Competition, Financial Constraints, and Appraisal Inflation

Transcription

1 Competition, Financial Constraints, and Appraisal Inflation James Conklin, University of Georgia N. Edward Coulson, University of Nevada Las Vegas Moussa Diop, University of Wisconsin-Madison and Thao Le, Pennsylvania State University September 30, 2016 Abstract In this paper, we examine whether competition in the appraisal industry affects valuation. We find that competition at the MSA/year level has two competing effects. First, appraiser competition decreases the probability of an at-price appraisal. In our baseline regression, a one standard deviation increase in appraiser competition reduces the likelihood of an at-price appraisal by 3.4% relative to the mean. However, the relationship between competition and the likelihood of an at-price appraisal reverses when the borrower is likely to be financially constrained, as proxied by combined loan to value (CLTV) thresholds. For borrowers that are most likely to be financially constrained (100% CLTV), appraiser competition in fact increases the probability of an at-price appraisal. Thus, the effect of appraiser competition on valuation depends on the borrower s financial situation. Key Words: Competition, Appraisal Bias, Credit Constraints, Mortgages JEL Classification: G2, G01, G10, G18, D1, R2 We thank Brent Ambrose, Lynn Fisher, Laurie Lambie-Hanson, Anthony Yezer, and participants at the 2016 American Real Estate and Urban Economics Association National Conference for helpful comments and suggestions. We also thank the Penn State Institute for Real Estate Studies for providing access to the New Century Mortgage database. Terry College of Business, University of Georgia, Athens, GA University of Nevada Las Vegas, Las Vegas, NV and University of California, Irvine School of Business, University of Wisconsin-Madison, Madison, WI Smeal College of Business, Pennsylvania State University, State College, PA

2 I. Introduction Mortgage financing is generally conditional on the certification of the value of the property by an appraiser. In residential purchase transactions, for example, lenders base pricing and loan limits on the lower of the agreed upon purchase price and the appraised value of the property. 1 Appraisers are supposed to provide independent, unbiased estimates of the value of real properties used as collateral to secure the loans since true property values are not directly observable in the market due to asset heterogeneity and infrequent trading. However, the independence of appraisers and the accuracy of appraisals have been widely questioned, particularly during real estate market booms (Cho and Megbolugbe (1996), Chinloy et al. (1997), Lang and Nakamura (1993), Quan and Quigley (1991), Calem et al. (2015), Shi and Zhang (2015), Agarwal et al. (2014), and Tzioumis (2016)). The empirical studies overwhelmingly indicate a persistent upward bias in appraisals, and suggest that interference by mortgage lenders, brokers, and realtors explains the bias (Cho and Megbolugbe (1996)). 2 Moreover, inflated real estate valuations have been identified as a contributor to the 2007 real estate market crash and ensuing severe economic downturn (Agarwal et al. (2014)). The perceived lack of appraiser independence became a serious concern to policymakers as well. In May 2009, Fannie Mae agreed to the Home Valuation Code of Conduct (HVCC) with the Office of Federal Housing Finance Agency (FHFA) and the New York Attorney General to help reinforce the independence of appraisers. The HVCC was later included in the Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 as the Appraiser Independence Regulations (AIR), requiring, among other things, the use of Appraisal Management Companies (AMCs) as a buffer between lenders and appraisers. Although appraisal inflation is well documented, and appraiser independence has been called into question, the root cause of appraisal bias has not been adequately explored in the literature. Put simply, under what conditions is an appraisal likely to be inflated? In this paper, we provide evidence that appraiser competition helps answer this question by showing that appraiser inflation (appraiser s valuations relative to purchase price) varies with the level of appraiser competition. 1 The refinancing of mortgage loans is also dependent on the confirmation of the value of the real estate. In this paper, the discussion centers primarily on purchase mortgages since the empirical tests can only be conducted on those mortgages. 2 This is separate from bias stemming from the use of past transaction data that causes appraisals to lag real estate market cycles. 1

3 Theoretically, the direction of the effect of competition on appraisal quality is ambiguous. On the one hand, it is possible that appraisers facing higher competition have to maintain their reputation by providing credible estimates of value. In this scenario, less skilled appraisers will be pushed out of the market if they have higher operating costs or if lenders demand, and are able to identify, appraiser quality. Alternatively, competition may reduce the informativeness of appraisals. As competition intensifies, appraisers may engage in less costly information production to preserve profit margins. Observed inflation also may be related to appraiser competition even if all appraisers are reporting unbiased opinions of value, since in competitive markets loan originators can more easily shop across appraisers for the most favorable valuation. But a suspicion of many analysts (Cho and Megbolugbe (1996), Shi and Zhang (2015), and Agarwal et al. (2015)) is that appraisers may cater to the interest of end users (e.g, lenders,brokers, realtors, and borrowers) to win future business, and this inclination may be stronger in competitive markets. Evidence suggests that the tendency to inflate values also depends on borrower financial constraints (Agarwal et al. (2015)). On purchase transactions, a below transaction-price appraisal either increases the required downpayment (to keep CLTV constant) or the interest rate (due to higher CLTV), ceteris paribus. A larger required downpayment or a higher interest rate reduces the likelihood of a transaction, particularly if the borrower is financially constrained. There are several ways that an appraiser can become informed regarding the borrower s financial constraints. For example, an informed party (e.g., loan originator) can communicate this information to the appraiser by stating that the appraisal needs to be a certain value in order for the transaction to go through. Alternatively, the appraiser could infer it from the requested loan amount. If the buyer is applying for 100% CLTV financing (no downpayment), an appraiser may infer that the borrower is financially constrained. Financing information may be stated in the appraisal order form (see Figure 2 for a sample) or sales contract (see Figure 3 for a sample), both of which are often available to the appraiser. Thus, the appraiser can determine whether the borrower is financially constrained, and this, in turn, can affect the appraiser s proclivity to cater to end users. Using CLTV thresholds (e.g., 100%, 90%, 80%) to proxy for financial constraints, Agarwal et al. (2015) provide evidence consistent with this hypothesis on refinance loans. In this paper, we ask if the willingness of the appraiser to cater based on the borrower s financial constraints also depends on the level of appraiser competition. 2

4 We examine the effect of appraiser competition on appraisal inflation using a large sample of purchase mortgages originated nationwide during the recent housing boom by one of the largest subprime mortgage originators. Following Calem et al. (2015), we investigate appraisal inflation when it matters most by using the probability of an appraisal being at the contracted purchase price. 3 Calem et al. (2015) argue that an at-price appraisal i) is evidence of appraisal inflation, and ii) makes the appraisal less informative ( information loss ). In line with previous studies, Figure 1 shows that a staggering 47% of appraisals in our sample are at purchase price, with 2% of properties appraised below price and the remaining 51% representing above price appraisals. Assuming appraisal errors are randomly distributed around the expected property value, the high percentage of at-price appraisals is suspicious. Anchoring could explain this left-side bunching up; appraisers receive copies of purchase contracts and they may consciously or unconsciously use those prices as anchors. 4 If anchoring alone explains the distribution of appraisals, then appraiser competition should have no bearing on the likelihood of an at-price appraisal. However, if the high frequency of at-price appraisals is driven by interference in the process by other interested parties (e.g., mortgage brokers and lenders) or by loan originators shopping across appraisers for favorable valuations, then appraiser competition will also be important for appraisal inflation. In this paper, we test whether competition among appraisers affects the likelihood of an at-price appraisal. Our primary measure of appraiser competition is market concentration of appraisers at the MSA level. We also test whether the effect of appraiser competition varies with borrower financial constraints. Similar to Agarwal et al. (2015) we use combined loan-to-value ratio (CLTV) thresholds (100%, 90%, 80%) as proxies for financial constraints. Our results show that appraisal competition reduces the probability of an at-price appraisal by 3.4% for a one standard deviation change in our baseline estimations. Interestingly, appraisal competition has the opposite effect on the likelihood of an at-price appraisal for financially-constrained borrowers. The estimated coefficients of the appraiser competition-cltv threshold interaction terms are positive and statistically 3 Since financing terms (e.g., rate and loan amount) are based on the minimum of the transaction price and appraised value, appraisal inflation is most important when the appraiser s true opinion of value is below the transaction price. An appraisal below the negotiated sales price will generally make the financing terms more burdensome (larger downpayment or higher rate) for the borrower. Moreover, this increased burden will be most problematic for financially constrained buyers. In practice, a below transaction price appraisal reduces the likelihood that the property sale will be completed. 4 It is unclear why appraisers receive copies of purchase contracts. It is possible that the benefits from making this information available to appraisers outweigh potential drawbacks, such as price anchoring. 3

5 and economically significant. We also provide evidence suggesting that the effect of appraiser competition on the probability of an at-price appraisal increased as the housing boom gained steam. To investigate whether appraiser competition affects an individual appraiser s valuation, we control for appraiser fixed effects, thus exploiting within appraiser variation in competition. Intuitively, does the same appraiser value properties differently in competitive versus non-competitive markets? Our main result, that competition reduces the likelihood of an at-price appraisal, unless the borrower is financially constrained, remains after controlling for appraiser fixed effects. Our results are also robust to alternative measures of appraiser competition. In addition, we try to distinguish between appraisal catering versus originator shopping as the mechanism driving our results. The former hypothesizes that appraisers voluntarily bias their valuations in order to win future business from lenders, while the latter postulates that lenders, especially brokers, shop across appraisers for favorable appraisals. The evidence suggests that both appraisal catering and originator shopping likely occurred. II. Literature Review The first paper to document the upward bias problem in appraisal is Cho and Megbolugbe (1996), who compare the appraised values with purchase prices of residential properties from the 1993 Fannie Mae loan acquisition data. In their sample, 30% of the appraisals were identical to and more than 60% were above the purchase prices. Only 5% of the appraisals came below the transaction prices. The authors argue that this highly asymmetric distribution of appraised values is evidence of moral hazard on the part of appraisers. In particular, a low appraisal might prevent the buyer from obtaining the necessary finance from the loan originator and the sale cannot be completed, resulting in lost opportunities for both the buyer and the lender. They thus have an incentive to influence the appraiser to report a high appraisal relative to the true appraised value in his opinion. A subsequent article by Chinloy et al. (1997) supports this contention with an estimated upward bias of 2%. Such findings are at odds with appraisal models theorized in earlier literature, in which appraisers follow an optimal updating rule or backward-looking expectations (Lang and Nakamura (1993); Quan and Quigley (1991)). In a recent paper Calem et al. (2015) propose a simple behavioral rule that is consistent with the observed conduct of appraisers. In particular, to 4

6 avoid unsuccessful transactions due to low appraisals, appraisers substitute the transaction price for the actual appraised value if the latter is below the former (information loss). On the other hand, when the appraisal is above the transaction price, there is no need for inflation and the appraiser reports his estimated value without bias. While it helps facilitate transactions, appraisal inflation inevitably increases default risk through large loan amounts that are too high compared to the true collateral values. LaCour-Little and Malpezzi (2003) show that appraisal bias was associated with more frequent defaults in the Alaska market in the 1980s. Research on the recent mortgage crisis by Ben-David (2011) also indicates that excessive valuation of collaterals was to blame for the ensuing turmoil, because borrowers could have already been in a negative equity position at the time of mortgage origination. The model proposed in Calem et al. (2015) implies that appraisers have to balance the tradeoff between the increased default risk versus the cost of a failed transaction. Accordingly, when the perceived credit risk of a mortgage is lower, the incentive for appraisers to engage in substitution is strengthened. Consistent with this reasoning, Calem et al. (2015) find that rising house prices and decreased foreclosure rates reduce the probability of low appraisals, while a high transaction price relative to other properties in the same neighborhood, which is viewed as an indicator of high perceived default risk, is more often associated with a low appraisal. Using data on mortgage refinancing and their subsequent sale transactions, Agarwal et al. (2015) again confirm the presence of appraisal bias. Further analysis reveals that the extent of bias depends largely on the motivations of mortgage originators and borrowers to influence appraisers. For example, valuation inflation is much more pronounced among refinance mortgages originated by brokers because they do not bear the default risk, and among loans with high LTV ratios because financially constrained borrowers have stronger incentives to influence appraisers. Judgment biases due to feedback and anchoring have also been found in commercial appraisals in Hansz and Diaz III (2001) as well as Clayton et al. (2001). However, Tzioumis (2016) offer an alternative observation that there is no association between appraisal inflation and subsequent work volume, implying no incentives for appraisers to engage in such behaviors. A few papers examine whether the use of AMCs as an intermediary between appraisers and lenders helps increase the objectivity of the former (Agarwal et al. (2014); Calem et al. (2015); Ding and Nakamura (2015)). The 2009 Home Valuation Code of Conduct, which mandated the use of AMCs for government- 5

7 sponsored enterprise loans, is found to have reduced the magnitude of appraisal bias by as much as 45% in Agarwal et al. (2014). One important missing piece in the literature discussed above is the impact of competition on the behaviors of appraisers, which is the focus of this paper. A rich literature exists that studies the link between competition and reporting bias of financial agencies, especially in the case of credit ratings and analyst forecasts. For example, that rating agencies often engage in rating inflation is a well documented phenomenon. The issuer-pay business model adopted by the majority of credit rating agencies naturally creates a classic conflict of interest problem: issuers will shop for favorable ratings among agencies, leading to widespread rating inflation. Generally, there are two competing views regarding the disciplinary role of competition among credit rating agencies and stock analysts. On one hand, it is argued that agencies facing higher competition have to maintain their reputation by providing credible reports. On the other hand, the alternative view holds that competition forces agencies to bias their opinions to cater to the interest of the end users (investors, security issuers, consumers, etc.) in order to win business. Developing a theoretical model on competition in the credit rating industry, Camanho et al. (2009) posit that the level of reputation of the competing agencies determines which force will ultimately prevail. Specifically, if all agencies have similar reputation, each of them has the potential to become the market leader through providing more accurate ratings, implying that competition reduces bias. However, if some agencies have a much better reputation than others, the latter effect will dominate because the lower reputation group have an incentive to inflate ratings to gain market share. Supporting the first view that competition has a positive effect, Hong and Kacperczyk (2010) find that a decrease in the number of analysts covering a particular stock leads to higher optimism bias in the earnings forecasts of that stock. The authors use mergers of brokerage houses and their ensuing firing of analysts as an exogenous shock to competition. In the context of credit rating, Xia (2014) studies how Standard and Poors (SP) responded to the entry of Egan-Jones Rating Company (EJR), with the two representing issuer-paid and investor-paid agencies, respectively. He shows that ratings issued by SP became more conservative and responsive to changes in credit risk following EJRs entry. These findings are in contrast with another camp in the literature advocating the alternative view that competition results in more reporting distortions. Using the market share of Fitch as a measure of competition with Moodys and SP, Becker and Milbourn 6

8 (2011) document a positive association between competition and corporate bond ratings during the period A similar finding is found in by Cohen and Manuszak (2013) for the ratings of commercial mortgage-backed securities (CMBS) by Fitch. However, in a recent paper, Bae et al. (2013) argue that Fitchs market share is subject to a potential endogeneity issue caused by industry-wide characteristics. After correcting for the endogeneity problem, the authors do not find any significant relation between Fitchs market share and rating inflation. Another working paper by Flynn and Ghent (2015) employ the entry of Morningstar Credit Ratings LLC and Kroll Bond Ratings into the CMBS rating market as an exogenous increase in competition for the four incumbent agencies (Moodys, SP, Fitch, and Dominion Bond Rating Service). Their data on the ratings of CMBS from 2009 to 2014 show that the entrants issued higher ratings than the incumbents in order to win business. In response, the ratings by the incumbents also became more generous as the entrants increased their market share. Focusing more on the information content of rating rather than rating inflation, Doherty et al. (2012) theorize that the entry of a new credit rating agency helps improve information disclosure. The entrant can provide additional information to buyers/investors by differentiating their rating scale from that of the incumbent. Consistent with this theoretical prediction, their empirical test using the entry of SP into the rating market for insurance companies, which had been dominated by the monopoly A.M. Best Company until the late 1980s, provide evidence that SP applied a more stringent rating policy. More notable is the observation that large, successful but opaque insurers are more likely to seek a second rating by SP in order to signal their quality. Another paper by Bolton et al. (2007) also finds that competition enhances information disclosure in the case of financial services. III. Methodology Since the focus of this paper is on appraiser competition and its effect on valuation, we must first define a measure of appraiser competition. Empirical studies typically assume that market concentration is inversely related to competition (Berger et al. (2004)), with the Herfindahl-Hirschman Index (HHI) serving as the most widely used proxy for market competition (Bikker and Haaf 7

9 (2002)). Following this practice, we calculate our primary measure of competition as follows: n Compet mt = (HHI mt ) = ( Skmt 2 ) (1) where Compet mt (HHI mt ) measures market competition (concentration) in MSA m and year t and S kmt represents the market share of appraiser k in MSA m and year t. Both S kmt and HHI mt range between zero and one. A HHI of one implies a monopolistic market while a HHI near zero implies perfect competition. We take the negative of the HHI so that an increase in Compet mt can be interpreted as an increase in competition. The construction of Compet mt will be discussed further in Section IV.A. 5 In the first part of our analysis, we investigate whether appraiser competition affects the probability of an at-price appraisal by estimating a linear probability model of the following form 6 : k=1 At P rice icmst = β 0 + β 1 Compet mt + β 2 Constrained i + β 3 Compet mt Constrained i +X iβ 4 + Z mtβ 5 + β 6 W ct + α s + α t + ε imst, (2) where At P rice imst is a dummy variable that takes a value of one if the appraisal on application i, in county c, MSA m, state s, and year t equals the purchase price. Constrained i measures whether an individual borrower is likely to be financially constrained; X i is a vector of loan characteristics, Z mt is a vector of MSA characteristics at time t, α s are state fixed effects that control for time-invariant state characteristics, α t are year fixed effects that control for nation-wide changes in economic conditions, and ε imst is the error term with standard properties. X i includes the combined loan to value ratio at origination, the natural logarithm of the purchase price, and a dummy variable that takes a value of one if the loan application was submitted by a third-party originator (e.g. mortgage broker). 7 Our MSA-level controls, Z mt, include house price appreciation over the previous 5 In robustness checks discussed below, we also use two alternative measures of competition commonly used in the literature (Bikker and Haaf (2002)). 6 Our main results remain unchanged when we use a logit model rather than a linear probability model. 7 In estimating Equation 2, we assume that the sales price and downpayment, and thus the initial CLTV, are negotiated between the borrower and the lender prior to, but contingent on, the completion of the appraisal. Put differently, we assume that CLTV is exogenous. Since the CLTV is based on the minimum of the appraisal or the transaction price, as long as the appraisal is at or above the transaction price then this assumption is not violated. However, if the appraisal is below the contract price, CLTV will be endogenous. We believe this to be a relatively minor issue since the appraised value is rarely below the transaction price. 8

10 quarter and the previous year and house price volatility over the previous 20 quarters. Since market liquidity affects the appraisers ability to find comparable sales for the subject property, Z mt also includes the natural logarithm of the total number of sales transaction in MSA m in year t from the Home Mortgage Disclosure Act (HMDA) data. W ct is an analogous liquidity measure at the county level. We are primarily interested in the coefficients β 1, β 2, and β 3. β 1 tests whether appraiser competition affects the probability of an appraisal equal to the transaction price. β 2 indicates whether financial constraints are related to an at-price appraisal. A positive coefficient on β 2 suggests that appraisers are more likely to hit appraisal targets when borrowers are financially constrained. Finally, β 3 tests whether the relationship between competition and At P rice varies according to the borrower s financial constraints. IV. Data A. Sample Construction Our primary database comes from New Century Financial Corporation (NCEN), one of the largest subprime mortgage lenders in the years leading up to the mortgage crisis. The data includes both funded loans and unfunded loan applications. NCEN collected detailed loan-level information during the underwriting process, including, but not limited to, borrower characteristics, property characteristics, and contractual features of the loans. Both the purchase price and the appraised value are reported in the data. The appraised value can be below, above, or exactly at the agreed upon purchase price. The main dependent variable of interest in our study takes a value of one if the appraised value equals the purchase price. Our second source of data is the Federal Housing Finance Agency s (FHFA) quarterly MSA house price index. To account for local house price trends, we match the NCEN data with the FHFA data and calculate the house price appreciation over the previous quarter and year for each mortgage application as well as house price volatility over the previous 20 quarters. MSA unemployment data at the time of origination is collected from the Bureau of Labor Statistics. Finally, we collect MSA level income data from the Bureau of Economic Analysis. Since we want to examine the impact of appraiser competition on the probability of an at-price appraisal, we need to be able to identify individual appraisers to construct our competition mea- 9

11 sure. An important feature of the NCEN data is that the appraiser s full name is recorded. From this information we are able to track each appraiser s market share within geographic locations over time, which allows us to construct our main independent variable of interest, the appraiser Herfindahl-Hirschman index (HHI). As noted, the HHI is calculated for each MSA/year as the sum of the squared market shares of all the appraisers operating within that MSA in that year. 8 To ease interpretation, we multiply this index by negative one so that a higher value of our appraiser competition measure represents greater competition. The possible range for our appraiser competition variable is negative one to zero, with negative one indicating a monopoly and zero suggesting perfect competition. 9. Figure 4 shows the level of appraiser competition for MSAs in our sample in Perhaps not surprisingly, competition among appraisers is relatively high in the so-called Sand States (Arizona, California, Florida and Nevada). However, even within these states, variation in appraiser competition exists across MSAs. In estimating Equation 2, we include state fixed effects, so identification depends on within-state variation in appraiser competition. A potential concern with our competition measure is that we only observe appraisals (and appraisers) for applications that ended up with NCEN. By using applications from only one lender, our measure of competition assumes that competition in the NCEN data serves as a good proxy for the overall level of appraiser competition within an MSA/year. We believe this is a reasonable assumption with the intuition as follows. Residential appraisers rarely (if ever) work exclusively for one mortgage broker. In this sense, appraisers operate independently from mortgage brokers. 11 If this independence holds, then the sample of appraisers in each MSA in the NCEN data should be representative of the population of appraisers in that MSA as a whole. The same logic holds for each appraiser s market share. 12 As a result of appraiser independence, our measure of appraiser competition should provide a reasonable proxy for true MSA/year appraiser competition. 8 Since appraisers are typically paid a fixed fee for an appraisal, we calculate the HHI based on the number of loan applications, rather than the dollar value of the applications. 9 Our MSA/year-level competition measure is calculated using both refinance and purchase applications. However, for reasons outlined below, our primary loan level regression include only purchase applications. Also, when calculating competition, we drop all observations where the appraiser s name is blank. 10 We calculate appraiser competition in each year for each MSA. As discussed below, appraiser competition within and MSA is highly persistent. 11 Appraiser independence (or lack thereof) is commonly used in the literature to refer to other agents (e.g. borrower, broker, lender) ability to influence the appraiser s valuation. Here, the term takes on a different meaning. We define an appraiser as independent if she does not work exclusively with a single broker or lender. The distinction is important. For our competition measure to be representative, we require that the appraisers do not work exclusively for a single broker, but we do not require that the appraiser s valuation cannot be influenced by other agents. 12 Tzioumis (2016) uses a similar argument regarding the independence of appraisers and appraiser market share. 10

12 Since At price can only be estimated on purchases, and several of our control variables are measured at the MSA level, we restrict our sample to purchase applications located within an MSA for loan-level regressions. Also, because the construction of our competition measure requires the appraiser s name to be filled out, and this field is sparsely populated on non-funded applications prior to 2003, our sample period covers 2003 through the first quarter of We drop observations in MSAs that have less than 100 applications in a given year, and we include only first lien mortgages with a minimum loan amount of $40,000. To avoid data entry errors, we drop observations with FICO scores below 350 or above 850. We also exclude observations where the reported CLTV is below 25% or above 100%. Finally, we drop observations where the appraised value is more than three times larger than the purchase price. The resulting sample includes 364,257 purchase applications. B. Summary Statistics Consistent with previous studies, a large portion (47%) of the appraisals in our initial sample come in exactly at the purchase price (Figure 1). In contrast, only 2% of appraisals are below the purchase price, with the remainder above purchase price. The summary statistics presented in Table I are based on the final sample of 356,440 transactions, after excluding below-price appraisals for the reasons explained in Section V.A. Consequently, the proportion of at-price appraisals is slightly higher in our final sample. Our primary independent variable of interest, appraiser competition, ranges from to , with a mean of We also include in Table I the three alternative measures of appraiser competition (CR2, CR4, and appraisers per capita) used in the robustness tests. They will be discussed in section V.E. Turning to loan characteristics, the average CLTV at origination is 94%. This reflects the fact that NCEN originated subprime mortgages, and that for subprime purchases, down payments were typically small. In fact, almost half of the borrowers made no down-payment (100% CLTV). Consequently, it is fair to assume that many of the borrowers were likely financially constrained. 14 The average purchase price and loan amount are $223,000 and $185,000, respectively. 15 Although 13 NCEN stopped originating loans and filed for Chapter 11 bankruptcy in March An alternative interpretation is that borrowers had the financial means to make large downpayments, but chose not to. We cannot rule out this explanation. However, we view this as unlikely since the loans in our sample are subprime, and subprime loans were concentrated in lower income areas (Mayer and Pence (2008)). 15 The CLTV at origination is much higher than that implied by the average loan amount and purchase price. This 11

13 NCEN originated loans through both the wholesale and retail channels, nearly 90% of the applications came from third party originators (e.g. brokers). V. Results A. Multinomial Analysis First, we investigate the role of appraiser competition on the likelihood of an at-price appraisal using a multinomial approach. There are three possible appraisal outcomes for purchase transactions: appraisals that equal the transaction price, those that fall below the transaction price, and those that are above the transaction price. To account for the polytomous nature of the appraisal price outcome for purchase transactions, we model the probability p ij of appraisal outcome j {below price, at price, above price} on loan i with the multinomial logit specification: p ij = exp(γ i β j) 3 j=1 exp(γ i β j) (3) where, for the sake of simplicity in notation, Γ i includes all control variables from the binary model (2). The β j s recovered from the estimation of the multinomial logit model will be used to determine whether the appraisal outcome varies with appraiser competition. For a given control variable, for example appraiser competition (Compet mt ), the estimated marginal effects of a one standard deviation increase in our competition measure (0.022) for the three possible appraisal outcomes will sum to zero. In other words, if competition increases the probability of an at-price appraisal, it must decrease the probability of one or both of the other outcomes. Table II presents the estimated marginal effects of a one standard deviation increase in appraiser competition on the likelihood of different appraisal outcomes. Outside of the CLTV thresholds included in our model, appraisal competition is negatively associated to the likelihood of a property being appraised at below-price or at-price and positively related to an above-price appraisal. Even though these estimates are statistically significant, most of the variation in probabilities is between at-price and above-price appraisals. The marginal effects of competition at CLTV thresholds also is due to the fact that many of the purchases had second mortgages associated with the transaction. We exclude second mortgages from our analysis to avoid double counting of the same transaction. Also, since silent seconds are not reported to the originator of the first mortgage, our estimate of CLTV does not reflect unreported mortgage liens. 12

14 display a similar pattern in magnitude across the appraisal outcomes. However, for financiallyconstrained borrowers (100%-CLTV home buyers), 16 appraiser competition has a positive effect on the probability of at-price appraisal and a negative effect on above-price appraisals. Figure 5 provides a visual representation of these marginal effects for the three appraisal outcomes. Interestingly, we find no evidence that competition significantly affects the probability of a below price appraisal, regardless of the CLTV threshold (top right corner of Figure 5). A priori, we would expect the probability of a below-price appraisal to be affected by competition. For example, in a competitive environment, it seems plausible that an appraiser would be more inclined to inflate the value of a property to either the contract price or above to avoid the cancellation of the sale. However, the below-price results suggest that the likelihood of a below-price appraisal, conditional on other covariates, is not affected by appraiser competition. There are two possible explanations of this result. First, it is possible that below-price appraisals lead to sale cancellation before the application ever reaches the lender (NCEN). Thus, most below-price appraisals never show up in our data. A second and more plausible explanation, particularly during the period covered by our study, is that initial below-price appraisals were intentionally revised upward by appraisers (the catering hypothesis) or shopped around by lenders or brokers (the shopping hypothesis), thus appearing as at- or above-price in our sample. However, at-price adjustments appear more likely given the large fraction of at-price appraisals found in our data and the fact that above-price appraisals represent roughly 50% of the sample. We will explore this issue later. Thus, virtually all of the shifting of probabilities associated with appraiser competition occurs between the at-price and above-price groups. Table II and Figure 5 show that appraiser competition decreases the probability that an appraisal equals the transaction price and increases the likelihood of above-price appraisals at non-threshold CLTV values. For at-price appraisals, the marginal effect of appraiser competition decreases with CLTV, with higher CLTVs (more credit-constrained borrowers) being associated with higher probabilities of being appraised at price. The graph in the bottom left corner of Figure 5 depicts this negative relationship between the likelihood of appraisal at price and CLTV. Given the requirement that marginal effects add to zero across the outcomes, appraiser competition is positively related to CLTV thresholds for above price appraisals (top left 16 Borrowers making no downpayment are likely to be financially constrained. Thus, on a transaction with no downpayment, an appraisal below contract price, which requires either contract renegotiation or additional cash from the borrower for the downpayment, will increase the likelihood of sale cancellation. 13

15 corner of Figure 5). In fact, this relationship is almost the mirror image of the corresponding at-price relationship due to the trivial impact of below-price estimates. It is clear from Figure 5 that appraiser competition largely affects the probability of appraisal at price or above price. Since the multinomial estimation effectively reduces to a dichotomous outcome analysis, we will use a linear probability model to estimate the likelihood of at-price appraisal following equation (2) for the remainder of the paper. After excluding below-price observations, our final sample includes 356,440 observations. B. The Relationship Between Appraiser Competition and At-Price Appraisals Table III presents the coefficient estimates from the linear probability model of Equation (2). The CLTV thresholds have a large impact on the the appraisal outcome. The probability of an at-price appraisal increases by 7.5% if the borrower is not making a downpayment (CLTV 100%), consistent with the idea that appraisers try not to be an obstacle in getting a transaction completed. As noted, it is highly probable that most of the at-price appraisals were adjusted upward as a result of appraiser catering or broker shopping activities. We also include other leverage thresholds (90% CLTV and 80% CLTV). Like the 100% CLTV indicator, the 90% and 80% CLTV thresholds may indicate financial constraints, but they are also important because interest rates increase significantly above these thresholds. Thus, an appraisal hitting the transaction price when the CLTV is 90% (or 80%) affects the downpayment amount and loan pricing. These CLTV thresholds are also positively related to the likelihood that the appraised value equals the transaction price. The coefficients of the CLTV thresholds are basically of similar magnitude; Wald tests fail to reject the null hypothesis that the coefficient on 100% CLTV differs from that of 90% or 80% CLTV. To test the effect of appraiser competition across various home buyer groups, equation (2) includes our appraiser competition measure and its interactions with the three CLTV thresholds. Table III shows that appraiser competition is significantly negatively related to the probability of an at-price appraisal (β 1 < 0) outside the CLTV thresholds. A one standard deviation increase in appraiser competition is associated with a 1.6 percentage point decrease in the likelihood of an at-price appraisal, or a 3.4% reduction in the probability of At P rice relative to the mean This is calculated as the coefficient estimate time the standard deviation divided by the share of at-price appraisals ( /0.483). 14

16 Even though appraiser competition reduces the likelihood of the property being appraised at price for this group of borrowers who are less likely to be financially constrained, it could result in few business losses for lenders. When the borrower is financially constrained, however, the effect is reversed: appraiser competition is now positively or less negatively related to the probability that the appraisal value equals the sale price. The reversal of the relationship is stronger when the financial constraints become more binding. At the 100% CLTV threshold, the effect of appraiser competition is positive (β 1 + β 3 = 0.288). At this threshold a one standard deviation increase in appraisal competition is associated with a 0.6 percentage point increase in the likelihood of an at-price appraisal, or a 1.3% increase relative to the mean. 18 For 90% and 80% CLTVs, a similar calculation yields 0.3% and 1.1% decreases in the likelihood of appraisal at price at the mean for a one standard deviation change in appraisal competition, respectively. Compared to the 3.4% decrease in probability for financial unconstrained borrowers, these results are consistent with the hypothesis that appraisers compete by making sure that the appraisal comes in at the transaction price when a below price appraisal would likely result in lost business or a significantly higher contract rate for the borrower. Alternatively, our results are also consistent with loan originators being able to shop for a value among appraisers in areas where appraisers compete more fiercely. In either case, competitive pressure increases the probability that an appraisal equals the contract price when the expected costs of foregone lending are high or when a higher CLTV will result in an increase in interest rate. Through the loan screening process, originators observe whether a borrower is financially constrained, and thus the importance of shopping for a favorable valuation. However, it is perhaps less clear how the appraiser would know if the borrower is financially constrained. There are several ways that an appraiser can become informed regarding the borrower s financial constraints. First, an informed party (e.g., loan originator) can communicate this information to the appraiser. For example, a loan originator could communicate that the appraisal needs to be a certain value or the transaction will fall-through. This type of interference by originators was supposedly common practice during the last housing boom. The requirement that appraisal requests must now be channeled through AMCs in charge of independently assigning appraisers is meant to address this issue. But appraisers may still feel the need to cater to AMCs to secure future business since 18 That is: /

17 appraisers compete for market share from AMCs. In other words, even though regulations limit the scope for appraisal interference by lenders and brokers, appraiser competition may still matter for the informativeness of appraisals. A second way that appraisers may learn of the financial constraints of borrowers is through knowledge of the loan amount on the transaction. Figure 2 shows an appraisal company s order form from The form requests information on the purchase price, the amount of the first mortgage, and the amount of the second mortgage. With this information, the appraiser can determine whether the borrower is financially constrained as proxied by CLTV thresholds. Discussions with industry participants confirm that disclosing loan information to the appraiser via the order form was common practice leading up to the housing bust. Finally, appraisers typically receive a copy of the sales contract. Along with the agreed upon purchase price, the sales contract can contain information on the requested loan amount. Figure 3 shows a sample agreement of sale where the mortgage information is disclosed in Sections 1.B - 1.E. 20 Taken together, it seems likely that appraisers were aware of borrower financial constraints in the run-up to the crisis. Although recent regulations may reduce the problems of appraiser shopping and appraiser catering, it is not likely that the problems have been fully eliminated. We turn to a discussion of the control variables in equation (2). One of the more surprising results is the negative (and statistically significant) coefficient on third-party originations, given the common perception that brokers are more inclined to make riskier loans which would in turn be more likely to have at price appraisals. The unconditional correlation between broker-originated loans and at price appraisals is indeed positive, although small and insignificant. The inclusion of other loan characteristics, especially CLTV thresholds, presumably controls for that particular aspect of the brokerage business. The coefficients on (log) purchase price and MSA price appreciation are positive, with small standard errors. This suggests that at price appraisals are more likely to occur within a milieu of active markets with strong demand. This is not simply a question of difference in behavior between so-called bubble vs. other states, nor is it simply a question of boom behavior, since both state and time fixed effects are included in the model. But it does suggest that more active markets lead to less accurate appraisals, other things equal. The temptation to use short cuts, such as using the 19 The order form is available at 20 The sample agreement is available at Sample-AgreementToPurchaseRealEstate.pdf. 16

18 purchase price as an anchor, might be greater in such circumstances. Moreover, in a market with consistently rising prices, the cost (in the form of future default probability) of an informal at-price appraisal is surely lower. The positive coefficient on number of transactions in the same county accords with this. There are two other coefficients to mention. One is the positive, though small and imprecise, coefficient on price volatility. The time period we examine is one in which prices were consistently rising in all markets, so that relative volatility does not have a lot of explanatory power. Finally, while the coefficient on (log) county sales is, as noted above, positive, that of (log) MSA sales is negative, though of smaller magnitude. This seems to be contrary to the analysis of Lang and Nakamura (1993) who posit that a large number of local sales would lead to more accurate appraisals. The estimates suggests that it is the market activity in the broader geographic area that creates more informative appraisals. To summarize, Table III shows that appraiser competition is negatively related to the probability that an appraisal equals the transaction price. However, when the borrower is likely to be financially constrained or the CLTV is at an interest rate threshold, appraiser competition has a positive or a smaller negative effect on the probability that the appraised value equals the contract price. C. Appraiser Competition Over Time We examine whether the effect of appraiser competition varies over time by estimating Equation (2) separately for each origination year and report the coefficient estimates in Table IV. The relationship between appraiser competition and At P rice for transactions not at the included CLTV thresholds appears to change over our sample period. Although the coefficient estimates are negative, in 2003 and 2004 appraiser competition was not significantly related to At P rice. However, for the origination cohorts, loans originated in more competitive appraiser markets are less likely to have an at-price appraisal. Furthermore, the magnitude of the appraiser competition effect increases over time. Again, home buyers outside of the CLTV thresholds are less likely to be financially constrained and therefore less affected by a below-price appraisal. The main effects of the CLTV thresholds (100%, 90%, and 80%) are positive and significantly related to the probability that an appraisal comes in exactly at the purchase price in every year, consistent with the hypothesis that appraisers make sure that their appraisal does not interfere 17

19 with a successful transaction when financial constraints are likely to be binding. Turning to the interactions between CLTV thresholds and appraiser competition, Table IV shows that when borrowers are most likely to be financially constrained (100% CLTV), appraiser competition increases the likelihood that an appraisal comes in exactly at the transaction price for loan applications from Furthermore, the effect increased almost monotonically with the housing market boom. The interaction between appraiser competition and 90% CLTV is positive in all years as well, and statistically significant in Although the interaction between 80% CLTV and appraiser competition is positive, it is not statistically significant in any of the years. This is consistent with borrowers at 80% CLTV being less financially constrained. These results suggest that competitive pressures increase the probability that the appraisal equals the transaction price when financial constraints are likely to be binding. Overall, the fact that stronger housing market activity is associated with more at-price appraisals is confirmed in these regressions. Moreover, the mechanism appears to obtain most strongly in competitive appraiser markets. Appraiser competition appears to decrease the probability At P rice for non-financially constrained home buyers, with a larger effect during the peak of the housing market boom. Across all years, borrower financial constraints, as proxied by CLTV thresholds, increase the probability that the appraisal equals the transaction price. Moreover, the relationship between financial constraints and At P rice is stronger in markets with higher levels of appraisal competition. D. Appraiser heterogeneity A potential concern with the preceding analysis is that appraiser heterogeneity may be driving our results. If unobservable characteristics of appraisers vary with market competition, and these characteristics are correlated with At P rice, then we would incorrectly attribute the effects of appraiser unobservables to appraiser competition. To assuage these concerns, we estimate the linear probability model of Equation (2) with appraiser fixed effects to exploit within appraiser variation in appraiser competition across MSAs over time. Intuitively, does the same appraiser estimate value differently depending on the level of local appraiser competition? For this exercise, we limit our sample to appraisers that completed at least 20 appraisals to allow for within-appraiser 18

20 variation, reducing the number of observations to fewer than half the original sample. 21 Column 1 of Table V presents the coefficient estimates from this estimation. Even after controlling for appraiser heterogeneity, appraisal competition is still negatively related to At P rice for the nonthreshold CLTV observations. Moreover, the magnitude of the coefficient (-0.809) is similar to the coefficient reported in Table III where we did not control for appraiser fixed effects. Also, all three CLTV thresholds measuring borrower financial constraints are significantly positively related to the probability that an appraisal matches the transaction price. Consistent with the results in Table III, appraiser competition is also positively related to the probability that the appraisal equals the transaction price at these CLTV thresholds. The estimated effects of the other control variables are generally similar to the results in Table III. Column 2 of Table V presents the estimation results for appraisers that completed fewer than 20 appraisals. Since we did not include appraiser fixed effects in this estimation, the effect of appraiser competition is therefore derived from variation across appraisers. The results from the estimation with appraiser fixed-effect still obtain, although some coefficient estimates are slightly smaller. The results in Table V suggest that unobservable differences in appraisers across MSAs and time are unlikely to drive the relationship between appraiser competition and At P rice. The same appraiser is less likely to estimate the value of a property exactly at the transaction price in more competitive markets. However, if the borrower is likely to be financially constrained, the relationship is reversed. Thus, for a given appraiser, competition will increase the likelihood of appraising at transaction price if the borrower is financially constrained. E. Alternative measures of appraiser competition In the previous sections, we used industry concentration (-HHI) as our measure of appraiser competition. As a robustness check, we repeat our main regression using three alternative measures of market competition. Our first alternative measure of competition is the concentration ratio of the two largest appraisers in each market (CR2). This is calculated by summing the market shares (based on the number of transactions) of the top two appraisers in an MSA in a given year. Similarly, we calculate the concentration ratio of the top four appraisers (CR4) as our second alternative 21 There are 3,573 appraisers that complete at least 20 purchase appraisals. Of these, 70% appraise properties in more than one MSA. 19

21 measure of appraiser competition at the MSA level. Concentration ratios range from zero to one, with higher values indicating lower competition, and are commonly used in the literature as proxies for market competition (Bikker and Haaf (2002)). 22 As our third alternative measure of appraiser competition, we divide the number of appraisers in an MSA each year by the MSA population. As expected, our three alternative measures of competition are highly correlated with our primary competition measure. 23 Table VI reports the coefficient estimates from Equation (2) using our alternative measures of appraiser competition. All of the results using CR2 in column (1) are consistent with our main results in Table III. First, higher levels of appraiser competition are associated with a lower likelihood of at-price appraisal for non-financially constrained borrowers (control group). Second, financial constraints are positively related to the likelihood of at-price appraisal. Finally, when the borrower is likely to be financially constrained, competition is positively related to At P rice, with a larger effect observed when the constraint is more likely to be binding (e.g. 100% vs. 80%). In column (2), we use our second alternative measure of appraiser competition (CR4) and the results are nearly identical to those in Column (1). Using the third alternative measure of competition (Appraisers per Capita) in column (3) largely leaves our results unchanged. One notable difference, however, is that the coefficient on 100% CLTV is not statistically significant in column (3). Table VI suggests that, ceteris paribus, an appraisal equaling transaction price is in general less likely in markets with greater appraiser competition. However, competitive pressures increase the likelihood of At P rice when a below transaction price appraisal is most likely to inhibit transaction completion. The signs and significance of the other control variables in Table VI are similar to those in Table III. In summary, our results regarding appraiser competition and the probability of an at-price appraisal are robust to our alternative measures of appraiser competition. F. Appraiser Catering or Broker Shopping? Appraisers have been accused of colluding with lenders to inflate home values, and that this collusion helped fuel the housing market boom and subsequent bust. 24 As a key component of the 22 As with our primary measure of appraiser competition, we multiply the concentration ratios by negative one so that higher values of these variables reflect greater market competition. 23 The correlation between our primary competition measure and CR2, CR4, and appraisers per capita is 0.95, 0.93, and 0.20, respectively. 24 This belief led to the enactment of the HVCC requiring the use of AMCs as part of the Dodd-Frank Act. 20

22 appraisal process, appraisers are expected to instill their educated opinion in valuation, but this might be subverted in two ways. The appraiser catering hypothesis postulates that for economic reasons, an appraiser may be more likely to issue an at-price appraisal if his initial assessment of the property value falls below the contract price. Appraisers may want to protect their market shares or win additional business from lenders (Chinloy et al. (1997), Agarwal et al. (2015), Calem et al. (2015)). Consequently, at-price appraisals should be more prevalent in competitive markets. 25 An alternative explanation of the prevalence of at-price appraisals blames originators, particularly brokers, of shopping around below-price appraisals that could result in lost business. We refer to this theory as the shopping hypothesis. Appraisals shopping by brokers is more likely to succeed in competitive markets where the broker has many appraisers to shop from. 26 Since both the catering and shopping hypotheses lead to the same outcome, it is difficult to disentangle these two explanations. Nonetheless, we undertake a preliminary examination of this subject by exploring the effects of appraiser experience and origination channel on the likelihood of at-price appraisals. Table VII shows the results of the at-price estimations in competitive and uncompetitive appraisal markets using a model specification controlling for appraiser experience. Competitive areas (column 1) regroups loans in MSAs where appraisal competition is above the median of our sample, while column (2) includes MSAs at below-median appraisal competition. For these estimations, we restrict our sample to brokered loans 27 and control for overall appraiser experience as well as appraisers experience with brokers. First, appraiser experience (whether measured by the number of transactions previously completed or in years) does not appear to matter in competitive markets. However, appraisers that have worked with more brokers during the previous year are more likely to appraise properties at price, particularly in competitive markets. This result supports the catering hypothesis. In contrast, the likelihood of at-price appraisal is negatively related to the share of business from the same broker during the previous year, particularly in competitive markets. The non-existence of a quid pro quo arrangement between appraisers and brokers represents a challenge to the catering hypothesis. Next, we examine the determinants of at-price appraisals for brokered and retail loans using 25 But the size of the appraiser should also be a determinant factor since larger appraisers should be able to withstand that pressure. 26 It is also possible that appraisal shopping is countercyclical. But we cannot test this since the data only cover the expansionary period. Appraisal competition could also vary with the business cycle. 27 We exclude retail loans since NCEN is sole lender in our data. 21

23 equation (2). Both brokered and retail originators face costs in the event that appraisal shopping is detected by the lender. However, these costs are likely much higher for the retail originator who is an employee of the lender. If the lender discovers that one of its employees deliberately withheld information in the origination process (e.g., a below-price appraisal), the employee is likely to incur significant penalties such as termination of employment. For a broker, the consequences are less severe since he typically has business relationships with many lenders. Intuitively, if the lender terminates the business relationship, the broker still has many other lenders he can work with. We estimate the regression of equation (2) separately for brokered and retail loans in Table VIII. In line with our baseline results, Table VIII shows that appraiser competition is negatively related to At P rice, particularly for brokered loans not at the CLTV thresholds. More importantly, appraiser competition has a stronger positive effect on the likelihood of an at-price appraisal for brokered loans made at 90% and 100% CLTV relative to retail loans at the same CLTV thresholds. This provides some support for the appraisal shopping hypothesis. In summary, although the evidence of appraisal smoothing at the contract price is undeniable, distinguishing between the shopping and catering hypothesis is difficult. Most likely, both shopping and catering occurred in practice. Although the use of AMCs (as required by Dodd-Frank) may curb originator shopping and influence on valuation, it will not address the catering problem since an appraiser typically has access to the transaction price via the sales contract. VI. Conclusion According to the Appraisal Institute The role of the appraiser is to provide objective, impartial, and unbiased opinions about the value of real property. 28 However, researchers have long questioned whether appraisals are in fact objective, impartial and unbiased, and two consistent findings in the literature seem to be at odds with unbiasedness and objectivity. First, multiple studies provide evidence of an upward bias in appraisals. Second, a large share of appraisals on purchase transactions have an estimated value that equals the negotiated contract price. What drives these facts, however, remains elusive. Obviously, the appraiser s incentives must be considered for any explanation of appraiser bias or inflation. To the authors knowledge, no previous study has inves

24 tigated the role of competition in appraisal bias. However, competition should be relevant, since in the absence of competition, appraisers may have little incentive to exert costly effort to produce accurate appraisals. On the other hand, with no competition from other appraisers, the influence of other agents (e.g., lenders and brokers) on the appraiser s valuation may be minimal. The point is, market competition between appraisers needs to be considered in any explanation of appraisal bias or inflation. In this paper, we examine whether competition in the appraisal industry affects the accuracy of appraisals using a sample of purchase applications from a large subprime lender. In our analysis, we find that competition has two opposing effects. First, appraiser competition decreases the probability of an at-price appraisal. In our baseline regression, a one standard deviation increase in appraiser competition at the MSA/year level reduces the likelihood of an at-price appraisal by 3.4% relative to the mean. However, the relationship between competition and At P rice reverses when the borrower is likely to be financially constrained, as proxied by CLTV thresholds. Thus, depending on the borrower s financial situation, competition can have positive or negative effects on appraisal quality. We also find that the relationship between competition and at-price appraisals changed over time. In the later years of our sample ( ), competition has a larger impact in reducing the likelihood of an appraisal exactly equaling the transaction price. Additionally, we exploit within appraiser variation in market competition, and find that the opposing effects of competition on appraisals holds within individual appraisers. Our results are also robust to several different measures of appraiser competition. We also provide some support for both the appraisal catering and appraisal shopping hypotheses. 23

25 References Agarwal, S., Ambrose, B. W., and Yao, V. (2014). The limits of regulation: Appraisal bias in the mortgage market. Available at SSRN Agarwal, S., Ben-David, I., and Yao, V. (2015). Collateral valuation and borrower financial constraints: Evidence from the residential real estate market. Management Science. Bae, K.-H., Kang, J.-K., and Wang, J. (2013). Does increased competition affect credit ratings? A reexamination of the effect of fitch s market share on credit ratings in the corporate bond market. Journal of Financial and Quantitative Analysis (JFQA), Forthcoming. Becker, B. and Milbourn, T. (2011). How did increased competition affect credit ratings? Journal of Financial Economics, 101(3): Ben-David, I. (2011). Financial Constraints and Inflated Home Prices During the Real Estate Boom. American Economic Journal: Applied Economics, 3(2): Berger, A., Demirgüç-Kunt, A., Levine, R., and Haubrich, J. (2004). Bank Concentration and Competition: An Evolution in the Making. Journal of Money, Credit, and Banking, 36(3): Bikker, J. and Haaf, K. (2002). Measures of Competition and Concentration in the Banking Industry: A Review of the Literature. Economic & Financial Modeling, 9: Bolton, P., Freixas, X., and Shapiro, J. (2007). Conflicts of interest, information provision, and competition in the financial services industry. Journal of Financial Economics, 85(2): Calem, P. S., Lambie-Hanson, L., and Nakamura, L. I. (2015). Information Losses in Home Purchase Appraisals. Federal Reserve Bank of Philadelphia Working Paper No Camanho, N., Deb, P., and Liu, Z. (2009). Credit rating and competition. In 22nd Australasian Finance and Banking Conference. Chinloy, P., Cho, M., and Megbolugbe, I. F. (1997). Appraisals, transaction incentives, and smoothing. The Journal of Real Estate Finance and Economics, 14(1-2):

26 Cho, M. and Megbolugbe, I. F. (1996). An empirical analysis of property appraisal and mortgage redlining. The Journal of Real Estate Finance and Economics, 13(1): Clayton, J., Geltner, D., and Hamilton, S. W. (2001). Smoothing in commercial property valuations: Evidence from individual appraisals. Real Estate Economics, 29(3): Cohen, A. and Manuszak, M. D. (2013). Ratings competition in the cmbs market. Journal of Money, Credit and Banking, 45(s1): Ding, L. and Nakamura, L. I. (2015). The impact of the home valuation code of conduct on appraisal and mortgage outcomes. Forthcoming Real Estate Economics. Doherty, N. A., Kartasheva, A. V., and Phillips, R. D. (2012). Information effect of entry into credit ratings market: The case of insurers ratings. Journal of Financial Economics, 106(2): Flynn, S. and Ghent, A. (2015). Competition and credit ratings after the fall. Working Paper. Hansz, J. A. and Diaz III, J. (2001). Valuation bias in commercial appraisal: A transaction price feedback experiment. Real Estate Economics, 29(4): Hong, H. and Kacperczyk, M. (2010). Competition and bias. The Quarterly Journal of Economics, 125(4): LaCour-Little, M. and Malpezzi, S. (2003). Appraisal quality and residential mortgage default: Evidence from Alaska. The Journal of Real Estate Finance and Economics, 27(2): Lang, W. W. and Nakamura, L. I. (1993). A model of redlining. Journal of Urban Economics, 33(2): Mayer, C. J. and Pence, K. (2008). Subprime mortgages: what, where, and to whom? Technical report, National Bureau of Economic Research. Quan, D. C. and Quigley, J. M. (1991). Price formation and the appraisal function in real estate markets. The Journal of Real Estate Finance and Economics, 4(2): Shi, L. and Zhang, Y. J. (2015). Appraisal inflation: Evidence from 2009 GSE HVCC intervention. Journal of Housing Economics, 27:

27 Tzioumis, K. (2016). Appraisers and valuation bias: An empirical analysis. Real Estate Economics. Xia, H. (2014). Can investor-paid credit rating agencies improve the information quality of issuerpaid rating agencies? Journal of Financial Economics, 111(2):

28 Figure 1: This figure shows the distribution of the appraised value relative to the purchase price for loan applications in our sample. 27

29 Figure 2: Sample Appraisal Order Form. 28

30 Figure 3: Sample Sales Contract. 29