Conflicts of Interest in Residential Real Estate Transactions: New Evidence

Transcription

1 Conflict of Interet in Reidential Real Etate Tranaction: New Evidence Marcu T. Allen Alico Eminent Scholar Department of Economic & Finance utgert College of Buine Florida Gulf Coat Univerity Fort Myer, F timallen@fgcu.edu Jeica Rutherford Department of Finance Muma College of Buine Univerity of South Florida Tampa, F , jmrutherford@mail.uf.edu Ronald Rutherford Bank of America Endowed rofeorhip Finance Department Muma College of Buine Univerity of South Florida Tampa, F , rrutherford@uf.edu Abdullah Yava* Robert E Wangard Real Etate Chair School of Buine, Univerity of Wiconin Madion 5251 Grainger Hall 975 Univerity Avenue Madion, WI 53706, ayava@bu.wic.edu h: * Correponding Author Draft: December 1, 2015

2 Conflict of Interet in Reidential Real Etate Tranaction: New Evidence Abtract Rutherford and Yava 2005 and evitt and Syveron 2008 document premium of approximately 3.5% for real etate agent that ell their own houe. The general argument i that agent exploit their information advantage when elling their propertie. We examine the other ide of the tranaction comparing the price paid by real etate agent buying houe to purchae by individual. We find that in the Miami-Dade County MS market real etate agent purchae houe at price approximately 4.0 % below the price paid by individual. We how that agent-buyer gain from their informational advantage even when the eller i repreented by another informed agent. In addition, conitent with prior literature, we find that owner agent ell houe they own at a price premium of approximately 3.5%. Our reult ugget that liting agent failure to protect eller bet interet become greater when the buyer of eller property i another agent. We alo find that agent to agent tranaction occur at market value, offetting any competitive advantage an agent might have in the market. Thee reult ugget that the degree of the conflict of interet between the principal and agent in real etate tranaction i greater than previou etimate. We etimate total agency cot to be 7.7%, capturing agency cot on both buying and the elling ide of a tranaction. Keyword: Agent Buyer, Individual Buyer, Invetor Houing rice Buyer ower

3 Conflict of Interet in Reidential Real Etate Tranaction: New Evidence 1. Introduction Agent are often conidered expert in their field and may ue their informational advantage when tranacting with other individual who are not expert. rofeional agent are expected to be relatively well informed regarding the value of aet and provide their client with expertie when elling or purchaing aet. In general, agent have a reponibility to repreent the interet of their principal to the bet of their ability, that i, in the ame way they would repreent themelve. In Florida, agent have a reponibility to dicloe known fact that materially affect the value of a reidential real etate property and are not readily obervable to the buyer. In addition, there i a duty to deal honetly and fairly. Under the dicloure claue, it i aumed that an agent ha an obligation to dicloe their vat knowledge and licening to circumvent taking advantage of buyer or eller who are not licened. The National Aociation of Realtor Code of Ethic, Article 4 alo indicate that a Realtor dicloe their Realtor tatu when buying or elling for their own account. In a world of perfect information, an agent would pay the ame price and ue the ame effort for hi client property a he would in elling or buying hi own property. That i, equivalent aet will be tranacted at the ame price and remain in the market for the ame length of time regardle of whether or not they are old and/or purchaed by a real etate agent. Earlier tudie on potential agency problem in reidential brokerage indutry have largely focued on the mialignment of incentive between the eller and the agent. Thi agency problem between the eller and hi agent ha been empirically documented in two recent tudie. Rutherford et al and evitt and Syveron 2008 provide empirical evidence that agent obtain higher price when elling their own aet, uggeting that agent may exploit their informational advantage, or expend more effort on elling their own propertie, depite their fiduciary reponibility to repreent the interet of their principal in the ame way they would repreent themelve. 3

4 However, there i a bigger, and omewhat aburd, mialignment of incentive between the buyer and the agent he work with. The reaon i that, a in the cae of the eller agent, the buyer agent typically receive a percentage of the price a commiion. Thu, the commiion ytem fail to align not only the magnitude, but alo the direction of the interet of the buyer and her agent. In thi paper, we offer the firt direct empirical evidence of the agency problem between the buyer and her agent. We utilize a comprehenive data et that allow for the identification of propertie purchaed by agent. Thu, our unique date et allow u to invetigate the purchae price difference, and to directly tet whether real etate agent expand the ame effort level and ue their informational advantage for the benefit of their buyer client the ame way they would for themelve. The empirical reult in thi paper how that real etate agent purchae houe at price approximately 4.0% below the price paid by individual. Thi reult i critical for a number of reaon. Firt, it how that incentive problem exit on the buying ide a they do on the elling ide. Agent pay le for imilar propertie than their buyer client. Second, it how not only that liting agent don t obtain imilar price for their client propertie a they do for their own propertie, liting agent failure to protect the bet interet of the eller become aggravated when the buyer of hi client property i another agent. That i, not only doe the liting agent not treat the eller liting a hi/her own, the liting agent alo fail to protect eller interet againt another agent; agent purchaing from non-agent eller that are neverthele repreented by a liting agent are able to purchae at lower price of approximately 4.0%. In other word, agent-buyer may gain from their informational advantage even when the eller i repreented by another informed agent. Thee reult etablih that the agency problem in reidential real etate are broader than uggeted by Rutherford et al and evitt and Syveron In addition, we alo find that the price premium the agent enjoy when elling their own propertie to non-agent i imilar to the reult in Rutherford et al and evitt and Syveron Rutherford et al report that agent-owned houe ell at a price premium of approximately 4.5% while evitt and Syveron 2008 report a premium of 3.7%. 4

5 We how that, when we control for agent purchae, the price difference remain at approximately 3.5%. In a mall ubet of the ample where the property i owned and old by a real etate agent and a different real etate agent purchae, the coefficient for price premium i tatitically inignificant. Additional analyi ugget no difference in price when a property i old by an agent and purchaed by an agent. Thu, agent to agent ale tranact at market price, wherea elling agent to non-agent tranaction obtain a price premium of 3.5% and non-agent to purchaing agent tranaction ell for a 4% dicount. %. We alo find that agent to agent tranaction occur at market value, offetting any competitive advantage an agent might have in the market. We etimate total agency cot to be 7.7%, capturing agency cot on both buying and the elling ide of a tranaction. The agency problem in market for real aet ha been the focu of variou recent tudie. The general concluion of the earlier model i that although the percentage commiion ytem enure the interet of the agent to be in the ame direction a thoe of the client, it fail to align the magnitude of the interet of the agent with thoe of the client. For example, Geltner et al. 1991, Anglin and Arnott 1991 and Miceli 1991 point out the agency problem with repect to the agent earch effort level and argue that the agent will expend le than the efficient level of effort for the eller while Arnold 1992 tudie the pricing apect of the agency problem and how that the agent reervation price for the eller houe will generally be different than the eller reervation price. The ource of the agency problem in thee model can be ummarized a follow. The eller want to maximize the elling price while minimizing the time the aet tay on the market. The agent, on the other hand, eek to maximize hi expected commiion revenue while minimizing the time on the market. Given that the agent receive a mall portion of the tranaction price a commiion, the agent goal of maximizing the expected commiion may diverge from the eller goal of maximizing the elling price. Thi divergence of interet give trong incentive to broker to get the deal done quickly, and try to convince the eller to accept a uboptimal price. In fact, Bernheim and 5

6 Meer 2008 find that a eller ue of a broker reult in a decreae in the elling price by 5.9 to 7.7 percent. 1 A recent tudy by Han and Hong 2015 identify another conequence of aymmetric information in reidential brokerage market. They focu on in-houe tranaction where both the eller liting agent and buyer agent work for the ame brokerage company. Since in-houe tranaction help brokerage firm clear inventorie fater and ecure commiion from both end of a tranaction, brokerage firm often pay a higher portion of the commiion to reward agent engaged in in-houe tranaction. The author find evidence that an agent' promotion of internal liting impoe a ubtantial cot on buyer. The buyer agent utilize their informational advantage to direct buyer interet to internal liting, even when thi reult in a uboptimal match for buyer. A indicated earlier, agency problem between the eller and hi agent have been empirically documented in Rutherford et al and evitt and Syveron Thee tudie how that agent obtain higher price when elling their own aet. The current paper offer an additional empirical tet of the agency problem in real etate brokerage. We focu on the agency problem on the buying ide of real etate tranaction. Our unique data et allow u to invetigate whether real etate agent are able to ue their information to their advantage when purchaing a property even when a eller i repreented by an informed real etate agent. The current paper contribute to a wider and growing literature on the role of aymmetric information in real etate market. The role of aymmetric information in houing and mortgage market ha gained a ignificant amount of attention ince the recent financial crii. Garmaie and Mokowitz 2004, for intance, how that participant in real etate market reolve aymmetric 1 Earlier tudie on the impact of employing an agent on the tranaction price offer mixed reult. While Bernheim and Meer 2008 find negative impact, Frew and Jud 1987 find a poitive impact, and Hendel, Nevo, and Ortalo-Magné 2009 find no effect on the tranaction price. Elder et al 2000 focu on the impact of buyer ue of a broker and find that buyer broker have no effect on home price. 6

7 information problem by purchaing nearby propertie and propertie with long hitorie while avoiding tranacting with informed broker. On the other hand, Bayer, Geiler and Robert 2011, Chinco and Mayer 2012, Kurlat and Stroebel 2015 and Stroebel 2015 examine how informed real etate trader or lender can exploit their uperior information in their tranaction. An, Deng and Gabriel 2011 offer evidence of advere election problem in commercial mortgage ecuritization while Elul 2009, Key et al. 2010, Agarwal, Chang and Yava 2012, Albertazzi, et. al., 2015 and Ambroe, Conklin and Yohida 2015 offer evidence of advere election in reidential mortgage ecuritization. A related line of reearch invetigate agency iue aociated with mortgage ecuritization Key et al. 2010; Mian and Sufi 2011; Mian, Sufi, and Trebbi Relative to thi literature, the current paper i the firt to document aymmetric information problem between buyer and buyer agent in reidential real etate market. It alo i the firt to combine the agency problem on buying and elling ide of the tranaction, and acertaining that earlier paper by Rutherford et al and evitt and Syveron 2008 underetimate the degree of ummative agency problem in reidential real etate market. The next ection of the paper review agency problem between the buyer and her broker and propoe empirical prediction. The third ection of the paper provide an overview of the data. In ection four we dicu the etimation of the model uing data from 34,655 Multiple iting Service MS liting from Miami-Dade County Florida during the year 2009 through September An important feature of thi data et i that we are able to identify 4, % of the obervation a real etate agent purchae. The empirical reult preented in ection five how that agent-purchaed 2 We indirectly contribute to a related body of literature that look that ha provided evidence on the behavior of aymmetrically informed agent in financial market e.g, in and Howe, 1990; Seyhun, 1992; Ealey, Hvidkjaer, and Ohara. 2002; Kelly and jungqvit, 2012, and Choi, Jin, and Yan, The current tudy alo contribute to the literature on middleman that relied on aymmetric information and earch friction to explain the role of intermediarie in variou market e.g., Biglaier, 1993; Yava, 1994, 1996; i, 1998; Shevchenko,

8 houe ell no fater than other houe, but they do ell at a price dicount of approximately 4.0% below the price of a comparable houe not purchaed by a real etate agent. The final ection of the paper ummarize the reult and offer ome concluding remark. 2. Real Etate Agent urchaing in the Single-Family Houing Market Real etate market are characterized by imperfect information. Buyer and eller do not know each other location or reervation price. Furthermore, ince real etate i a heterogeneou good and houehold have idioyncratic tate, the final tranaction price will depend on the characteritic of the buyer and eller at the negotiation table a well a the market condition. Thee uncertaintie, coupled with the fact that mot people get involved in a real etate tranaction on a very infrequent bai, create a market for intermediarie, namely real etate agent, who pecialize in facilitating real etate tranaction. The vat majority of real etate tranaction in the US go through an MS Multiple iting Service. An MS i owned and operated by an aociation of local of real etate broker, and i a vehicle for member broker to hare information about their liting. That i, under MS, the member broker combine their liting and create a ingle pool of liting, and by doing o they create a platform where almot all the eller and buyer are brought together. However, the primary purpoe of an MS i to provide a facility for broker to cooperate with each other in elling their liting. It i a mechanim through which liting broker offer compenation to buyer broker who bring a buyer for their lited property. The liting broker typically charge the eller 6% of the tranaction price and offer 40-60% of hi/her commiion to the broker repreenting the buyer. The commiion rate that i offered by the liting broker to buyer broker i publihed within the MS. In thi paper, we focu on the buy ide of real etate tranaction and invetigate the agency problem between the buyer and her agent. What i triking about the buy ide i that the buyer agent 8

9 receive a percentage of the price a commiion. Thi create an obviou conflict of interet. The buyer want to minimize the purchae price and the time the aet tay on the market. The buyer agent, on the other hand, eek to maximize hi expected commiion revenue while minimizing the time on the market. Given that the agent receive a mall portion of the tranaction price a commiion, the agent goal of maximizing the expected commiion will conflict with the buyer goal of minimizing the elling price. Since the agent effort level i not directly obervable and the agent ha aymmetric information about the market value of the aet, thi divergence can lead to hirking by the agent. A a reult, the agent may be motivated to convince the buyer to accept a uboptimal purchae price at the bargaining in order to facilitate a fater ale and obtain a higher commiion. What i ironic i that in mot tate the broker that the buyer work with while earching for a houe i typically a ub-agent of the liting broker, who i an agent of the eller. In other word, the buyer broker i expected to repreent the bet interet of the eller, not the buyer. Thi often lead to confuion for the buyer, if they undertood the agency relationhip, and occaionally led ome buyer to file legal complaint againt buyer broker for mirepreentation. A a reaction, many broker today offer excluive buyer agency contract to buyer, under which they are committed to repreent the bet interet of the buyer, intead of being a ubagent of the eller. Oddly, even under excluive buyer agency contract, the buyer agent commiion i till typically a percentage of the price and the buyer agent commiion i till being paid by the eller. In Florida, where the data for the current tudy come from, the apparent conflict of interet and the confuing relationhip between the buyer and buyer broker led to a change in tate tatute in July The new Florida Statute, tate: It hall be preumed that all licenee are operating a tranaction broker unle a ingle agent or no brokerage relationhip i etablihed, in writing, with a 9

10 cutomer. 3 A tranaction broker i not an agent of either party in the legal ene of the word, but merely help to facilitate a tranaction between the buyer and eller. Mot broker in Florida today prefer to only work a tranaction broker in order to avoid the liability iue aociated with the traditional ub-agency relationhip. Compared with liting broker and buyer broker who are both agent or ub-agent of eller, broker who erve a tranaction broker do not have the legal liability to repreent the bet interet of eller. Thu, they are not required to try their bet to obtain a higher price for the eller when they are helping a buyer to find a home. However, ince broker till receive a percentage of the price a commiion, they till have the financial incentive to convince the buyer to agree to pay a higher price than they would otherwie. Earlier paper by Rutherford et al and evitt and Syveron 2008 ue data from Texa and Illinoi, repectively, and document premium of approximately 3.5% for real etate agent that ell their own houe. Thi i very imilar to the premium we obtain for agent owned houe in the current tudy. The reult of thee three tudie indicate that having tranaction broker do not help buyer eliminate or reduce the price premium they pay when they purchae agent owned houe. That i, eliminating legal liability, without eliminating the financial incentive, for the buyer broker to repreent the bet interet of eller ha not been an effective policy in protecting buyer againt the informational diadvantage they face in dealing with eller who are agent. Empirical rediction In thi ection we propoe everal prediction that allow u to analyze empirically the importance and outcome of information aymmetrie between buyer and buyer agent. Thee prediction follow from the imple theoretical model we offer in the appendix

11 The firt prediction follow from the fact that agent will exhort greater effort when purchaing for their own account and make a more effective ue of their aymmetric information advantage in price negotiation. A a reult, they will obtain a lower price for themelve. rediction 1: Real etate agent will purchae houe at lower price than individual. If the agent are able to purchae at lower price, and if lower price are not due to unobervable inferior characteritic of the propertie they purchae, then they hould enjoy greater capital gain when they ell their propertie. rediction 2: Real etate agent hould obtain higher appreciation for their propertie. The data we have enable u to replicate the analyi of Rutherford et al and evitt and Syveron Rutherford et al tudy ue data from Texa while evitt and Syveron 2008 tudy i baed on data from Illinoi. Buyer broker in thee two tate typically erve a ubagent of the eller. 4 In Florida, on the other hand, tranaction brokerage i the default cae unle the broker and her client agree to an alternative contract where the broker will have fiduciary reponibility to repreent her client interet. Thu, while broker are a neutral party in tranaction in Florida, they have legal liability to repreent the bet interet of their client in Illinoi and Texa. Since the client principal i typically the eller, thi hould lead to a better alignment of interet of the eller and broker in Texa and Illinoi than in Florida. A a reult, when comparing the price premium that agent receive when elling their own propertie, we hould ee a bigger price premium 4 Texa allow for tranaction brokerage when the broker i repreenting both partie in a tranaction. In The Hitory of The Texa Real Etate icene Act, the Texa Real Etate Commiion, on September 2005, amendment were made, page 14, item 7: A broker who agree to repreent both a buyer and a eller mut agree to act a an intermediary. Texa Statutory TREA

12 in the current tudy than thoe reported in Rutherford et al and evitt and Syveron However, broker in all three tate receive a percentage of the price a commiion, paid by the eller. Thu, they have the ame financial incentive in the three tate. rediction 3: Financial incentive would prevail over fiduciary reponibilitie; price premium for the agent-owned propertie in Florida hould be imilar to that in Rutherford et al and evitt and Syveron Data In order to conduct the empirical analyi, we obtain data from a number of dataet. The primary data are MS ale for Miami-Dade county. The MS data lack grantor and grantee information that i needed to identify real etate agent purchae REA. However, Miami-Dade county ha maintained grantor and grantee information in their ale data ince approximately We match the Miami-Dade county data from 2009 through September 2013 with the MS dataet uing the property unique property ID Folio Number 5. The Miami-Dade ale dataet include grantee and grantor information, ale price, date of ale, a unique property ID Folio number, deed book and deed page, property addre, DORcode type of property, SaleCode type of ale, 6 quare feet of the building, quare feet of the land, number of bedroom, number of bathroom, number of torie, year built, and effective year built. A econd dataet from Miami-Dade contain information about propertie with pool that we ue to create a pool dummy variable. A third et of yearly dataet are obtained from the Florida Department of Revenue FDOR. Each year, every Florida County provide 5 The rationale for the time period i that grantor and grantee information i available from January 2009 and we extracted the data in September/October See for example Real roperty Tranfer Qualification Code for ue by DOR & roperty Appraier Beginning January 1, 2012 at: 12

13 a dataet that contain the aeed value of the land and aeed value of each property to FDOR. We ue the FDOR Miami-Dade dataet to etimate the percentage of value from the land and match with the ale dataet by year and by property ID. The dataet alo contain a quality decription each year that we match with the ale data to obtain an etimate of the quality of the property. 7 In addition, cenu block and cenu block group are available in the FDOR dataet and we ue cenu block group to control for location. We match the data from the above-decribed dataet with the local MS data by the tax ditrict property information number. The MS dataet include time on the market, ale price, lit price, cah ale and REO ale and liting and elling agent ID. 8 We alo create dummy variable for the type of tranfer baed on the FDOR tranfer qualification code. The initial coding i in SAS with viual verification in Excel. In addition we match the grantee information with a lit of Real Etate Agent obtained from the Florida Real Etate Commiion. Name are initially matched in SAS and verified in Excel. 9 The data et contain 34,655 MS ale from 2009 through September Tranaction with a price below $20,000, a price of $10,000,000 or greater or propertie with lot quare footage le than 645 quare feet and propertie with two or more living unit were alo deleted. ropertie with more than 1,095 day on the market were deleted along with tranaction with miing data. The final ample of 34,655 ale include 4,517 agent purchae and 2,696 owner agent ale. We preent variable ued in the analyi and their decription in Table 1. Table 2 provide decriptive tatitic for the full ample, real etate agent purchae and other buyer with a difference in mean t-tet. Sale price are lower for agent purchae $260,265 v. $325,031, but they are alo 7 We aume that the quality meaure i at bet a relative meaure of quality, not an abolute meaure of quality. 8 The cah percentage i approximately 41% for the MS ale. Thi number i conitent with the 43% etimate by Realty Trac, Augut 29, Agent name from the Florida Real Etate Commiion are matched with the grantee name provided by Miami-Dade county in their ale databae. In any matching by name there i the poibility of error. If a matched grantee i not a real etate agent, but ha the ame name a an agent and i coded a a real etate agent purchae, then the bia i toward 0 for that coefficient. Ideally, the MS or county would require that a buyer reveal he or he i a real etate agent and record thi in the MS or county record. 13

14 maller, with fewer bedroom, and bathroom. Agent purchaed propertie have fewer torie and a lower percentage with a pool which i conitent with lower valued propertie relative to other purchaed propertie. A higher percentage of agent purchaed propertie are vacant, are REO ale, need work, are of average quality and are purchaed with cah compared to other buyer purchaed propertie. One variable of interet that ha generated a number of article in the popular pre i the cah tranaction. 10 In Table 2, we note that real etate agent pay cah for about 41% of their purchae and other purchaer buy with cah in about 29% of their tranaction over the ample period, o the difference ugget that agent ue cah at a higher rate than other buyer ue cah in thi market. Controlling for cah tranaction i important becaue paying with cah give the buyer a competitive advantage when negotiating the price of a home becaue of two reaon. Firt, a cah buyer may preent le rik to the eller of the deal falling apart due to the mortgage-contingency claue in a ale contract. Second, a cah purchae may reduce the time required to complete the tranaction becaue cah buyer do not have to pend time to obtain a loan approval for the purchae. A a reult of thee two reaon, a eller would be willing to accept a lower price when he face a cah buyer. Indeed, Aabere, Huffman, and Mehdian 1992 and uht and Hanz 1994 report dicount for cah financing of 13 and 16 percent, repectively. In the firt et of regreion and the probit model, we compare agent purchae to non-agent purchae. Mot of the remaining regreion have control for REA with the control group being individual grantee purchaing a houe for their own account in the dataet. The only exception i when we compare propertie owned by an agent and old to an agent to propertie owned by an 10 For example a report by Goldman Sach, in the Mortgage Analyt, Augut 14, 2013 titled How much upide to purchae mortgage origination? etimate an increaing percentage of cah tranaction with approximately 30% cah tranaction in 2009 and roughly 58% in the ummer of RealtyTrac, Augut 18, 2014 tate: Among metropolitan tatitical area with a population of at leat 500,000, thoe with the top ix highet percentage of cah ale were all in Florida: Miami-Fort auderdale-ompano Beach 64.1 percent i the highet. 14

15 agent and old to a non-agent and when we compare propertie owned by an agent and old to an agent to propertie owned by a non-agent and old to an agent with thee reult preented in Table Method We initially etimate a model with cenu block group fixed effect and ale year month fixed effect. The initial empirical model we etimate allow u to compare agent purchaed propertie to propertie purchaed by other buyer, and take the following form: yi = β0 + β1 RE Agent urchae + β2 RE Agent Own + Σ βi Qi + Σ βi TSi + Σ βi Xi + Σ βi Ri + εi, 1 where the dependent variable y i the logged ale price Table 4, 7 and 8, or time on the market Table 5 and 6. RE Agent Own i a dummy for a property owned by a real etate agent and Q i a et of variable decribing the relative quality of a property in a given year and TS i a et of additional variable decribing the type of ale. The vector Xi for the ale price model include a full et of houing characteritic a indicated in Table 5. Thee include phyical characteritic uch a ize, effective age, bathroom and bedroom, and pool. Ri i a vector of dummy variable created from real etate agent public remark, and ε i a random error term. In addition, in Table 4 and 5 we replace the REA variable with two binary variable defined in Table 1 that refine the type of real etate agent purchaer into Smaller REA one or two purchae and arger REA, 3 or more purchae. Thee binary variable take the value of 1 if the tranaction involve grantee who fit the ize categorie defined above and 0 otherwie. The omitted category in Table 4 and Table 5 i the individual ingle-purchae buyer. The purpoe of thi categorization of REA variable i to identify and differentiate the impact of agent purchaer who tranact frequently and act a invetor. In addition the regreion model include a et of dummy variable for the ale 15

16 year and month for the ale price model along with dummy variable for each cenu block group. The time on the market model include lit year and month in place of ale year and month in the price model and include DO a a dependent variable. The primary reaon to etimate a time on the market model i a a robutne tet. The dummy variable for ale year and month, or lit year and month, and cenu block are not reported in the table. We alo etimate a robit model to examine the characteritic aociated with REA veru other grantee a follow: rob REA = πx, OwnerAgent, Q, TS, 4 In Equation 4, the dependent variable rob REA i the probability of a real etate agent purchaing the property, and X, Q, and TS are a defined above and in Table Empirical Reult The reult of the robit model, provided in Table 3, marginally indicate that overpriced propertie and propertie lited by experienced agent are le likely to be purchaed by REA. REO propertie and cah purchae are more likely to be purchaed by a real etate agent. The reult in the robit model provide no evidence that agent purchae more or le often from owner agent. That i, agent buyer do not particularly try to avoid, or attracted to, houe owned by other agent. It appear that REA are better equipped to compete for ditreed propertie, prefer propertie that are average or below average in quality, and more frequently purchae with cah. Reult obtained from the initial regreion model in Table 4 indicate that REA are purchaing propertie at approximately a 4.7% dicount relative to other buyer after controlling for phyical characteritic, a proxy for quality, and for type of ale along with REO and Cah. In the econd model we include dummy variable decribing or commenting on the quality and attractivene of certain apect of the property, created from the MS public remark. The coefficient on REA decreae to 4.0%. We then plit the REA variable into two group, REA grantee that purchaed 1 16

17 or two propertie and REA grantee that purchaed 3 or more propertie. The reult in model 3 indicate that agent purchaing 3 or more propertie are able to obtain a larger dicount than agent purchaing 1 or 2 propertie when we compare them to the group of other buyer. All remaining regreion have control for the MS public remark and for REA with the control group being individual grantee purchaing a houe for their own account in the dataet. In thi data we alo have owner agent elling their own propertie. The reult are imilar to thoe found by Rutherford and Yava 2005 and evitt and Syveron 2008, with a premium of approximately 3.5% for real etate agent that old their own property through the MS. The other variable are a expected, with cah purchae, lower quality and older propertie elling for lower price and larger houe elling for higher price. More bedroom or more bathroom are alo aociated with higher price. There i a mall ample 276 tranaction of the 4,517 REA that ell in the dataet. We include thi et of owner agent propertie eparately from the Owner Agent Sale and label it Subequent Sale after urchae by Agent. Thi variable ha a coefficient of approximately 3.9% in the ale price model, indicating that the ubequent ale by a purchaing agent now an owner agent ell at a imilar price a other owner agent ale in the data. The mall ample of REA reale may overtate the capital gain, but if we take the broader meaure of Owner Agent, that coefficient i a 3.5% premium, again indicating a ubtantial capital gain for propertie old by an agent relative to propertie purchaed by an agent. Thee reult are conitent with finding by Kurlat and Stroebel 2015 that real etate agent who are buyer of reidential real etate are able to earn abnormal capital gain compared to individual in the ame market It i poible that higher capital gain enjoyed by agent buyer are due to more pending by agent on repair and maintenance after the purchae. We addre thi concern by including a number of variable in our analyi that capture characteritic and condition of the property and whether the property wa remodeled. 17

18 The quality proxy variable have the expected ign compared to average quality, with fair quality elling at about a 22% dicount, above average quality propertie elling at a 3.6% premium and excellent quality propertie elling at a 10.9% premium. In Table 5, we tudy the day each property tayed on the market DOM. We conduct fixed effect regreion analyi and a Weibull duration model. Both model how that agent purchae are on the market the ame amount of time a a typical property in the ample. Another variable of interet in Table 5 i the degree of overpricing DO variable, which meaure the percentage by which the lit price exceed the expected lit price etimated from a regreion model. The coefficient in both model of day on the market indicate that while overpricing ha a ignificant and negative impact on the number of day the property tay on the market, the magnitude of the impact i very mall. In Table 6, we provide tatitic for new variable that we ue in place of Owner Agent and Agent urchaing. In Table 7, uing the full ample and the variable in equation one, we provide the etimate of the new variable keeping all other variable a in Table 4 and Table 5 model. We find no tatitical difference in the price of a property that i owned by an agent and purchaed by an agent. When an agent ell to an agent, we obtain the ame price a individual elling to other individual. When an agent own the property and ell to a non-agent, we ee a premium of 3.6% that i conitent with our prior etimate of 4.0% for REA. We identified a number of liting where the lit and ale agent are the ame. In Florida, we do not claify thi a dual agency, but imply a tranaction brokerage. We examined what happen when the liting and elling agent are the ame and an agent purchae thi property. We found that if the liting and elling agent are the ame, then an agent purchaing the property obtain an even better deal. In thi cae, the purchaing agent buy at a dicount of -5.4%, uggeting the liting and elling agent i helpful to hi fellow agent and either recommend to hi client to accept a lower price or may provide information to the buyer that allow him to purchae at a deeper dicount. In the cae where the lit and ale agent are different and agent purchae, the REA obtain a 3.5% dicount, cloer to the 4% we aw above. 18

19 6. Robutne Tet It i poible that agent-buyer obtain lower price becaue they target eller who are more deperate to ell. To invetigate thi argument, we examine day on the market DOM, the number of day each property tayed on the market before it wa old. We find in the difference in mean tet that DOM i lightly horter 155 day veru 160 day for the propertie purchaed by agent indicating that the propertie purchaed by agent have been on the market lightly horter, not longer. We alo examine whether the term Motivated eller /Reduced price were mentioned in the MS public remark and find no tatitical difference between agent-purchaed propertie and other ale. In the Sale rice fixed effect regreion model, the coefficient on DOM i , but tatitically ignificant, uggeting that longer time on the market are aociated with a very mall reduction in price. However, the coefficient i o mall that taking the average day on the market time the coefficient reult in le than 1 day, o eentially there i no impact. However, in our fixed effect model and a Weibull duration model of DOM reported in Table 5 we find that an agent purchaing i not aociated with horter or longer DOM. All three model have an Agent urchae coefficient cloe to zero and it i inignificant. Thu, we find no upport for the poibility that eller elling to agent might be different with more of an urgency to ell than eller elling to non-agent. In unreported reult from a 3SS etimation of price and time on the market, we find reult that are conitent with the fixed effect and Weibull model preented in Table 5. REA take the ame amount of time to market a propertie purchaed by individual One other poible explanation for the lower price obtained by agent buyer i that they imply identify and target mipriced / below-market priced propertie, a oppoed to uing their aymmetric information to increae their bargaining power and convince the eller to accept a lower price at the negotiation tage. To check for thi poibility, we compare two meaure of mipricing for agent buyer and non-agent buyer: iting rice-to-selling rice ratio and iting rice-to--etimated 19

20 Selling rice ratio. We find that thee two ratio are not different acro the propertie purchaed by agent and non-agent buyer. It i poible that we may not be controlling for all the property characteritic, though our reult on the owner agent variable i conitent with Rutherford and Yava 2005 and evitt and Syveron Though the ample i mall, 2,696 agent owned propertie with an average ale price of $406,722 or $159 per quare foot with 321 of thoe purchaed by an agent with an average price of $350,641 or $140 per quare foot, we are able to compare price of propertie owned by an agent & old to an agent with propertie owned by an agent & old to a non-agent. Since the owner agent didn t know who the future buyer would be at the time he purchaed the property i.e., they did not purchae propertie with a certain et of characteritic baed on their conjecture of whether or not the future buyer would be an agent, thi comparion yield the pure impact of whether the buyer i an agent or not. That i, a comparion of thee two group would erve a nice robutne tet againt the argument that buyer agent are paying le imply becaue they are buying different lower quality propertie, and eller agent are enjoying higher price imply becaue they are elling different higher quality propertie. We etimate a fixed effect model and preent the reult for only the variable of interet in Table 8a. All variable etimated in the regreion model in Table 4 are included in the analyi of Table 8a, except owner agent and agent purchae variable. Our variable of interet i a dummy variable repreenting propertie owned by an agent and old to an agent compared to the group of propertie owned by an agent and old to a non-agent. In thi mall ample, our reult indicate that the group of propertie owned by an agent and old to an agent ell for 7.7% le than propertie owned by an agent and old to a non-agent. A an extenion of the comparion above, we compare propertie owned by an agent and old to an agent with propertie owned by a non-agent and old to an agent. Thi include all propertie purchaed by an agent. In thi cae the reult are preent in Table 8b and the coefficient i and the t-tatitic i -0.06, again including all additional other variable. Thi indicate that the group of 20

21 propertie owned by an agent and old to an agent have imilar price to the group of propertie owned by an individual or non-agent and old to an agent. A a robutne tet, in Table 9, we expand the ample of Table 8 to include all owner agent propertie and all purchaing agent propertie and run a fixed effect ale price model. In thi model, we alo include a control for thoe 321 propertie old by an agent and purchaed by an agent. The reult indicate overall that a purchaing agent buy at a dicount of 7.66% relative to what an owner agent ell their propertie. For the 321 propertie old by an agent and purchaed by an agent, the propertie ell at market. Thu, one can interpret the 7.66% difference between the price paid by purchaing agent and the price obtained by agent elling their own propertie a the total potential agency cot in the market, capturing the agency cot between the agent and hi client on both buying and the elling ide of a tranaction. The reult of Table 9 are conitent with the reult preented in Table 8 where we break the ample into group owned by an agent and old to an agent relative to two other group, one owned by an agent and old to a non-agent and the other owned by a non-agent and old to an agent. 6. Concluion Our reult indicate that real etate agent purchae reidential real etate at dicount relative to individual ingle-purchae buyer during the year 2009 through We find that REA obtain ubtantial dicount of 4.0% and that Owner Agent propertie ell at ignificant premium of 3.5%, with a ubet of the REA elling a econd time a an Owner Agent property with a premium of 3.9%. In either cae we contend that real etate agent ue their informational advantage to purchae at dicount and ell at premium. Comparing the price of agent-purchaed propertie with agent-owned propertie, we find that agent purchae at a 7.7% dicount relative to elling price of agent-owned propertie. However, agent to agent tranaction occur at market value, offetting any competitive advantage an agent might have in the market. We believe 7.7% dicount erve a a proxy 21

22 for the total potential agency cot in the market, capturing agency cot on both buying and the elling ide of a tranaction. 22

23 Reference Agarwal, S., Chang, Y. and A. Yava Advere Selection in Mortgage Securitization. Journal of Financial Economic,105, Albertazzi, U., Eramo, G., Gambacorta,. and C. Salleo Aymmetric information in ecuritization: An empirical aement. Journal of Monetary Economic, 71, Ambroe, B. W., Conklin, J. and J. Yohida Reputation and exaggeration: Advere election and private information in the mortgage market. Working paper. An, X., Deng, Y. and S. A. Gabriel Aymmetric information, advere election, and the pricing of CMBS. Journal of Financial Economic, 1002, Anglin,. M Determinant of Buyer Search in a Houing Market. Real Etate Economic, 25, Anglin,. M., and R. Amott Reidential Real Etate Brokerage a a rincipal-agent roblem. Journal of Real Etate Finance and Economic, 4, Amold, M. A The rincipal-agent Relationhip in Real Etate Brokerage Service. Journal of the American Real Etate and Urban Economic Aociation now Real Etate Economic, 20, Aabere,., Huffman, F., and S. Mehdian The rice Effect of Cah veru Mortgage Tranaction. Journal of the American Real Etate and Urban Economic Aociation now Real Etate Economic, 201, Bayer,., C. Geiler, and J. Robert Speculator and middlemen: The role of flipper in the houing market. Working aper, 16784, NBER. Bernheim, B. D. and J. Meer How Much Value Do Real Etate Broker Add? A Cae Study. NBER working paper no Biglaier, G Middlemen a expert. The RAND journal of Economic, Chinco, A., and C. Mayer Ditant peculator and aet bubble in the houing market. Working aper, 19817, NBER. Choi, J.,. Jin, and H. Yan Informed trading and expected return, Working aper, 19420, NBER. Ealey, D., S. Hvidkjaer, and M. Ohara I information rik a determinant of aet return? Journal of Finance, 57: Elder, H. W.,. V. Zumpano, and E. A. Baryla Buyer Broker: Do They Make a Difference? Their Impact on Selling rice and Search Duration. Real Etate Economic, 28,

24 Elul, Ronel Securitization and Mortgage Default: Reputation v. Advere Selection. Federal Reerve Bank of hiladelphia Working aper Frew, J., and G. D. Jud Who ay the Real Etate Broker' Commiion? Reearch in aw and Economic 10, Geltner, D., B. D. Kluger, and N. G. Miller Optimal rice and Selling Effort from the erpective of the Broker and Seller. Journal of the American Real Etate and Urban Economic Aociation now Real Etate Economic, 19, Garmaie, M. and T. Mokowitz Confronting Information Aymmetrie: Evidence from Real Etate Market. Review of Financial Studie, 172, pp Han, and S. Hong Undertanding In-Houe Tranaction in the Real Etate Brokerage Indutry, U of Toronto Working paper. Hendel, I., Nevo, A. and F. Ortalo-Magné The Relative erformance of Real Etate Marketing latform: MS veru FSBOMadion.com. American Economic Review, 995, Kalai, E. and M. Smorodinky Other Solution to Nah' Bargaining roblem. Econometrica 43: Kelly, B., and A. jungqvit Teting aymmetric-information aet pricing model. Review of Financial Studie 25: Key, B. J., Mukherjee, T., Seru, A. and V. Vig Did Securitization ead to ax Screening? Evidence from Subprime oan. Quarterly Journal of Economic 125 1: Kurlat, and J Stroebel Teting for Information Aymmetrie in Real Etate Market. Review of Financial Studie, 28 8: evitt, S., & Syveron, C Market ditortion when agent are better informed: The value of information in real etate tranaction. The Review of Economic and Statitic 904, i, Y Middlemen and private information. Journal of Monetary Economic, 421, in, J. C., and J. Howe Inider trading in the OTC market. Journal of Finance 45: Mian, A, and A. Sufi Houe rice, Home Equity-Baed Borrowing, and the US Houehold everage Crii. American Economic Review 101 5: Mian, A, Sufi, A. and F. Trebbi The olitical Economy of the US Mortgage Default Crii. American Economic Review 100 5: Nah, J. C The Bargaining roblem. Econometrica. 18:

25 Rutherford, R. C., Springer, T. M. and A. Yava Conflict between principal and agent: evidence from reidential brokerage. Journal of financial Economic, 763, Rubintein, A erfect Equilibrium in a Bargaining Model. Econometrica 50: Sahin, M. A., Sirman, C. F. and A. Yava Buyer brokerage: Experimental evidence. Journal of Houing Economic, 224, Seyhun, N Why doe aggregate inider trading predict future tock return? Quarterly Journal of Economic, 107: Shevchenko, A Middlemen. International Economic Review, 451, Stroebel, J Aymmetric information about collateral value. Journal of Finance, forthcoming. Yava, A., and. Colwell Buyer brokerage: Incentive and efficiency implication. The Journal of Real Etate Finance and Economic, 183, Yava, A Middlemen in Bilateral Search Market. Journal of abor Economic 12,

26 Appendix A Simple Model of Agency roblem between Buyer and Buyer Broker The purpoe of thi appendix i to offer a imple model to how the incentive problem between the buyer and the buyer agent. Conider a rik-neutral agent who chooe an effort level to find a property that matche the buyer need and to obtain a low price for the property. In order to tudy whether the agent repreent the interet of hi client in the ame way that he would repreent himelf, we firt tudy the cae in which the agent i buying a property for himelf and then compare thi to the cae in which the agent i helping a buyer to find a property. The purpoe here i not to deign the efficient and incentive compatible contract between the buyer and the agent. Rather, we take the current commiion tructure in the indutry a given and compare the effort level choice of the agent depending on whether he i the buyer of the property or not. et repreent the earch effort of the broker. The probability of finding a property that fit the pecification of the buyer i an increaing function of the earch effort of the broker and i given by Θ where Θ >0 and Θ <0. The cot of earch effort, C, i trictly increaing and convex C >0, C >0 with C 0=0. Suppoe the buyer value the houe at b. If an acceptable match i found, the buyer and eller bargain over the price. I aumed that bargaining will reult in a trade a long a there are gain from trade, i.e., a long a the buyer valuation exceed the eller valuation, b>. The negotiated price divide the urplu b - uch that the buyer receive ω portion of it, and the eller receive 1-ω portion of it. Thi implie that the negotiated price will be = +1-ωb-. For implicity, we will aume equal diviion of urplu, ω=½. 12 Thi yield a negotiated price of = ½+b when buyer with valuation b meet eller with valuation. The valuation of the partie are private information. For the buyer, the valuation of the eller,, i a random draw from the denity probability function f with the upport [, ] and < 0. That i, lower bound of the ditribution i a decreaing function of the effort level of the broker. A higher effort level lead to the poibility of contacting eller who have a lower valuation of the property, which in turn lead to lower expected purchae price of the property. We aume the buyer work with only one broker in earching for a houe to buy. Thi i in line with what we typically oberve in houing market, and eliminate the need to introduce competing broker in the model. We aume the buyer broker i a member of the local multiple liting ervice and help buyer earch through, inpect and viit liting poted on the multiple liting ervice. Following the current practice in the indutry, if a tranaction occur, the buyer broker receive k percentage of the price a commiion from the liting agent for ecuring a buyer for the property, which i typically 40-60% of the total commiion the liting agent ha negotiated with the eller. We aume that the total commiion rate and the buyer agent hare of the commiion, k, are determined in the market and taken a exogenou by the player. 12 It can be hown that the equal plit of the urplu in thi imple etup i the bargaining olution propoed by Nah 1950, Kalai and Smorodinky 1975 and Rubintein