset at the tme t, µ denotes the rsk ree rate or the rate o return o benchmark. 4

5 12 th Pacc Rm Real Estate Socety Conerence January The Concept o Unsystematc Downsde Rsk One o the essental propertes o CAPM s to acltate rsk decomposton and quantcaton n whch total rsk can be decomposed nto two orthogonal components: a market rsk and rm-specc resdual. A tme-seres regresson or CAPM s as ollows: R R = α + β ( R R + ε (7 M where R = return o asset at the tme t, R = rsk-ree rate, α = non-ndex related return to asset, β = beta o asset, RM = return to market ndex at the tme t, resdual return o asset at the tme t. Takng the varance o both sdes; the varance o returns on asset s shown as: 2 Var( R = β Var( R + Var( ε (8 M ε = where Var ( R s the varance o the asset, Var( ε unsystematc rsk. 2 s volatlty measure or As dscussed above, the computaton o CAPM and D-CAPM are the exactly same except or beta. Thereore, the unsystematc downsde rsk can also be estmated by amendng Equaton (8 n whch total rsk s substtuted by total downsde rsk; beta s changed to downsde beta; varance o the market portolo s replaced by semvarance o the market portolo and unsystematc rsk s substtuted by unsystematc downsde rsk. The relatonshps between total downsde rsk, systematc downsde rsk and unsystematc downsde rsk are wrtten as ollow: SV R D 2 ( R = β SV ( R + SV ( ε (9 R m R Where SV R ( R SV R ( R m SV R ( e = Total Downsde Rsk (Semvarance = Semvarance o market portolo = Unsystematc downsde rsk 4. Data and Methodology 4.1 Data In ths study, annual returns rom Malaysan Property Shares (PSs were utlsed. As at 31 st December 2003, there were 90 property companes lsted on Bursa Malaysa (ormerly known as Kuala Lumpur Stock Exchange. From whch 30 lsted property 2 See Beckers (1996, Chaudhry et al. (2004 and Sanders et al. (2001 or the detals. 5

6 12 th Pacc Rm Real Estate Socety Conerence January 2006 shares have been selected n ths study. The analyss spans the tme perod n whch covers the boom (1993 and recesson (1997 phases o the most recent property shares cycle n Malaysa. It must be noted that Bursa Malaysa experenced hghly speculatve actvty n 1993; the daly turnover surged to RM4.8 bllon on 22 December 1993 (Central Bank o Malaysa, 1994b. There was a steep rse n share prces due to the speculatve actvtes (Central Bank o Malaysa, 1994a. In addton, the monthly returns or most PSs n 1993 were exceptonally hgh, or example, the monthly return o AHPLANT n October 1993 at %. Ths aected the stablsaton o the money market and the exchange rate (Department o Valuaton and Property Servces Malaysa, Hence, annual data s utlsed n ths study n order to avod the basng eect o the speculaton perod or the stock market n 1993 and better relect the true potental perormance o PSs. Addtonally, market captalsatons o ndvdual PSs were employed as ndcators o sze and Property Stock Index was used as the market benchmark. All o these data were obtaned rom Bursa Malaysa. 3-month Treasury Bll (TB was used as the rsk-ree rate, whch was obtaned rom Central Bank o Malaysa. 4.2 Methodology As dscussed above, the ratonale o usng varance s doubtul the asset return dstrbutons are not symmetrcally or normally dstrbuted. As such, t s crucal to examne the asset return dstrbuton to establsh whether return s normally or asymmetrcally dstrbuted. Several tests can be used n order to examne the normalty o asset return dstrbutons. In ths study, skewness, kurtoss and Jera-Bera Test were used or examnng the PSs return dstrbutons. Ths s consstent wth the methodology o Brown & Matysak (2000 and Kshore (2004. Skewness Skewness characterses the degree o asymmetry o a dstrbuton around ts mean. Zero skewness ndcates the dstrbuton s symmetry. Whereas, postve skewness ndcates a dstrbuton wth an asymmetrc tal extendng toward rght. Conversely, negatve skewness ndcates a dstrbuton wth an asymmetrc tal extendng toward let. Skewness s computed as ollow: S 1 = T 1 T = 1 ( R R 3 σ 3 (10 where R s the return or asset, R s the mean return o asset, σ s the standard devaton o asset and T s total number o returns. 6

7 12 th Pacc Rm Real Estate Socety Conerence January 2006 Kurtoss Kurtoss descrbes the degree o latness or peakedness o an asset return dstrbuton. Kurtoss s estmated rom the ollowng: K 1 = T 1 T = 1 ( R R 4 σ 4 (11 where R s the return or asset, R s the mean return o asset, σ s the standard devaton o asset and T s total number o returns. Jarque-Bera Test Jarque & Bera (1980, 1987 proposed a combnaton o skewness and kurtoss (s also known as the Jarque-Bera Test n order to examne the normalty o a dstrbuton. The Jarque-Bera statstc s computed as ollow: 2 n 2 ( K 3 JB = S + ( where S s a measure o skewness, K s a measure o kurtoss and n s the sample sze. The Jarque-Bera statstc has a ch-squared dstrbuton wth two degrees o reedom (one or skewness, one or kurtoss. 4.3 Downsde Rsk The downsde rsk (total rsk can be estmated by Lower Partal Moment, whch s dened by Bawa (1975 and Fshburn (1977: τ ( R = α LPM α τ, ( τ R df( R = 1 T [ Max(0,( R t ] α T 1 t= 1 τ (13 where df (R s the cumulatve dstrbuton uncton o the nvestment return R, τ s the target return, α s the degree o the LPM, R s the return o asset and T s total number o returns. Notably, sem-varance s a specal case o the more general LPM n whch the α value s equal to 2. Thus, t can also be reerred to as a target sem-varance (Harlow, In ths study, the degree o the LPM s equal to 2 n order to estmate the semvarance. Furthermore, theτ, target rate s set as the rsk-ree rate. 7

8 12 th Pacc Rm Real Estate Socety Conerence January Regresson Model In ths study, two regresson models were employed n order to ascertan the relatonshp between sze and systematc downsde rsk and unsystematc downsde rsk. Both systematc downsde rsk and unsystematc downsde rsk are regressed aganst sze. The regresson models are estmated as ollows: Log ( SystematcDownsdeRsk = α + β Log( Sze + (14 e Log ( UnsystematcDownsdeRsk = α + β Log( Sze + (15 e where systematc downsde rsk ( β D and unsystematc downsde rsk are derved rom Equaton (6 and Equaton (9 respectvely, sze s the actual sze o the PSs; α, β and are estmated rom the models. e 5. Results and Analyss 5.1 The Dstrbuton o the Asset Table 1 reveals the dstrbuton o derent sze groups (Small, Medum and Large rom 1992 to In general, all derent PSs sze groups exhbted postve skewness. Interestngly, all ndvdual PSs also dsplayed a postve skewness except SIMEPTY ( and FIMACORP (0. 3 It s not surprsng that no sze group revealed a kurtoss wth 3; whle all sze groups dsplayed a kurtoss o more than 3 (leptokurtoss. The heterogenety n kurtoss was also apparent; the Medum group showed 6.94 kurtoss; whle the Large group exhbted 4.70 kurtoss. It s more notceable or ndvdual PSs, or example, PTGTIN had 9.37 kurtoss whereas SIMIPTY revealed kurtoss. Consstently, all sze groups were statstcally sgncant n the Jarque-Bera Test. As a result, the assumpton that these sze groups are normally dstrbuted should be rejected by the Jarque-Bera test. Surprsngly, all ndvdual PSs were asymmetrcally dstrbuted except FBO (1.54, SPB (2.07, BRAYA (2.74, PARAMOUNT (3.80, CRIMSON (3.91, IGB (4.02, FIMACORP (4.60, BOLTON (4.66 and IOI (5.86. These results are consstent wth the results or developed real estate markets such as Unted States (U.S., Unted Kngdom (U.K. and Australa where return dstrbutons o real estate (ncludng securtsed and unsecurtsed real estate are not necessarly 3 See Appendx 1 or the detals o ndvdual PSs. 8

9 12 th Pacc Rm Real Estate Socety Conerence January 2006 normally dstrbuted (Gra et al., 1997, Kshore, 2004, Matland-Smth & Brooks, 1999, Myer & Webb, 1993, 1994, Peng, 2005, Young & Gra, In lne wth ths, the assumpton o Mean Varance Analyss and the ratonale o usng varance as rsk measure are questonable. Table 1: The Dstrbuton o Assets ( Group(s/Asset(s Skewness Kurtoss Jarque-Bera Test Small * Medum * Large * Note: (* sgncant at 5% level 5.2 Rsk Analyss Accordng to the Fgure 1, the volatltes o assets were substantally lower when downsde rsk was employed. The sem devatons o all sze groups were only 1/3 o the standard devatons or correspondng sze groups. These results are consstent wth the results o Peng (2005, Sng & Ong (2000 and Svtandes (1998 n whch the varance (standard devaton s hgher sgncantly than the downsde rsk or Australan Lsted Property Trusts, Sngapore real estate and U.S. REITs respectvely. Another pont must be notced s the Large group (80.42% has a smaller volatlty compared to Medum (97.30% and Small (103.84%. Smlarly n the downsde rsk ramework, the sem devaton o the Large group was 29.39%, whch was lower than Medum (29.82% and Small (34.77%. In other words, the Large PSs have lower rsk compared to smaller PSs. These results are consstent wth the results rom Gyourko & Nellng (2003 or standard devaton and Devaney & Lee (2005 or sem-devaton (total downsde rsk. Fgure 1: Standard Devaton and Sem-Devaton or Derent Sze Groups Standard Devaton and Sem-Devaton or Derent Sze Groups Rsk(% % % 80.00% 60.00% 40.00% 20.00% 0.00% 97.30% % 80.42% 29.39% 29.82% 34.77% Large Medum Small Sze Group Standard Devaton Sem Devaton 9

10 12 th Pacc Rm Real Estate Socety Conerence January Perormance Analyss Table 2 presents the perormance analyss o PSs n derent sze groups rom 1992 to Over ths 12-year perod, the average return o Large group (24.23% was slghtly hgher than Medum (22.61% and Small (20.96%. However, average return cannot smply be used or rankng purposes wthout ponderng rsk. As a result, n ths study, rsk-adjusted return was used or ths purpose and the rskadjusted returns were calculated usng the Sharpe Rato and the Sortno Rato. 4 Usng the Sharpe Rato, t was shown that the Large group was the best perormer group compared to the Small and Medum groups on a rsk-adjusted bass. Conversely, the Sortno Rato provded contradctory results, whch the Medum group acheved the hghest rskadjusted perormance whle the Small group was the worst perormer. Thus, the results rom Sharpe Rato and Sortno Rato are always contradctory. Ths can be attrbuted to derent rsk measures are employed by both ratos. Addtonally, ths s consstent wth the results rom Ells & Wlson (2005 and Stevenson (2001, whch Sharpe Rato and Sortno Rato exhbt dverge results. Table 2: Perormance Analyss ( Asset Average Return Sharpe Rato Sortno Rato (% Small ( (3 Medum ( (1 Large ( (2 Note: Fgures n parenthess are rsk-adjusted rankng. 5.4 The Downsde Beta and Unsystematc Downsde Rsk Table 3 presents the downsde beta (systematc downsde rsk and unsystematc downsde rsk o PSs. The downsde beta or the small sze group was Ths ndcates that when market return alls by 1%, ths group wll all by 1.13%. Thus, magnyng by 13% the downsde swngs n the market wth respect to the rsk ree rate. Conversely, the downsde betas or Medum and Large sze groups were 0.93 and 0.96 respectvely. In other words, on average, both groups wll all only 0.95% wth a 1% all n the market. Another mportant observaton s that there s no obvous relatonshp between downsde beta and sze. Moreover, the downsde beta or varous sze groups dd not vary notceably. However, the mpact o sze was ound aect unsystematc downsde rsk. The Large sze group has the lowest unsystematc downsde rsk (1.38%, whch was around hal o the 4 Sharpe Rato employs standard devaton as the rsk measure and can be estmated as: Sharpe Rato = (return o asset at the tme t the rsk-ree rate/standard devaton o asset at the tme t. Whereas Sortno Rato utlses sem devaton as the rsk measure and s computed as: Sortno Rato = (return o asset at the tme t the rsk-ree rate/sem devaton o asset at the tme t. 10

11 12 th Pacc Rm Real Estate Socety Conerence January 2006 Small group (2.18%. Whle, the systematc downsde rsk or Medum group was 1.79%. The overall pcture that emerges rom Table 3 s that Large PSs have a lower unsystematc rsk, but there s no smlar notceable evdence or the downsde beta. The mportance o sze to downsde beta and unsystematc downsde rsk reducton are stll vague. Hence, t s essental to nvestgate ths ssue n more depth. Table 3: Downsde Beta and Unsystematc Downsde Rsk Asset Downsde Beta Unsystematc Downsde Rsk (% Small Medum Large The Relatonshp between Sze and Rsk (Downsde Beta and Unsystematc Downsde Rsk Table 4 llustrates the emprcal relatonshp between sze and rsk. Consstent wth the results rom the prevous secton, the regresson model depcted that downsde beta has an nsgncant negatve relatonshp wth sze. In other words, the Large PSs have only a slghtly lower systematc downsde beta. There s no consderable reducton n systematc downsde rsk or nvestors by nvestng n Large PSs. Ths conrms the assertons o CAPM n whch only unsystematc rsk can be elmnated and aected by sze. Moreover, ths s also consstent wth results rom Ltt et al. (1999 or U.S. REITs and Byrne & Lee (2003 or U.K. property unds by controllng the nvestment characterstcs. But, t counters the ndngs rom Gyuorko & Nellng (1996, whch s analysed wth Mean Varance Analyss ramework. On the other hand, sze has a sgncant negatve correlaton wth the unsystematc downsde rsk ( Ths ndcates that doublng the sze o the PSs wll lead to a sgncant reducton n unsystematc downsde rsk o just over 32%. In other words, the smaller PSs have larger unsystematc downsde rsk. Ths s consstent wth the contentons o CAPM and ndngs rom Byrne & Lee (2003, Chaudhry et al. (2004, Ltt et al. (1999 and Malkel & Xu (1997, whch are analysed under the Mean Varance Analyss ramework. Interestngly, the sze coecents or downsde beta and unsystematc downsde rsk are qute close wth the sze coecents or beta and unsystematc rsk that were ound by Ltt et al. (1999 or U.S. REITs. One o the possble explanatons s they employed the Rsk-Adjusted Model or ther estmaton. 5 5 Ltt et al. (1999 reported that the sze coecents wth NAREIT Beta and rm-specc rsk were and respectvely. 11

12 12 th Pacc Rm Real Estate Socety Conerence January 2006 Table 4: The Emprcal Relatonshp between Sze and Rsk Varable Log(Downsde Beta Log(Unsystematc Downsde Rsk Constant ( ( Log(Sze ( ( R-square 3.93% 5.09% Note: Fgures n parenthess are Standard Errors. 6. Conclusons and Implcatons Ths paper nvestgates the relatonshp between sze and rsk (systematc rsk and unsystematc rsk n a downsde rsk ramework. Several mportant ndngs can be drawn rom the analyses. Frst, the Malaysan PSs return dstrbutons are lke other real estate (securtsed and unsecurtsed real estate n developed markets; they are not necessarly normally dstrbuted. As such, the ratonale o usng varance as the rsk measure s precluded the return dstrbutons are n asymmetrc orm and t strengthens the motvaton to use downsde rsk, partcularly or emergng markets. Second, the varance (standard devaton exhbts hgher rsk or nvestors. Ths can be attrbuted to the ncluson o upsde potental n rsk estmaton under the Mean Varance Analyss ramework, whch s not logcal. It s not surprsng that the rsk o Malaysan PSs, as an emergng market, s over-estmated consderably by usng varance. In lne wth ths, downsde rsk s suggested to be used as a rsk measure partcularly or emergng markets. Thrd, sze s negatvely correlated to the unsystematc rsk, whle there s no smlar evdence or systematc rsk n a downsde rsk ramework. Ths supports the assertons o CAPM n whch nvestors only can dversy ther unsystematc rsk through sze nvestment strategy but would not gan any systematc downsde rsk reducton va ths strategy. In other words, portolo und managers and nvestors can only gan the ull unsystematc downsde rsk dverscaton benet by nvestng n Large Market Captalsaton PSs. Ths probably s the rst study to explore the relatonshp between sze and rsk (systematc rsk and unsystematc rsk n a downsde rsk ramework n a real estate context and uture work s needed to conrm the relatonshp. Future research should be drected to mprovng the model by employng more actors. Larger samples should be employed to help researchers to more accurately assess the relatonshp between sze and systematc rsk and unsystematc rsk n a downsde rsk ramework. 12

13 12 th Pacc Rm Real Estate Socety Conerence January 2006 Reerences Bawa, V.S. (1975 Optmal Rules or Orderng Uncertan Prospects, Journal o Fnancal Economcs, 2(1, pp Bawa, V.S. & Lnderberg, E.B. (1977 Captal Market Equlbrum n a Mean-Lower Partal Moment Framework, Journal o Fnancal Economcs, 5(2, pp Beckers, S. (1996 A Survey o Rsk Measurement Theory and Practce, n: C. Alexander (Ed The Handbook o Rsk Management and Analyss (West Sussex, John Wley & Sons Ltd. Bhardwaj, R.K. & Brooks, L.D. (1993 Dual Betas rom Bull and Bear Markets: Reversal o the Sze Eect, Journal o Fnancal Research, 16(4, pp Brown, G.R. & Matysak, G.A. (2000 Real Estate Investment: A Captal Market Approach (Harlow, Fnancal Tmes Prentce Hall. Byrne, P. & Lee, S. (2003 An Exploraton o the Relatonshp between Sze, Dverscaton and Rsk n UK Real Estate Portolos: , Journal o Property Research, 20(2, pp Byrne, P. & Lee, S. (2004 Derent Rsk Measures: Derent Portolo Compostons?, Journal o Property Investment and Fnance, 22(6, pp Central Bank o Malaysa (1994a Annual Report o Central Bank o Malaysa 1994(Kuala Lumpur, Central Bank o Malaysa. Central Bank o Malaysa (1994b Economc Report 1994/1995(Kuala Lumpur, Central Bank o Malaysa. Chaudhry, M.K., Maheshwar, S. & Webb, J.R. (2004 REITs and Idosyncratc Rsk, Journal o Real Estate Research, 26(2, pp Cheng, P. (2005 Asymmetrc Rsk Measures and Real Estate Returns, Journal o Real Estate Fnance and Economcs, 30(1, pp Chang, K.C.H., Lee, M.-L. & Wsen, C.H. (2004 Another Look at the Asymmetrc REIT-Beta Puzzle, Journal o Real Estate Research, 26(1, pp Conover, C.M., Frday, H.S. & Howton, S. (1998 The Relatonshp between Sze and Return or Foregn Real Estate Investments, Journal o Real Estate Portolo Management, 4(2, pp Conover, C.M., Frday, H.S. & Howton, S.W. (2000 An Analyss o the Cross Secton o Returns or EREITs Usng a Varyng-Rsk Beta Model, Real Estate Economcs, 26(1, pp Department o Valuaton and Property Servces (1994 Property Market Report (Kuala Lumpur, Department o Valuaton & Property Servces, Malaysa. Devaney, S. & Lee, S. (2005 Real Estate Portolo Sze and Rsk Reducton. Workng Papers n Real Estate and Plannng 24/05 (Readng, Unted Kngdom, The Unversty o Readng. Ells, C. & Wlson, P.J. (2005 Can a Neural Network Property Portolo Selecton Process Outperorm the Property Market?, Journal o Real Estate Portolo Management, 11(2, pp Elton, E.J. & Gruber, M.J. (1995 Modern Portolo Theory and Investment Analyss (New York, John Wley & Sons, Inc. Estrada, J. (2000 The Cost o Equty n Emergng Markets: A Downsde Rsk Approach, Emergng Markets Quarterly, 4(1, pp Estrada, J. (2002 Systematc Rsk n Emergng Markets: The D-CAPM, Emergng Markets Revew, 3(4, pp Evans, J.L. (2004 Wealthy Investor Atttudes, Expectatons, and Behavors toward Rsk and Return, The Journal o Wealth Management, 7(1, pp Fshburn, P.C. (1977 Mean-Rsk Analyss wth Rsk Assocated wth Below-Target Returns, Amercan Economc Revew, 67(2, pp Gra, R.A., Harrngton, A. & Young, M.S. (1997 The Shape o Australan Real Estate Return Dstrbutons and Comparsons to the Unted States, Journal o Real Estate Research, 14(3, pp Gyuorko, J. & Nellng, E. (1996 Systematc Rsk and Dverscaton n the Equty REIT Market, Real Estate Economcs, 24(4, pp Hargtay, S.E. & Yu, S.-M. (1993 Property Investment Decsons: A Quanttatve Approach (London, E & FN Spon. Harlow, W.V. (1991 Asset Allocaton n a Downsde-Rsk Framework, Fnancal Analysts Journal, 47(5, pp

14 12 th Pacc Rm Real Estate Socety Conerence January 2006 Harlow, W.V. & Rao, R.K.S. (1989 Asset Prcng n a Generalzed Mean-Lower Partal Moment Framework: Theory and Evdence, Journal o Fnancal and Quanttatve Analyss, 24(3, pp Hogan, W.W. & Warren, J.M. (1974 Toward The Development o an Equlbrum Captal-Market Model Based On Semvarance, Journal o Fnancal and Quanttatve Analyss, 9(1, pp Jarque, C. & Bera, A. (1980 Ecent Tests or Normalty Homoscedastcty and Seral Independence o Regresson Resduals, Econometrc Letters, 6, pp Jarque, C. & Bera, A. (1987 A Test or Normalty o Observatons and Regresson Resduals, Internatonal Statstcal Revew, 55(2, pp Kshore, R. (2004 The Impact o Sze and Value Eects on Lsted Property Trust Perormance (Sydney, Unversty o Western Sydney. Lee, C.L., Robnson, J. & Reed, R. (2005 Mean Varance Analyss versus Downsde Rsk Framework: Revewng the Applcaton o Both Frameworks n Real Estate Markets. The 10th Asan Real Estate Socety (AsRES Internatonal Conerence (The Swss Grand Resort & Spa - Bond Beach, Sydney, Australa, Ltt, J., Me, J. & Webber, P. (1999 A Rsk Adjustment Model or REIT Evaluaton, Real Estate Fnance, 16(1, pp Matland-Smth, J.K. & Brooks, C. (1999 Threshold Autoregressve and Markov Swtchng Models: An Applcaton To Commercal Real Estate, Journal o Property Research, 16(1, pp Malkel, B.G. & Xu, Y. (1997 Rsk and Return Revsted, Journal o Portolo Management, 23(3, pp Mao, J.C.T. (1970 Survey o Captal Budgetng: Theory and Practce, Journal o Fnance, 25(2, pp Markowtz, H.M. (1952 Portolo Selecton, Journal o Fnance, 7(1, pp Markowtz, H.M. (1959 Portolo Selecton: Ecent Dverscaton o Investment Second Edton (Massachusetts, John Wley & Sons, Inc. Myer, F.C.N. & Webb, J.R. (1993 Return Propertes o Equty REITs, Common Stocks, and Commercal Real Estate: A Comparson, Journal o Real Estate Research, 8(1, pp Myer, F.C.N. & Webb, J.R. (1994 Statstcal Propertes o Returns: Fnancal Assets Versus Commercal Real Estate, Journal o Real Estate Fnance and Economcs, 8(3, pp Nantell, T.J. & Prce, B. (1979 An Analytcal Comparson o Varance and Semvarance Captal Market Theores, Journal o Fnancal and Quanttatve Analyss, 14(2, pp Nawrock, D. (1992 The Characterstcs o Portolos Selected by n-degree Lower Partal Moment, Internatonal Revew o Fnancal Analyss, 1(3, pp Peng, V. (2005 Rsk Measurements and Lsted Property Trusts Investment Strateges: Focusng on the Downsde, Pacc Rm Property Research Journal, 11(2, pp Prce, K., Prce, B. & Nantell, T.J. (1982 Varance and Lower Partal Moment Measures o Systematc Rsk: Some Analytcal and Emprcal Results, Journal o Fnance, 37(3, pp Sanders, A.B., Paglar, J.L., Jr. & Webb, J.R. (2001 Portolo Management Concepts and Ther Applcaton to Real Estate, n: J.L. Paglar, Jr. (Ed The Handbook o Real Estate Portolo Management (New York, McGraw-Hll Prms Custom Publshng. Sharpe, W. (1963 A Smpled Model or Portolo Analyss, Management Scence, 9(2, pp Sharpe, W. (1964 Captal Asset Prces: A Theory o Market Equlbrum under Condtons o Rsk, Journal o Fnance, 19(3, pp Sng, T.F. & Ong, S.E. (2000 Asset Allocaton n a Downsde Rsk Framework, Journal o Real Estate Portolo Management, 6(3, pp Svtandes, P.S. (1998 A Downsde-Rsk Approach to Real Estate Portolo Structurng, Journal o Real Estate Portolo Management, 4(2, pp Stevenson, S. (2001 Emergng Markets, Downsde Rsk and the Asset Allocaton Decson, Emergng Markets Revew, 2(1, pp Young, M.S. & Gra, R.A. (1995 Real Estate Is Not Normal: A Fresh Look at Real Estate Return Dstrbutons, Journal o Real Estate Fnance and Economcs, 10(3, pp

15 12 th Pacc Rm Real Estate Socety Conerence January 2006 Appendx 1 Table 6: Summary Statstcs PSs Return Standard Devaton Skewness Kurtoss JB Large SIMEPTY 11.76% 42.14% * HLPROP 28.62% % * L&G 25.60% % * IOI 56.61% % IGB 10.89% 40.86% BRAYA 27.36% 72.45% SPB 9.62% 34.17% PELANGI 6.29% 54.42% * FACBRES 21.89% % * AMDB 10.95% 85.78% * Medum PGARDEN 3.46% 32.87% * UMLAND 31.13% % * DBHD 15.19% % * MUIPROP 12.06% 68.98% * BOLTON 18.72% 73.87% NEGARA 12.89% 60.02% * AHPLANT 68.53% % * TALAM 30.73% % * PJDEV 18.01% 89.59% * SDRED 15.38% 92.79% * Small ASIAPAC 15.95% % * YTL 51.92% % * MENANG 25.50% % * FBO 2.66% 46.76% E&O 33.20% % * PTGTIN 45.66% % * CRIMSON 4.22% 63.35% PARAMOUNT 19.25% 69.81% AHTIN 27.44% % * FIMACORP 16.54% 52.77% Note: * sgncant at 5% level 15