Are Appraisers Rational? Evidence from T-REITs

Transcription

1 Are Appraisers Raional? Evidence from -REIs Fong-Yao Chen (Corresponding auhor) Associae professor el: Fax: Dep of Land Economics, Naional Chengchi Universiy Jen-Hsu Liang Dep of Naural Resources, Chinese Culure Universiy Submi o 16h acific Rim Real Esae Sociey Conference Wellingon, New Zealand JANUARY 2010

2 Are Appraisers Raional? 1 Evidence from -REIs Absrac A herd menaliy is driving financial evens worldwide. he grea damage ha he curren financial crisis has brough o he world has revealed he excessive opimisism characerisic of financial bubbles in he boom phase of economic cycles. I also leads o he re-evaluaion of appraisal smoohing. In imes of prosperiy, do appraisals have o flucuae wih he marke? Mos of he previous sudies examined appraisal smoohing from an aggregae level and used exensive daa ses o de-smooh. his paper uses individual re-appraisal daa of -REIs and modifies he parial adjusmen model developed by Quan and Quigley (1991) o observe appraisers raional behavior in aiwan. he resuls show ha he confidence parameer is 0.85 and verifies ha parial adjusmens exised. We find ha appraisers in aiwan place less weigh on marke informaion because of marke noise. Key words: arial adjusmen, Appraisal behavior, REIs 1 ar of NSC H

3 1. Inroducion An appraiser's proper role is o offer raional value informaion in he marke. heoreically, appraisers should esimae he unbiased marke value of propery. However Fisher e al. (1999) found ha propery sales prices exceeded he appraised values in up markes, and he reverse in down markes. Yiu e al. (2006) argue ha he persisence of esimae errors will grealy affec invesors judgmens. ha means ha appraised values have insufficienly reaced o marke flucuaions (so-called appraisal smoohing). he recen financial sunami shows excessive opimism accompanying a "prosperiy bubble". Akerlof and Shiller (2009) indicae ha a herd menaliy seems o be driving invesors' behavior. So much for invesors, bu wha abou professional appraisers, whose behavior one would expec o be more raional? heir raionaliy should be differen. If appraisers, jus as Fisher e al. (1999) and Gelner (1998) menioned, end o underesimae in prosperiy and overesimae in recession, is appraisal smoohing sill a bias or a raional expecaion? revious appraisal smoohing sudies have mosly used an aggregae level index and have concluded ha here are wo major characerisics: lower volailiy and lag srucure. hey have always ied lower volailiy and lag feaures o appraiser lack of confidence or anchor on previous appraised value. his paper adops disaggregae daa o find ou wheher appraisal smoohing has exised, and wha behavior causes he insufficien reacion o marke flucuaion. his paper is organized as follows: in addiion o his preamble, he firs par reviews he appraisal smoohing and parial adjusmen lieraure. We modify he raional parial adjusmen model developed by Quan and Quigley (1991). Secion 3 briefly inroduces -REIs marke and daa descripion. An empirical model and is resuls are hen presened, and finally our conclusions.

4 2. Lieraure review 2.1 Appraisal smoohing Appraisal smoohing can be sudied from aggregae and individual levels. Gelner (1989a), based a sudy upon aggregae level or asse porfolio calculaion, defined appraisal smoohing o be he siuaion when he raio of he ransacion price index o he appraisal sandard deviaion is greaer han 1, or when he appraisal price index falls behind he ransacion price index such ha when he marke price has a differen rend he appraisal price index does no cach up immediaely. Fisher e al.(1999) found ha when he marke reverses he rend o a growing marke, he appraisal price index is lower han he marke price index; while he marke declines, he appraisal price index is higher han he marke price index. revious appraisal smoohing research sudies have all assumed he exisence of appraisal smoohing. his assumpion is criicized by Lai and Wang (1998). hey showed ha he use of appraisal based daa can resul in a higher variance han ha of rue reurns. hey suggesed sudying he characerisics of real esae as possible explanaions for he low variance observed beween appraisal and ransacion indexes. Gelner (1998) argued ha Lai and Wang (1998) did no disinguish beween disaggregae level random error and sysemaic error which carries over o he aggregae index. A broader perspecive for concepualizing he problem of appraisal smoohing and more producive direcions for fuure research is recommended. Using he de-smoohing model o sudy he ime varying characerisic of appraisal smoohing, he smoohing coefficien may be differen in various economic cycles. oo much pas informaion may cause appraisal smoohing. Brown and Maysiak (1998) relaxed he consan smoohing coefficien assumpion o calculae he ime varying smoohing coefficiens, and used he Sae Space Model (SSM) o sudy he raional adjusmen model. Clayon e al.(2001) is based on individual appraisal daa, using he

5 Quan and Quigley (1991) parial adjusmen model o sudy 202 reappraisal repors for 33 real esae cases. By seing he confidence level as he ransacion price daa available o he appraisers, hey found ha he confidence level varies over ime. In differen economic condiions, appraisers will have differen confidence levels and use differen appraisal adjusmens. herefore, a de-smoohing model should use differen coefficiens over ime. revious appraised resuls can affec appraisers valuaion on he same real esae in consecuive periods and have more lagging han he firs ime appraisals. Roaing appraisers may be a good way o avoid he lagging effec by he previous appraisal on he same appraiser. Gelner (1991) claimed ha mos of he aggregae level research sudies commercial real esae. ransacion prices of commercial real esae are hard o collec. Mos of he research can only use appraisal prices o sudy risk and reurn relaionships or porfolio analysis. Mos of he research a he aggregae level has focused upon appraisal daa adjusmens o consruc price indexes and develop a de-smoohing model under he assumpion of he exisence of appraisal smoohing. Brown e al. (1998) criicized his assumpion. Wihou more deailed analysis of he reason for appraisal smoohing we can make sure o de-smooh and can sae wheher appraisers are using he wrong mehods, do no have enough experience, or do no use all of he marke informaion. In order o undersand he characerisics of appraisal smoohing, some research sudies focus on he individual level o sudy he appraisal process and appraiser behavior in order o beer undersand he facors for appraisal smoohing. Under he assumpion of incomplee informaion, cosly search, and varying expecaion, Quan and Quigley (1991)inroduced a real esae pricing model. he buyer and seller of real esae have less experience han he appraiser. he appraiser should exrac useful informaion from he marke. When he marke noise is bigger, and i is harder o observe marke price, he appraiser should adjus he price more conservaively. Conrary o he percepion of previous research, appraisal smoohing by he appraiser is boh raional and consisen

6 wih an opimal updaing behavior. Mcalliser e al.(2003)used a qualiaive inerview survey o sudy appraiser behavior in commercial real esae reurn performance. he appraisal smoohing may be due o he marke environmen. revious research sudies have claimed ha appraisal smoohing is because of he appraisal process, raher han marke inefficiency. Fuure research should undersand ha marke informaion is hard o obain, he appraisal process or lack of appraisal abiliy are no he only reasons for appraisal smoohing. Reappraisal no only need o consider weighed average prices, bu also many oher facors. 2.2 Appraisal behavior Using eiher he previous ransacion price (Ibboson and Siegel, 1984:222) or he previous appraisal (Ross and Siegel, 1987; Gelner, 1989) may cause auo correlaion and appraisal smoohing. During he appraisal process, appraisals may also be consrained by pas appraisals. Hansz (2004) used a conrolled experimen o sudy he impac of pas ransacion prices on parial adjusmen behavior of exper appraisers and non-appraisers. I is found ha pas ransacion price knowledge induces parial adjusmen behavior on exper appraisers. I could be for ha reason ha he Uniform Sandards of rofessional Appraisal racice (USA) required formal documenaion of appraisal pracice, such ha he appraiser canno ignore eiher previous appraisal prices or ransacion prices. Diaz (1997) and Gelner e al. (2001) also found oher people s opinion may also have impac on appraiser s parial adjusmen behavior. Anchoring means ha people will use a quaniaive reference as a basis for appraisal and adjusmen. However, anchoring may cause parial adjusmen problems (Kahneman and versky, 1973). Diaz and Wolveron (1998) used a longiudinal experimen o sudy reappraisal and found parial adjusmen phenomenon of he appraiser due o pas appraisal prices. Hansz (1004) found pas ransacion prices also have an impac upon appraiser s parial adjusmen behavior. However, behavior research can only show

7 parial adjusmen behavior caused by pas informaion, bu canno explain appraisal smoohing due o he appraiser s lack of confidence. Only Clayon e al. (2001) has shown ha he appraiser s confidence can cause appraisal smoohing. he moive of parial adjusmen remains an area for furher sudy. When an appraiser uses sales comparison mehods, collecs comparison cases, or obains capializaion raes from he marke, he or she is using pas ransacion informaion. his is very likely o resul in an appraisal lag problem. Alhough de-smoohing is a echnical issue, appraisal smoohing is caused by appraiser behavior. revious aggregae level sudies can only have limied and indirec implicaions as o he causes appraisal smoohing. We need o focus on appraiser behavior o sudy he moive for appraisal smoohing. his paper uses individual reappraisal cases on real esae securiizaion o sudy appraiser behavior. 2.3 arial adjusmen model Quan and Quigley (1991) used a ransacion model which is he weighed average of reservaion prices and offer prices o develop an individual appraiser parial adjusmen model o explain he appraiser s reappraisal behavior in he real esae appraisal marke. Assume ha he real marke price follows a random walk process and canno be observed. Volailiy is exogenous. = 1 + η and η ~ N(0, 2 σ η ) (1) Following appraisal rules, an appraiser can use available informaion and experience in making a real esae appraisal. Available informaion se a ime -1 is: {, } I 1 1, 2 3,..., 1 (2)

8 he ransacion price and unobservable marke price have a long erm equilibrium relaionship: = + υ,where 2 υ ~ N(0, σ v ). (3) Following his procedure, we can derive an appraiser opimal reappraisal process. Based on informaion I 1 a ime -1 and addiional informaion, appraiser s appraisal resul is he expeced real esae price a ime [, I ] = E 1 I Informaion se includes informaion 1 a ime -1 and addiional marke informaion a ime : E [, I ] E[ I ] + K [ E ( I )] 1 = K [ E( I )] 1 1 is he updaing componen.. (4) he appraiser does no use all he informaion a ime. Appraiser, based on informaion a ime o adjus he real esae price E I and pas appraisal ( ) 1 1, only uses adjusmen weigh K o parially adjus he real esae price. he appraiser s expeced real esae price a ime is he weighed average price of pas appraisal prices and marke ransacion price informaion. K K = + ( 1 ) 1... (5) Quan and Quigley only developed a heoreical model. Clayon e al. (2001) defined weigh K as he appraiser s confidence parameer o he informaion. However here is no research on appraisal smoohing in aiwan. his paper sudies he appraisers who, due o lack of confidence in marke ransacion informaion, and in valuing he same real esae in consecuive periods anchor ono heir previous appraisal values and have

9 parial adjusmen resuls. Quan and Quigley (1991) believed his is a raional behavior when appraisers have marke informaion uncerainies. Firs, we examine if parial adjusmens exised in reappraisal values. A sronger parial adjusmen affec will have a more serious appraisal smoohing resul. hen we sudy he facors, such as lack of confidence in available informaion, ha affec parial adjusmens. When marke noise is sronger, he reappraisal will be more conservaive. We also include a proxy variable for marke informaion qualiy ino our model. If an appraiser has ambiguiy aversion, raional behavior will give less weigh o uncerain marke informaion. When following raional behavior, marke informaion will have lower weigh; previous appraisal will have higher weigh. 3. Daa and mehodology 3.1 Descripive saisics his paper emphasizes he use of disaggregaed daa o es appraisal smoohing. here are 8 REIs cases in aiwan. According o he aiwan Real Esae Securiizaion Ac, rus properies should be reappraised every hree monhs. If here are more han wo appraisal values from differen appraisal firms, he average real price is he appraisal value. here are 26 real esae reappraisal cases in hese 8 REIs. he firs one is FuBan number 1 which was issued in he end of 2005 and he reappraisal in 2006Q1. he las day of a season is he reappraisal dae. Our daa se is panel daa. Since he firs -REI, real esae is a growing marke and does no have many decreasing prices. here are 120 reappraisal samples and we obained 38. he descripion is in able 2. From he dispersion degree, he sandard deviaion of marke value, , is higher han he oher hree mehods. I indicaes ha marke informaion is relaively dispersed, implying a valuaion smoohing phenomenon.

10 able 1. REIs Marke (by 2009 Ocober) Names of REI Dae Issued rus ropery Locaion Scale(US$ millions) FuBon REI#1 Mar offices, 1Serviced Aparmens,1 reail aipei Cahay REI#1 Oc office, 1 Hoel, 1 reail aipei Shin Kong REI#1 Dec offices, 1 reail, 2 aparmens. aipei, ainan FuBon REI#2 Apr offices aipei San Ding REI Jun office, 1 reail, 1 warehouse aipei, aoyuan Kee ai REI Aug office, 1hoel and office aipei Cahay REI#2 Oc offices aipei Gallop REI#1 May offices, 1 warehouse aipei oal volume able 2 Saisic descripion of marke value and appraised value value from marke informaion Value from marke approach Value from Direc capializaion mehod Value from DCF Mean Sd. Dev Kurosis Skewness Min Max N Appraiser s reappraisal sraegy Following he specificaion in Equaion (5), we can es he null hypohesis by esimaing he simple linear regression model. M j, α 0 + β1j, 1 + β2j, + ε j, =... (6) j, Where dependen variable is he appraised value of propery j a ime. Independen M, 1 variable includes j j and,. j, 1 is he appraised value of propery j a ime -1, M j namely previous value., is a proxy marke value variable of propery j esimaed by equaion (7).

11 M = NOI i ( 1 + ΔREN i CA ) (7) Where NOIi is ne operaion income a ime i, ΔREN i is he ren difference beween ime i and ime, CA is he mean value of capializaion rae a ime. o reduce variance heerogeneiy problems caused by large scale, his sudy will ake he naural logarihm on each variables, and hereby avoid coefficien esimaes bias. he double log model is: Ln ( ε, M j, ) = α 0 + β1ln( j, 1) + β2ln( j, ) + j.. (8) 4. Empirical Resul 4.1 Reacion on marke informaion able 3 presens he resul of our es of he smoohing hypohesis in he -REIs re-appraisal daabase. I shows in boh coefficien esimaes ha here is no much difference beween he wo models equipped wih appropriae well-adjused R square, However boh Jarque-bera es and Breusch-agan chi-square es indicae ha we can rejec he null hypoheses, and from lower J-B value we find a linear model can fi he requiremen of normaliy and variance heerogeneiy. herefore we used he double log model for coninuing analysis. As able 3 shows, we can rejec he null hypohesis a a 1% significance level. here exiss in he daa parial adjusmen behavior. Appraisers give higher weigh, , for he marke informaion, and less weigh, , for previous appraised value. his is similar o he resul Clayon e al. (2001) did (see able 4).

12 able 3 Empirical resul of parial adjusmen model linear regression model log log model Coefficien Sd Err. value Coefficien Sd Err. value consan E revious value Marke informaion N Adjused R F-es Skewness Kurosis J-B es B- value Significan a he 1% level. Significan a he 5% level. Joinly Null H: β = 0, β 2 F(2,35)= = 1 able 4 Hansz(2004) Comparison of appraiser s confidence on marke informaion Diaz & Hansz(1997) Hansz(2004) Clayon, Gelner, and Hamilon(2001) Diaz & Wolveron(1998) mehod Conrol experimen Conrol experimen Conrol experimen Reference poin as ransacion value(higher one) Confidence level, K Familiar wih he marke 0.48 No familiar Oher appraiser s esimae 0.54 No familiar as ransacion value(lower one) 0.66 No familiar Empirical as appraisal value 0.69 Familiar Conrol experimen as appraisal value 0.70 No familiar his paper Empirical as appraisal value 0.86 Familiar Clayon, Gelner, and Hamilon(2001) Diaz(1997) Source:Hansz(2004) Empirical daa Conrol experimen as appraisal value 0.87 Familiar as appraisal value 0.88 Familiar

13 4.2 Reacion on informaion qualiy In his secion we will invesigae adjusmen influence facors. Firsly, we es he raional behavior of appraisers reacing o low qualiy informaion. We replace he noise proxy variable wih he difference rae of marke-exraced values. As he value informaion exraced from he marke has greaer variaion, he appraiser will ake insufficien comparaives and know less abou he marke, or need o place far more adjusmen magniude on propery characerisics. Secondly, he ype of reference poin may have a differen impac on he appraiser's level of conservaism; appraisers could have more confidence in heir own appraised value raher han in ohers' valuaions. Finally we invesigae wheher he clien background will affec he adjusmen paern, a hypohesis ha he size of cliens will affec he adjusmen parameers will be esed. From equaion (4), we rewrie Quan-Quigley model o be equaion (9). ha is, appraisers will parial adjus o he marke change, he difference of conemporaneous marke informaion and las appraised value. M [ ] = + K j, j, 1 j, j, (9) j, j, hen K j, = M j, 1 j, 1 he parameer K is wha we concern he weigh of appraiser pu on marke informaion. o avoid K parameer o be zero and no o se aside he unchanged value, we define he dependen variable o be level of conservaism or named anchoring degree (AD), 1-K. he adjusmen influence facors model is specified as follow: n AD = α noise + β D + ε l l l = 1...(10) Noise is defined as he absolue value of he raio of he difference beween comparison

14 noise = value and capializaion value o he comparison value, comps comps cap. Higher difference beween comparison value and he capializaion value means more noise in he marke. A dummy variable se is o es wheher reference poin and clien size affecs he adjusmen. he variable descripion is in able 5. able 6 shows ha he regression model is significan a 1% level. -REIs appraisers do reac conservaively o low marke informaion as noise increases. he resul is he same wih Clayon e al. (2001). he dummy se of reference poin ypes shows appraisers refer o ransacion prices bu no oher appraiser s opinion. Appraisers have less anchoring effec o ransacion prices, which means ha appraisers have more confidence in heir own judgmen. Moreover, he model resul shows he larger he clien is, he more conservaive he adjusmen sraegy. able 5. Variable descripion Variable Mean Sd. Dev. measuremen Descripion AD coninuous How conservaive appraisers are when reappraised rus propery noise coninuous roxy variable of marke comparison qualiy D discree Caegories of reference poin (oher appraisers opinion=1, oher=0) D discree Caegories of reference poin (propery ransacion price =1, oher=0) D discree Relaive size of cliens (financial holding co. as originaor =1, ohers=0) Noes as dependen variable.

15 able 6. Resuls of AD = α noise + β D i i + ε i= 1 n Variable. Coefficien Sd. Err. -value noise D D D3.539 R-squared = Adjused R-squared = F(4,90)= rob.= Significan a 1% level. Significan a 5% level Conclusion Regression resuls show ha we rejec he joinly null hypohesis of full adjusmen o marke flucuaions and he confidence parameer is We find ha appraisers have parial adjusmen sraegies. Moreover, we find appraisers give less weigh o curren marke informaion because of marke noise. Marke noise does decrease appraisers confidence. ha means appraiser s parial adjusmen is a raional behavior in -REI s reappraisal. he resul is similar o Quan and Quigley (1991). Reference 1. Akerlof, G. A. and R. J. Shiller, 2009, Animal Spiris How Human sychology Drives he Economy, and Why I Maers for Global Capialism, rinceon Universiy ress. 2. Brown, G. R., and G. A. Maysiak, 1998, Valuaion Smoohing wihou emporal Aggregaion, Journal of ropery Research, 15(2): Clayon, J., D. Gelner, and S. W. Hamilon, 2001, Smoohing in Commercial ropery Valuaion: Evidence from Individual Appraisals, Real Esae Economics, 29(3) Diaz, J., 1997, An Invesigaion ino he Impac of revious Exper Value Esimae on Appraisal Judgmen, Journal of Real Esae Research, 13(1):57-66.

16 5. Diaz, J. and J. A. Hansz, 1997, How Valuers Use he Value Opinions of Ohers, Journal of ropery Valuaion and Invesmen, 15(3): Diaz, J. and M. Wolveron, 1998, A Longiudinal Examinaion of he Appraisal Smoohing Hypohesis, Real Esae Economics, 26(2): Fisher, J., M. Miles, and R. Webb, 1999, How Reliable are Commercial ropery Appraisals? Anoher Look, Real Esae Finance, 16(3): Gelner, D., 1989a, Bias in Appraisal-Based Reurns, Journal of he American Real Esae and Urban Economics Associaion (AREUEA Journal), 17(3): Gelner, D., 1989b, Esimaing Real Esae s Sysemaic Risk from Aggregae Level Appraisal-Based Reurns, Journal of he American Real Esae and Urban Economics Associaion, 17(4): Gelner, D., 1991, Smoohing in Appraisal-Based Reurns, Journal of Real Esae Finance and Economics, 4: Gelner, D., 1998, Appraisal Smoohing: he Oher Side of he Sory A Commen, Available a SSRN: hp://ssrn.com/absrac=131459, Working aper, Deparmen of Finance, Universiy of Cincinnai. 12. Hansz, J. A., 2004, rior ransacion rice Induced Smoohing: esing and Calibraing he Quan-Quigley Model a he Disaggregae Level, Journal of ropery Research, 21(4): Ibboson, R., and L. B. Siegel, 1984, Real Esae Reurn: a Comparison wih Oher Invesmen, AREUEA, 12: Kahneman, D. and A. versky, 1973, On he sychology of redicion, sychological Review, 80(4): Lai,.Y., and K. Wang, 1998, Appraisal Smoohing: he Oher Side of Sory, Real Esae Economics, 26(3): Mcalliser,., A. Baum, N. Crosby,. Gallimore, and A. Gray, 2003, Appraiser behaviour and appraisal smoohing: some qualiaive and quaniaive evidence, Journal of ropery Research, 20(3): Quan, D. C., and J. M. Quigley, 1991, rice Formaion and he Appraisal Funcion in Real Esae Markes, Journal of Real Esae Finance and Economics, 4: Ross, S., and R. Zisler, 1987, Managing Real Esae orfolio, ar 3: A Close look a Equiy Real Esae Risk, Goldman Sachs Real Esae Research, November. 19. Yiu, C. Y., B. S. ang, Y. H. Chiang, and L. H. Choy, 2006, Alernaive heories of Appraisal Bias, Journal of Real Esae Lieraure, 14(3):