A Closer Revew and Strategc Implcatons of the Comparatve Market Analyss n Settng the Lst Prce Chu V. Nguyen Unversty of Houston-Downtown Luclle L. Ponter Unversty of Houston-Downtown Charles Stran Unversty of Houston-Downtown Investgatng the belef that real estate propertes sellng wthn the tme of the lstng agent s contract duraton are correctly prced based on comparatve market analyss revealed that per square foot lsted and sold prces, not ther percentage devaton from the average of recent past sold prces, contrbute more to the Tme on Market (TOM) of lsted property. Comparatve Market Analyss (CMA) prcng s thus not as crtcal a factor n the predcton of Tme on Market (TOM). Instead, other contrbutng factors contrbute sgnfcantly to the TOMs of lsted propertes. As such, strategc planners may need to rethnk ther sales strateges. INTRODUCTION Hstorcally, the US has pursued a standng polcy promotng home ownershp through regulatons and nsttutonal arrangements. The ntroducton of Regulaton Q and the creaton of the Federal Home Loan Bank System, Gnne Mae and Fredde Mac are examples of how the U.S. government has encouraged and facltated the channelng of funds from economc unts wth surplus funds to the home mortgage markets. The net effect was that t made mortgage funds avalable to consumers at affordable rates. Consequently, almost sxty seven (67) percent of Amercan famles owned ther own homes before the subprme mortgage crss, and resdental real estate s by far the largest nvestment for the average Amercan as well as the largest component of ndvdual wealth. In 2007, realtors sold over 6 mllon new and used homes. Ths created a large, poltcally strong group of real estate agents and brokers n the U.S. The lterature s replete wth studes nvestgatng the dverse strateges used to sell real estate (Turnbull & Zahrovc-Herbert, 2012; Benefeld & Srmans, 2009; Johnson, Benefeld & Wley, 2009; Angln, Rutherford & Sprnger, 2003). Real estate agents nfluence many decsons n the marketng of property ncludng settng the lstng prce for property. The lst prce s the prce that a partcular property when t s put up for sale. Normally a realtor does a comparatve market analyss (CMA) to help the clent set the lst prce. A common belef s that the lstng prce and the attendant psychologcal reactons of the potental buyers to the lsted prce sgnfcantly nfluence both the fnal sales prce and the total number of days that the property remans on the market (TOM). Although prce s an mportant factor, evdence 48 Journal of Marketng Development and Compettveness vol. 6(5) 2012
ndcates other factors may nfluence the TOM. Other varables such school dstrct, the number of bedrooms and baths, mortgage rate and ts recent changes, number of lstngs n the partcular area, and so on, certanly can play a role n the TOM of the propertes. Gven the prevous dscusson, a logcal queston to ask s to what extent lst prce nfluences a property s TOM and to what extent other varables nfluence Tme on Market (TOM). Secondly, how useful s the CMA as an nstrument for settng lst prces. Because of the subjectve judgment n performng a CMA, the monetary beneft, and reputaton of the lstng agents n the ndustry for the quck sale of lsted propertes suggest the possblty that propertes are underprced. In lght of the aforementoned, ths study utlzes Cox s (1972) proportonal hazard and Tobt models to dscern the above ssues and evaluate the robustness of the emprcal results obtaned from them. Although researchers used the Cox, proportonal hazard Model n numerous studes, the Tobt has been used less n studes of ths nature. These models belong to the general class of the logstc regresson model, used to handle dscrete and truncated dependent varables. Revew of Lterature Gven the economc mpact of the housng market sales n the Amercan fnancal system, researchers have conducted studes to understand the factors predctng the length of tme property remans on the market (TOM) for close to forty years. Wthn ths tme, ndustry research found a myrad of factors mght be predctors of TOM. The vast quantty of research focused on three dstnct streams begnnng wth the physcal characterstcs of the property, lqudty ssues and lastly characterstcs of the sellers or a combnaton of these. Even though the lst prce has been ncluded n many models, there s less focus on how lst prces are determned n the housng market (Haurn, et al., (2010). The property lst prce s of crtcal concern for both sellers and prospectve buyers. Buyers vew the lst prce as a prmary factor n the length of tme property remans on the market because t captures the effects of the many hypotheszed drvers of sales (Knght 2002). Accordng to several researchers, the lstng prce provdes an upper boundary for expected sales offers and sgnals market nformaton to potental buyers (Haurn, et al., (2010); Angln 1997 and Horowtz 1992). Examnng the effects of sellng prce, TOM, and fnancng premums, Ferrera & Srmans (1989) found that the ntal lst prce postvely affect TOM and the greater the dfference between the lst prce and sales prce the longer the days on the market. Yavas & Yang (1995) research support the commonly prevalng noton that lst prce affects how long t takes for a real estate sell but they also show a recprocal effect because TOM nfluences the fnal sellng prce. Most studes consstently fnd a postve relatonshp between hgher than average lst prce and longer TOM (Jud, Seaks, & Wnkler, 1996; Asabere, Huffman, & Mehdan, 1993; Mller & Sklarz, 1987, Kang, & Gardner, 1989; Trpp, 1977; Angln, Rutherford & Sprnger, 2003; Haurn, et al., 2010), shows that sellers ncreasng senstvty to the arrval rate of potental buyers n a market, results n sellers reducng lst prces to mantan a steady flow of potental buyers. Knght (2002) and Huang & Palmqust (2001) examned the nterrelatonshp between lst prce and tme on the market fndng that t s crucal to lst homes at the rght prce ntally. Accordngly, the level of the ntal lst prce wll nfluences the rate at whch a seller learns about the buyers and the dstrbuton of offers. Hgher lstng prces, reduces the number of potental buyers and as a result fewer offers. Research shows that when we revse the orgnal lstng prces, the effect not only ncreases TOM but also results n lower sellng prces due to the exstence of a hypotheszed stgma effect, whch Taylor (1999) orgnally proposed. Benjamn & Chnloy (2000) show that property moves faster when prced at or below market value oppose to overprcng property. Ths effect must be qualfed dependng on geographc or spatal factors. In a study of the resdental housng major n a major U.K. cty where most resdental homes sell at a premum above the lsted prce, McGreal, et al., (2009), found that although the lst prce nfluences TOM, t s a more complex relatonshp. Propertes sellng hgher than lst prce experenced a shorter TOM and those sellng for less than the lst prce had a longer TOM. However, ths relatonshp only exsts for propertes on the market for less than 6 months. Journal of Marketng Development and Compettveness vol. 6(5) 2012 49
TOM and Sellng Prces In the US resdental real estate market, the actual sellng prce more often than not dffers from the lst prce (Horowtz, 1992; (Haurn, et al., (2010). Many studes document an nverse relatonshp between sellng prce and TOM compared to the postve relatonshp between lst prce and TOM (Belkn, Hempel & McLeavey, 1976). Examnng the relatonshp between sellng prce and marketng tme, Benefeld, Rutherford, and Allen (2011), found that across all homes ncludng both normal, foreclosed and homes classfed as estate sales, TOM had an nverse relatonshp to sellng prces. However, estate homes, whch were typcally older and smaller wth a sgnfcant lower lst prce and sellng, prce and spent slghtly less TOM than the non-estate homes. Whle nvestgatng the effects of lst prce, Haurn, et al., (2010) confrmed that homes atypcally rases the rato of lst prce to sales prce at a decreasng rate and dwellngs wth greater atypcally have a longer tme on the market. The study confrms that sellers of atypcal propertes tend to set prces relatvely hgh, and offer hgher dscounts from lst prces, whch supports Knghts (2002) fndngs. Research shows that seller heterogenety or constrants mpacts lst prce, sellng prce and TOM. As sellers holdng costs ncreases, reservaton prces decreases and TOM decreases (Cheng, Ln, & Lu, 2010; Srmans, Turnbull, & Dombrow, 1995; Glower, Haurn & Hendershott, 1998; and Arnold, 1992). Kang and Gardner (1989) note that the relatonshps between sellng prce, lst prce, housng characterstcs are complex ones that are often dependent on market condtons. Lst Prce, Comparatve Market Analyss and Agents The prevalng belef n the real estate ndustry s that correctly prced propertes sell wthn the tme of the lstng agent s contract duraton. Overprced propertes tend not to sell quckly, thus remanng on the market too long and eventually leadng to the expraton of the lstng agents contracts. Ths convcton s the bass for realtors to perform a comparatve market analyss. A CMA s supposed to help both the agents and consumers decde upon the correct prce for the lstng property. To perform a CMA, real estate agents search the Multple Lstng Servces (MLS) archve to fnd at least three comparable propertes, preferably n the same geographc area, that were sold n the past sx months. If agents cannot fnd comparable propertes n the same geographc area or wthn the last sx months, the agent must rely on propertes n another geographc area, whch should be n relatvely close proxmty. These ssues force the agent to make many adjustments to fnd the comparable propertes for a CMA. There are no hard rules for these adjustments. Instead, ther determnaton requres a subjectve evaluaton. Addtonally, durng the perod when the real estate property prces move rapdly and monotoncally such as n the recent resdental real estate market, sx months s a long tme. Therefore, to provde a good CMA, the agent must be very knowledgeable of the current economc and fnancal condtons of the economy as well as the condtons of the current market n the prevalng phase of the busness cycle. The real estate ndustry conssts of agents wth varyng levels of sklls and knowledge due to several factors. To become an agent, an ndvdual must meet the state educatonal requrements and provde proof of competency by passng the state examnaton. Subsequently, actve agents must satsfy an annual state educatonal requrement known as the mandatory contnued educaton requrement (MCE.). The current MCE requrements consst of a mnmum number of quanttatve calculatons n the areas of economcs and fnance. These mnmum requrements result n a hgh varablty n the sklls and abltes of the lcensed agents. Reputable real estate companes provde excellent tranng to ther agents. These companes n return keep a large porton of the sales commssons from ther assocates. Due to ths bas aganst the commsson sharng structure, newer real estate frms developed known as one hundred percent companes. In these new frms, the prncpal brokers are avalable to answer questons from ther sale assocates, but provde lttle tranng. In ths arrangement, the real estate sale assocates pay a mnmum annual fee and about one hundred dollars per transacton. Thus, agents n these establshments learn by dong, whch n turn exacerbates the problems assocated wth the qualty of agents and ther ablty to develop good CMAs. The prncple agent problems are a factor affectng the agents nputs on clents decsons on settng the lst prce and acceptance of offers. Levtt & Syverson (2008) show that realtors sell ther own homes at a hgher prce than they sell ther clents and leave them on the market 50 Journal of Marketng Development and Compettveness vol. 6(5) 2012
longer, whch led them to conclude that realtors often work to ther own optmal advantage rather than to ther clents best nterest. Turnbull & Dombrow (2007) demonstrate that agent specalzaton nfluences property sellng prce and tme on the market. Agents specalzng n lstng propertes obtan hgher prces for ther sellers whle those who specalze n sellng obtan lower prces for ther buyer. Ths hghlghts the mportance of agents beng experenced and knowledgeable n the market. A knowledgeable agent should help set the rght prce through ther CMA. The purpose of ths study s to nvestgate the relatonshp between lst prce, CMA and TOM. Do propertes lsted close to CMA sell wthn the tme of the lstng agent s contract perod? Potental Contrbutng Factors to TOM of Lsted Resdental Real Estate For the most part, real estate property owners tend to have an upward bas when prcng ther propertes. However, gven the above dscusson on the ndustry, the CMA s a useful nstrument for lstng agents to recommend the lstng prces for propertes. Lstng prces based on CMA may not account for all the varables nfluencng ther TOM. Based on pror research, t s logcal to postulate that the lsted prce of a property s an mportant factor n ts TOM; t s not the only factor contrbutng to ts TOM. Other factors specfc to that property may nfluence the TOM of a real estate property. Natonal economc and monetary polces, whch manfest n the personal ncome, nterest rate, unemployment, and nflaton affect the wllngness and the ablty of the populaton to purchase resdental propertes, are also contrbutng factors to the TOMs of lsted resdental propertes. To nvestgate the predctors of TOM, ths study uses commonly nvestgated home characterstcs wth the excepton of loan to value whch were dentfed from pror studes (see for example Haurn 1988; Kang & Gardner 1989; Glower, Haurn & Hendershott, 1998; Turnbull & Dombrow, 2007; Culp & Retzlaff, 2008; Bourassa, et al., (2009); McGreal, et al., (2009); Benefeld, Rutherford & Allen, (2011): per square foot lsted prces sold prces under or over the CMA (whch s the average of sale prces n the school dstrct where the lsted property s located) loan to value rato change n state unemployment rates from fve to four months ago change n the conventonal mortgage rates from four to three months ago ndependent school dstrct where the property s located perod when the property s lsted measured by dummy varable assumng dfferent values for dfferent perods when data s collected Aprls and Augusts of 2006, 2007, and 2008 number of full bathrooms number of bedrooms the cross between the number of bedrooms and full bathrooms to measure the functonal obsolescence of the property number of car garages square footage number of stores of the buldngs square footage of the lots whether the property has a prvate swmmng pool per square foot of lsted and sold prces amount the seller helps wth closng costs amount of money the seller spent on repars dollar lsted prce of the property year property bult bonus offered by the seller to buyer s agent Journal of Marketng Development and Compettveness vol. 6(5) 2012 51
Methodology The methodologcal challenge n studyng TOMs of lsted real estate propertes s whether the measures days n the market tself s dscrete and truncated from below,.e., TOM s an nteger and cannot be negatve. Ths truncatng phenomenon renders the conventonal econometrc procedure mproprate to dentfy the characterstcs contrbutng to the TOMs of lsted real estate propertes. To use any regresson model wth truncated data, the probablty densty of the dependent varable must relocate from plus and mnus nfnty to the range between zero and postve nfnty. To overcome ths statstcal challenge and to check for the robustness of the emprcal results, ths study follows several other studes (Haurn, 1988; Glower, Haurn & Hendershott, 1998) usng the sem-parametrc Cox proportonal hazard model that dscerns the aforementoned ssues and to check the robustness of the results by comparng them to the results obtaned from estmatng a parametrc Tobt model. COX PROPORTIONAL HAZARD MODEL Let the TOM be a random varable T wth a probablty dstrbuton functon f (, (where t s a realzaton of T) represent the TOM. As such, the followng equaton provdes the cumulatve probablty: F( = t f ( s) ds = Pr ob.( T (1) 0 The probablty that the TOM s of a length of at least n the termnology of ths topc matter, the property s sad to survve for the length of t can be descrbed by equaton (2) S( = 1 F( = Pr ob.( T. (2) Gven the lsted property has lasted untl tme t, the probablty that t sells n the next short nterval of tme, denoted by t, expresses as: (equaton (3) l( t, = Pr ob( t T t + t T. (3) A useful functon for characterzng ths aspect of the dstrbuton s the hazard rate whch s gven by equaton (4), Greene (2008, p. 933.) Pr ob( t T t + t T F( t + F( f ( λ ( = lm t 0 = lm t 0 = (4) t ts( S( As ponted out by Greene (2008, p. 934), the hazard rate s roughly the rate at whch the lsted resdental real estate propertes are sold after the duraton of t days, gven that they are on the market for t days. As such, the hazard functon dscerns the followng ntutve queston that the longer a lsted property s on the market, the more lkely that t sells wthn the next week. The speculatve poston of the real estate agents s that the longer a property has been on the market, the more dffcult t wll be to sell ths property because of the potental buyers psychologcal thnkng; thus, t s less lkely that t sells n the next short tme nterval. Greene, (2008) and Elandt-Johnson & Johnson, (1980) further artculated that the hazard functon, the probablty densty functon, the cumulatve densty functon, and the survval functon are all related. Equaton 5 provdes the hazard functon and equaton 6 descrbes the probablty densty functon. λ( = d ln S( dt (5) 52 Journal of Marketng Development and Compettveness vol. 6(5) 2012
f ( = S( λ(. (6) Another mportant functon for ths model s the ntegrated hazard functon provded by the followng expresson (7). t Λ( = λ ( s) ds (7) 0 Equaton (8) descrbes the survval functon. Λ( ) e t S( =. (8) Therefore, equaton (9) provdes an alternatve expresson of the ntegrated hazard functon. In ths settng, the ntegrated hazard functon s a generalzed resdual. Λ ( = ln S(. (9) As ponted out by Greene (2008, p. 937), one lmtaton of the above class of model s that external factors are not ncorporated as potental contrbutors to the survval dstrbuton and addng a set of covarates or explanatory varables to the models s farly straghtforward. To ths end, let x = ( z1, z2, z3,... zn ) be an n by 1 vector of covarates, then the followng equaton (10) specfes Cox s (1972) proportonal sem-parametrc hazard method of analyzng the effects of covarates on the hazard rate: λ = exp.( x β ) λ ( (10) ( 0 Thus, an alternatve to equaton (10) s: λ ( Ln = x β λ0 ( (11) In ths equaton, β s the vector of unknown regresson coeffcents we plan to estmate and λ0( s the unknown hazard functon for a lsted real estate property wth covarate vector x = ( 01,0 2,03,...0 n ). Also, as ponted out by Hopkns (1981, p. 576), ths specfcaton models the log-lnear effect of the covarates upon the hazard functon. Addtonally, when agents lst resdental real estate propertes, some of them wll be under contract quckly and contracts on some of the remanng propertes wll follow. Propertes that do not sale durng ths perod tend to reman on the market much longer and eventually lead to the expraton of the lstng agents contracts. In addton, TOM can never be negatve. The behavor of the TOM volates the normalty assumpton n conventonal OLS procedure. As ponted out by Greene (2008, p. 906), when modelng ths type of event, although an underlyng regresson model s n fact at work, t s not the condtonal mean functon that s of nterest. The objects of estmaton are certan probabltes of events. Researchers often use the exponental model for phenomena such as these. Econometrcally, Greene (2008, p. 941) artculates that the proportonal hazard model (semparametrc model) s a common choce for modelng these events because t s a reasonable compromse between the non-parametrc Kaplan-Meer estmators (Kaplan & Meer,1958), and possbly excessvely Journal of Marketng Development and Compettveness vol. 6(5) 2012 53
structured parametrc models. Cleves, Gould, Guterrez, & Marchenko, (2008) and Meyer (1988), have devsed another, sem-parametrc, specfcaton for hazard models. Hopkns, (1981) provdes a method by whch we may estmate the regresson parameters. Let ( t 1, t2, t3,... tk ) represent k dstnct tmes for a property to sell among the m observed tmes when the lsted propertes sell. The condtonal probablty that a lsted property wth covarate vector x s sold, gven that only one property s sold at tme t and that the set of pror to tme t ) s the rato of the hazards: j R x j R (ndces of propertes on the market exp.( x β ) (12) exp.( β ) If there s no multple sales (tes) among the lsted resdental real estate propertes at the tmes, t s then as Cox (1975) ponts out, multplyng these probabltes together for each of the sales tme, t, yelds the partal lkelhood functon (13):, L( β x) = k 1 = j R exp.( x β ) exp.( x j β ) (13) When there are tes or multple sales among the lsted resdental real estate propertes at the tmes, then Breslow (1974) proposes the followng lkelhood functon (14). t s, L( β x) = k = 1 j R exp.( s β ) m exp.( x j β ) (14) The equaton provdes that m s the number of sales of the lsted real estate resdental propertes at tme t, s s the sum of the covarates of the m of the sold propertes. Maxmzaton of the approprate partal lkelhood functon yelds estmators of β s wth the propertes smlar to those of the usual maxmum lkelhood estmators such as an asymptotc normalty, Hopkns (1981, p. 578.) As artculated by Kefer, (1988), the λ 0( t ), the baselne hazard wth unknown parameters, wll normally requre estmaton. In the above specfcaton, we obtan the effects of the covarates by multplyng the hazard λ 0( by a factor exp.( x β ), whch does not depend on the duraton of t. Addtonally, ths specfcaton s convenent because non-negatvty of exp.( x β ) does not mpose restrctons on β and the estmatons and nferences are straghtforward. More mportantly, estmaton of β n the above model does not requre the specfcaton and estmaton of the baselne hazard, λ ( ), Greene (2008, p. 940) and Kefer, (1988). TOBIT MODEL 0 t To check for the robustness of the emprcal results, we estmate the Tobt model (a parametrc member of the class of the logstc regresson model developed by economst James Tobt 1958) usng the same data set. Agan, the objectve of the analyss s to construct a probablty model that lnks the changes n a set of ndependent varables or covarates to the probablty of an outcome. Followng 54 Journal of Marketng Development and Compettveness vol. 6(5) 2012
Greene (2008), ths study specfes equaton (15) as the bass condton to construct the Tobt model, where y* s unobservable a dependent varable relatng to a set of covarates x as follows: y * = x β + ε y = 0 f y * 0 * y = y f * 0 < y (15) Econometrcally, equaton (15) specfes how a vector of factors, x nfluences the TOMs of lsted real estate propertes. Green (2008, p. 928), shows the development of the log-lkelhood of ths model from two sets of terms as follows: ln L = y 0 0 x β ln Φ + σ 0 < y 1 y ln Φ σ x β σ (16) Ths model can be used to estmate the coeffcent vector β of the covarates or ndependent varables x. DATA AND EMPIRICAL RESULTS Ths secton emprcally examnes whether or not other varables besdes the lsted prce (based on the CMA) provded by lstng real estate agents affect the TOM of lsted resdental real estate property. To ths end, ths study collected data for the aforementoned varables on 5,544 resdental real estate propertes sold n three ndependent school dstrcts, recorded n the archve of the MLS of Houston Assocaton of Realtors n Aprls and Augusts of 2006, 2007, 2008. Out of these sample propertes, 2,927 were lsted at the per square foot prce above that of the average of the recent past sold prces n the ndependent school dstrct, whch s used as a proxy measure for a CMA. To apprecate the beauty of the Cox proportonal hazard model, t s mportant to realze that equatons (1) and (2) combned or equaton (9) ndcates that f any gven covarate postvely affects the hazard rate of the propertes,.e. the reported hazard rato beng greater than one, that covarate n fact shortens ths property s TOM. For example, f the estmated coeffcent of an ndependent varable or a covarate s 0.18, then an ncrease n the measurement of that varable by one unt wll cause the hazard rate by 20% snce exp. (0.18) s 1.20. Alternatvely, f an estmated coeffcent of a covarate s -0.2231, then an ncrease n that ndependent varable by one unt wll result n a decrease of the hazard rate by 20% because exp.(-0.2231) = 0.8. Clearly, a postve estmated coeffcent of Tobt ndcates an ncrease n the TOM of the property; whle a postve estmated coeffcent of Cox s model ndcates an ncrease n the hazard that the property n queston to be censored or to be sold.,.e., shortenng the TOM of the property n queston. Snce some propertes were lsted under and some were lsted over CMA, ths study separated the sample n two subsamples one subsample ncludes propertes lsted under and the other conssts of those lsted above the CMA and each of the two selected model s estmated wth the full sample and these two subsamples as reported n Table 1. Addtonally, some sellers offer bonuses to buyng agents wth the dollar value of the bonus to the buyng agent expressed as the percentage of the sold prce for the emprcal analyss. Overall, the emprcal results reveal the goodness of ft as evdenced by the log lkelhoods and the Ch-square statstcs.e. fts the models for all the samples -- full sample and two subsamples well. In determnng factors nfluencng TOMs for the full data sample, the Tobt model reveals that the lstng prce s margnal sgnfcant, whle the Cox s model ndcates that t s not sgnfcant at any conventonal level. As to the bonus to agents representng buyers, the Tobt s model suggests that t s not sgnfcant, whle Cox s model ndcates that t s sgnfcant at the one percent level. Overall, the comparson Journal of Marketng Development and Compettveness vol. 6(5) 2012 55
between the two models of the logstc regresson model class suggests that Cox proportonal hazard model s more powerful n detectng factors that nfluence the TOMs of lsted real estate propertes. The consstency of the comparson results lends credence to the emprcal fndngs of ths nvestgaton. Statstcally, we determne the sgnfcance of ndvdual estmated coeffcents of Cox s model by the z-statstcs, whle the t-statstcs determne the sgnfcance of the estmated coeffcents of the Tobt model. From the full sample results - except for the aforementoned mnor dfference between results obtaned from the two models - an analyss of the estmaton results ndcate that many varables are sgnfcant at the 1 percent level. These nclude: the ndependent school dstrct where the property s located sze of the garage as measured by the number of cars number of stores of the unt, per square foot lsted prce and sold prce seller s help n closng cost, year bult of the unt bonus the seller offers to the buyer s agen are sgnfcant at the 1 percent level The emprcal results further ndcate that whether the property has a prvate swmmng pool s sgnfcant at the 5 percent level. Whereas a change n the conventonal mortgage rates from four to three months pror to the contract to purchase by the buyers, seller s pad repars, dollar lsted prce are margnally sgnfcant and all other ncluded factors are statstcally nsgnfcant at conventonal levels. Interestngly, the emprcal fndngs reveal that whle the levels of the per square foot lstng prce and sold prce but not ther devatons from CMAs lsted under or over the average of the recent past sold prces n the ndependent school dstrct sgnfcantly nfluence the TOMs of the lsted real estate property. Although usng the recent past sold prces n the ndependent school dstrct as a proxy measure for CMA n ths analyss may mtgate the belevablty of the emprcal fndngs, the emprcal results cast doubt on the prevalng belef n the real estate ndustry that propertes that sell wthn the tme of the lstng agent s contract duraton are prced rght. Propertes that do not sale fast and reman on the market too long and eventually lead to the expraton of the lstng agents contracts are not due to them not beng prced rght. The results seem to lend some support to the publc percepton reported n the Real Estate Industry survey 2005 and the aforementoned defcences of CMAs prepared by lstng real estate agents n the ndustry. 56 Journal of Marketng Development and Compettveness vol. 6(5) 2012
TABLE 1 Propertes Lsted Under CMA Propertes Lsted Over CMA Full Property Sample Varables β s - Cox β s -Tobt β s - Cox β s -Tobt β s - Cox β s -Tobt Lst Prce vs. CMA 0.00383 0.97-0.36798-1.27-0.00002-0.13 0.00060 0.07-0.00003-0.20 0.00124 0.14 Sell Prce vs. CMA -0.00198-0.52 0.17066 0.62-0.00119-0.79 0.07388 0.70 0.00094 1.09-0.09081-1.46 Loan Value Rato 0.00682 0.79-0.36878-0.53-0.00637-0.60 0.61825 0.95-0.00107-0.16 0.15270 0.32 Δ Unemployment rate from 5 to 4 mos. pror to contract 0.00153 0.19-0.26385-0.45 0.01044 1.38-0.67956-1.28 0.00741 1.37-0.52537-1.33 Δ mortgage rate. from 4 to 3 months. pror to contract 0.00760 0.72-0.88131-1.11 0.01227 1.22-0.96923-1.36 0.01233 1.70*** -0.94964-1.79*** School Dstrct Locaton 0.10292 2.94* -6.94591-2.70* 0.06435 1.78*** -4.66901-1.88*** 0.105212 4.30* -7.47747-4.27* Perod Property Sold 0.01134 0.95-0.52814-0.59-0.00133-0.12-0.18900-0.23 0.00478 0.58-0.27845-0.46 # Full Bath Rooms -0.38445-2.11** 21.94409 1.65*** 0.03759 0.23 0.278270 0.03-0.11451-1.00 7.81326 0.97 # of Bedrooms -0.14047-1.29 5.86655 0.74 0.19081 1.91*** -11.29692-1.67*** 0.05073 0.72-4.00335-0.81 Bath/Bedroom Functonal Obsolescence 0.07432 1.46-3.19452-0.86-0.03756-0.74 1.69076 0.49 0.00582 0.17-0.07598-0.03 Sze of Garage/ # of spaces 0.12913 3.95* -9.92914-4.04* 0.09965 3.53* -6.33306-3.15* 0.10966 5.26* -8.24734-5.33* Buldng Sq. Feet -0.00028-3.19* 0.02297 3.59* 0.00028 2.80* -0.01822-2.69* -0.00006-1.21 0.00584 1.51 # of Stores -0.02866-0.60 2.02170 0.56 --0.26619-5.41* 20.76292 6.13* -0.15088-4.60* 11..15824 4.63* Lot Sze 0.00001 0.71-0.00021-0.66-0.00001-0.67 0.00013 0.86-0.00005-0.29 0.00005 0.33 Prvate Pool 0.06149 0.83-5.92108-1.06 0.15491 2.08** -11.14090-2.14** 0.12929 2.40** -9.25521-2.43** Lstng Prce per Sq. Foot -0.05414-6.66* 4.01576 6.80* -0.01745-4.70* 1.07413 4.37* -0.02633-8.40* 1.77037 8.19* Sold Prce per Sq. Foot 0.04954 7,04* -3.54884-6.95* 0.02371 6.50* -1.43101-5.92* 0.02729 8.94* -1.76273-8.34* Seller Pad Buyer s Closng Cost -.00002-2.91* 0.00109 2.97* -.00002-1.97** 0.00113 1.91*** -0.00003-4.32* 0.00109 3.55* Seller Pad Repars -0.00003-4.08* -0.00013-0.88-0.00002-0.71 0.00058 0.44 0.00001 1.67*** -0.00011-0.76 Year Bult -0.00337-2.53** 0.22705 2.22** -0.0040-3.82* 0.29400 3.84* -0.00340-4.24* 0.24383 4.04* Lstng Prce n $000 s 0.00001 2.79* -0.00027-3.31* -0.0001-2.62* 0.00013 2.27** 0,00001 1.17-0.0000-1.74*** Bonus pad to Buyers Agent 0.00315 2.85* -0.12017-1.45 0.08347 1.39-4.89841-1.17 0.00303 2.80* -0.11840-1.48 Chqurare(22) 131.40* 124.83* 173.15* 167.56* 253.77* 236.55* Log lkelhood -17,934.97-15,002.26-20,374.19-16,66.73-42,165.20-31,644.57 Journal of Marketng Development and Compettveness vol. 6(5) 2012 57
Future research may refne the CMA by ncludng the comparable propertes very near to the lsted property and adjust for ther dfferent characterstcs to rectfy the CMA rudment. Arguably, the rudments of the CMA used n ths nvestgaton may mtgate the belevablty of the emprcal fndngs. The mportant contrbutons of ths study are that t confrms the use of the Cox s model as better model than the Tobt model for busness applcatons. It also confrms that there are several other mportant factors other than prce nfluencng the TOMs of lsted real estate propertes. A comparson of emprcal results obtaned from estmatng the two models reveals that some aforementoned factors becomes more/less or sgnfcant/not sgnfcant determnant of TOMs of some lsted real estate propertes n one sample versus another. More specfcally, a change n the conventonal mortgage rate from four to three months pror to buyers contracts to buy contrbutes sgnfcantly to TOMs for the full sample, but becomes nsgnfcant when the models were estmated usng two subsamples. In addton, number of bedrooms margnally contrbutes to TOMs of propertes wth per square foot lstng prce below the CMA level, but does not affect TOMs of other propertes. Addtonally, number of stores of the unt s hghly mportant n determnng TOMs of propertes when the per square foot lstng prce s above the CMA and s rrelevant regardng the TOMs of propertes when the per square foot lsted prce s below CMA level. We conjecture that these dfferences are attrbutable to the synergc effects of the characterstcs of the propertes n these samples and the buyers of these propertes. For example, a buyer of an upper qualty property, hence hgher per square foot lstng prce would have taste for dfferent amenty of the property than the taste of a buyer of a property wth lower than the CMA prce per square foot. CONCLUSION To nvestgate the belef that propertes that sell wthn the tme of the lstng agent s contract duraton have the correct prce and other factors, whch affect a property s Tme on Market (TOM), ths study used both the Cox s (1972) proportonal sem-parametrc hazard model and Tobt model. Interestngly, the emprcal fndngs reveal that the levels of the per square foot lsted prce and sold prce, but not ther devatons from the proxy measure of the CMA, sgnfcantly nfluence the TOM of lsted real estate propertes. Therefore, CMA prcng s not as crtcal a factor n the predcton of TOM. The model comparsons reveal that both models ft the data well, however, Cox s model s more powerful n determnng factors contrbutng to days on the market of lsted real estate propertes. For example, the estmated coeffcent of the bonus pad to buyers agents was hghly sgnfcant n Cox s model, but was not sgnfcant at any conventonal level n Tobt model. The estmaton results ndcate that the followng are sgnfcant at the one-percent level: the ndependent school dstrct where the property s located sze of the garage as measured by the number of cars number of stores of the unt, per square foot lstng prce and sold prce seller s help wth closng costs year bult of the unt bonus the seller offers to the buyer s agent The emprcal results further ndcates that whether the property has a prvate swmmng pool s sgnfcant at 5 percent level, whle change n conventonal mortgage rates from four to three months pror to the contract to purchase by the buyers, seller s pad repars, dollars lstng prce are margnally sgnfcant, all other ncluded factors are statstcally nsgnfcant at conventonal levels. Thus, real estate strategc planners may need to rethnk the varables they use n developng ther sales strateges. 58 Journal of Marketng Development and Compettveness vol. 6(5) 2012
REFERENCES Angln, P. M., (1997). Determnants of Buyer Search n a Housng Market, Real Estate Economcs, 1997, 25, 567-89. Angln, P.M., Rutherford, R. C. & Sprnger, T. M. (2003). The Trade-off Between the Sellng Prce of Resdental Propertes and Tme-on-the-Market: The Impact of Prce Settng, Journal of Real Estate Fnance and Economcs, 26, 1, 95 111. Angln, P. M. (2003). The Value and Lqudty Effects of a Change n Market Condtons, Mmeo, Department of Economcs, Unversty of Wndsor. Arnold, M. A. (1992). The Prncpal-Agent Relatonshp n Real Estate Brokerage Servces, Journal of the Amercan Real Estate and Urban Economcs Assocaton 20: 89 106. Asabere, P., Huffman, F. & Mehdan, S. (1993). Msprcng and Optmal Tme on the Market, Journal of Real Estate Research 8 (1), 149 156. Benefeld, J. D. & Srmans, G. S. (2009). The Influence of Contngent Closng Costs on Sale Prce, Tme on Market, and Proftablty of Sale, Journal of Housng Research, 18, 2, 121-142. Benefeld, J. D., Rutherford, R. C. & Allen, M. T. (2011). The Effects of Estate Sales of Resdental Real Estate on Prce and Marketng Tme, Journal of Real Estate Fnance Economcs (onlne). Benjamn, J. D. & Chnloy, P. T. (2000). Prcng, Exposure and Resdental Lstng Strateges, Journal of Real Estate Research, 20:1/2, 61-74. Belkn, J., Hempel, D. J. & McLeavey, D. W. (1976). An Emprcal Study of Tme on Market Usng Multdmensonal Segmentaton of Housng Markets, AREUFA Journal 4: 2 (Fall), 57-75. Bourassa, S. C., Haurn, D. R., Haurn, J. L., Hoesl, M. & Sun, J. (2009). Housng Prce Changes and Idosyncratc Rsk: Impact of Property Characterstcs, Real Estate Economcs 37; 2; 259-270. Breslow, N. E. (1974). Covarance Analyss of Censored Survval Data. Bometrca, 30, pp. 89-99. Cheng, P., Ln, Z. & Lu, Y. (2010). Home Prce, Tme-on-Market and Seller Heterogenety Under Changng Market Condtons, Journal of Real Estate Fnance Economcs, V. 41, 3, 272-293. Cleves, M., Gould, W., Guterrez, R. & Marchenko, Y. (2008). An Introducton to Survval Analyss Usng Stata, 2 nd Edton, Stata Press. Cox, D. R., (1972). Regresson Models and Lfe Tables, Journal of the Royal Statstcal Socety, Seres B, 34, pp. 187-220. Cox, D. R., (1975). Partal Lkelhoods Bometrca, 62, pp. 269-276. Culp, R. P. & Retzlaff, T. M. (2008). Duraton of Marketng Tme of Resdental Housng wth Dssmlar Buyer Search Effort, Southern Busness and Economc Journal 31; 3&4, p. 19-42. Elandt-Johnson, R. C. & Johnson, N, L, (1980). Survval Models and Data Analyss, John Wley and Sons, New York. Journal of Marketng Development and Compettveness vol. 6(5) 2012 59
Ferrera, E. J. & Srmans, G. S. (1989). Sellng Prce, Fnancng Premums, and Days on the Market, Journal of Real Estate Fnance and Economcs, 2: 209-222. Glower, M., Haurn, D. & Hendershott, P. (1998). Sellng Tme and Sellng Prce: The Influence of Seller Motvaton, Real Estate Economcs, 2: 209-222. Greene, W. H, (2008). Econometrc Analyss. 6 th edton, Pearson-Prentce-Hall. Haurn, D. R. (1988). The Duraton of Marketng Tme of Resdental Housng, Journal of the Amercan Real Estate and Urban Economcs Assocaton 16; 4 396-410. Haurn, D. R., Haurn, J. L., Nadauld, T. & Sanders, A. (2010). Lst Prces, Sales Prces, and Marketng Tme: An Applcaton to U.S. Housng Markets, Real Estate Economcs, 38, 4, p. 659-685. Hopkns, A. (1981). Regresson wth Incomplete Survval Data. BMDP Statstcal Software, 1981 ed., Unversty of Calforna Press, 1981, pp. 576-594. Horowtz, J. L. (1992). The Role of the Lst Prce n Housng Markets: Theory and an Econometrc Model, Journal of Appled Econometrcs 7(2): 115-129. Huang J. & Palmqust, R. (2001). Envronmental Condtons, Reservaton prces, and Tme on the Market for Housng, Journal of Real Estate Fnance and Economcs, 22, 203-219. Johnson, K. H., Benefeld, J. D. & Wley, J. A. (2009). Archtectural Revew Boards and Ther Impact on Property Prce and Tme-on-Market, Journal of Housng Research, 18, 1, 1 18. Jud, G. D., Seaks, T. G. & Wnkler, D. D.T. (1996). Tme on the Market: The Impact of resdental Brokerage, Journal of Real Estate Research, 12, 447 458. Kang, H. B. & Gardner. M. J. (1989). Sellng Prce and Marketng Tme n the Resdental Real Estate Market, Journal of Real Estate Research, Vol. 4, (1) 21-36. Kaplan, E. & Meer, P. (1958). Nonparametrc Estmaton from Incomplete Observatons. Journal of Amercan Statstcal Assocaton, 53, pp. 457-481. Kefer, N. M. (1988). Economcs Duraton Data and Hazard Functons. Journal of Economc Lterature, 26, June, pp. 646-679. Knght, J. R, (2002). Lstng Prce, Tme on Market and Ultmate Sellng Prces: Causes and Effects of Lstng Prce Changes, Real Estate Economcs, V30, 2, p 213-237. Levtt, S. D. & Syverson, C. (2008). Market Dstortons When Agents are Better Informed: The Value of Informaton n Real Estate Transactons, The Revew of Economcs and Statstcs, Vol. XC (Nov) (4), 599-611. McGreal, S., Adar, A., Brown, L. & James R. Webb (2009). Prcng and Tme on the Market for Resdental Propertes n a Major U.K. Cty, JRER, Vo. 31, 2, 209-233. Meyer, B. (1988). Sem-Parametrc Estmaton of Hazard Models. Northwestern Unversty, Department of Economcs.. 60 Journal of Marketng Development and Compettveness vol. 6(5) 2012
Mller, N. G. & Sklarz, M.A. (1987). Prcng Strateges and Resdental Property Sellng Prces, Journal of Real Estate Research, 2 (1): 31-40. Srmans, C. F., Turnbull, G. K. & Dombrow, J. (1995). Quck House sales: Seller Mstake or Luck?, Journal of Housng Economcs, 4, 230-243. Taylor, C. (1999). Tme on the Market as a Sgn of Qualty, Revew of Economc Studes, 66 (3), 555-578. Tobt, J. (1958). Estmaton of Relatonshp for Lmted Dependent Varables. Econometrca, 26, pp. 24-36. Trpp, R. R. (1977). Estmatng the Relatonshp between Prce and Tme of Sales of Investment Property, Management Scence, 23 (Aprl), 838-42. Turnbull, G. K. & Dombrow, J. (2007). Indvdual Agents, Frms and the Real Estate Brokerage Process Journal of Real Estate Fnance and Economcs, 35, pg. 57-76. Turnbull, G. K. & Zahrovc-Herbert, V. (2012). The Transtory and Legacy Effects of the Rental Externalty on House Prce and Lqudty, Journal of Real Estate Fnance and Economcs, 33, 3, 275 297. Yavas, A. & Yang, S. X. (1995). The Strategc Role of Lstng Prce n Marketng Real Estate: Theory and Evdence, Real Estate Economcs, 23, 3, 347-60. Journal of Marketng Development and Compettveness vol. 6(5) 2012 61