221 Week 3 A /For WIZARD KIM
z06jl
ContentServer.pdf
2016 V44 3: pp. 658–690
DOI: 10.1111/1540-6229.12120
REAL ESTATE ECONOMICS
The Impact of the Home Valuation Code of Conduct on Appraisal and Mortgage Outcomes Lei Ding* and Leonard Nakamura**
The accuracy of appraisals came into scrutiny during the housing crisis, and a set of policies and regulations was adopted to address the conflict-of-interest issues in the appraisal practices. In response to an investigation by the New York State Attorney General’s office, the Home Valuation Code of Conduct (HVCC) was agreed to by Fannie Mae, Freddie Mac and the Federal Housing Finance Agency. Using unique data sets that contain both approved and non- approved mortgage applications, this study provides an empirical examination of the impact of the HVCC on appraisal and mortgage outcomes. The results suggest that the HVCC has led to a reduction in the probability of inflated valuations, although valuations remained on average inflated, and induced a significant increase in the incidence of low appraisals. The well-intentioned HVCC rule made it more difficult to obtain mortgages to purchase homes dur- ing the housing price crash, possibly exacerbating the fall in prices.
Introduction
The fallout from the housing bubble raised questions about the accuracy of appraisals before the housing crisis, and, as a response, a set of policies and regulations was adopted to address the conflict-of-interest issues in the appraisal practices.1 With significantly tightened regulations and the decline in housing prices in many areas, there were concerns that more home val- uations were underestimated and new mortgages became harder to obtain
*Federal Reserve Bank of Philadelphia or [email protected]. **Federal Reserve Bank of Philadelphia or [email protected].
1Important regulations and rules related to appraisal include at least the Home Valu- ation Code of Conduct (HVCC); the Dodd-Frank Wall Street Reform and Consumer Protection Act; revised Interagency Appraisal and Evaluation Guidelines from the federal banking regulators issued in December 2010; and the government-sponsored enterprises’ new appraiser independence requirements that replaced the HVCC in October 2010 (U.S. Government Accountability Office, or GAO 2012).
C© 2015 American Real Estate and Urban Economics Association
The Impact of the Home Valuation Code of Conduct 659
during the crisis,2 though the upward bias in appraisals that had prevailed during the subprime boom has been reduced somewhat in many markets. Despite the controversial role of appraisers before and during the most re- cent housing crisis, there is a lack of empirical research about the pattern of appraisal outcomes and the effects of the interventions adopted since the crisis on appraisals and the housing market overall. This study provides the first empirical examination of the impact of a major appraisal rule, the now- superseded Home Valuation Code of Conduct (HVCC), which was adopted in the middle of the housing crisis, on low appraisals and mortgage outcomes. Appraisal ratio is defined as appraised value less the contract price as a per- cent of the contract price in this study, while low appraisal is defined as one in which appraised value falls below the contract price.3
The HVCC was enacted on May 1, 2009, as the result of a joint agreement between Fannie Mae and Freddie Mac (government-sponsored enterprises, or GSEs), the Federal Housing Finance Agency (FHFA),4 and the New York State Attorney General.5 The HVCC was set to expire in August 2010. The Dodd-Frank Wall Street Reform and Consumer Protection Act (Dodd-Frank Act), enacted on July 21, 2010, declared that the HVCC was no longer in effect, but it actually codified several of the HVCC’s provisions. The HVCC has several unique features. First, as a private agreement between the GSEs and the New York State Attorney General, the HVCC is an industry stan- dard instead of a federal regulation. In fact, the HVCC was implemented
2See the Reuters article http://www.reuters.com/article/2011/08/24/us-usa- economy-appraisals-idUSTRE77N2PM20110824 and the New York Times articles http://www.nytimes.com/2012/10/13/business/scrutiny-for-home-appraisers-as-the- market-struggles.html and http://www.nytimes.com/2013/09/15/realestate/when- appraisals-come-in-low.html. 3Similarly, significantly low appraisal is defined as one in which appraisal is at least 5% below the contract price. Share of low appraisals represents the share of appraisals with appraised values below the contract price. An appraisal is only an opinion of a property’s value so a deviation between appraised value and contract price does not necessarily mean the appraisal is wrong or biased. See similar measures of appraisal bias in Cho and Megbolugbe (1996), Chinloy, Cho and Megbolugbe (1997) and LaCour-Little and Green (1998). 4The agreement was initially known as the Home Value Protection Program and Co- operation Agreement. The Office of Federal Housing Enterprise Oversight (OFHEO) still existed as the independent regulatory agency of Fannie Mae and Freddie Mac when the HVCC was introduced in March 2008. In July 2008, the FHFA was formed by merging the OFHEO, the Federal Housing Finance Board, and the U.S. Department of Housing and Urban Development government-sponsored enterprise function. 5The HVCC, which was introduced on March 3, 2008, was a direct result of the Washington Mutual legal case. In November 2007, the New York Attorney General filed suit against Washington Mutual. Because GSEs purchased/securitized a large portion of its mortgages from Washington Mutual, the legal case pushed the GSEs to issue the HVCC.
660 Ding and Nakamura
despite opposition from major federal bank regulators (Abernethy and Hol- lans 2010). Second, while the HVCC initially only covered GSE loans, it had marketwide effects as a result of the oligopoly power of the GSEs and the lack of a robust alternative secondary market for residential mortgages.6
Third, the HVCC is believed to be a well-intentioned rule; however, some regulatory agencies and industry stakeholders have questioned it for its poten- tial jurisdictional problems and unintended consequences (U.S. Government Accountability Office, or GAO 2011). The rule introduced tighter scrutiny for appraisers, lenders, GSEs, and other stakeholders to ensure the independence of the appraisal process for GSE loans.7 However, as the HVCC’s efforts to address the conflict-of-interest issues in the middle of the crisis induced radi- cal changes of the entire appraisal industry, concerns arose about the possible decline in appraisal quality and increased difficulty in credit access (GAO 2011, 2012). For example, one direct effect of the HVCC was the greater use of appraisal management companies (AMCs).8 AMCs, which act as interme- diaries between lenders, only accounted for a small market share before the crisis and received little oversight by regulators during the crisis.9 So on the one hand, with less influence from lenders, brokers, and other stakeholders, appraisers are expected to achieve more objective appraisals and reduce the incidence of the previously widespread inflated appraisals. On the other hand, because of the greater use of AMCs partly induced by the HVCC and the potential overreaction by lenders and appraisers, the quality of appraisals may
6FHA adopted the HVCC on January 1, 2010, eight months later than the GSEs. GSEs accounted for about 69.4% of all mortgage originations in 2009; the GSEs and FHA together accounted for about 90% (Inside Mortgage Finance 2013). 7The HVCC was designed to enhance the independence and accuracy of the appraisal process primarily by the following: (1) prohibiting lenders and third parties with an interest in the mortgage transaction from influencing the development, reporting, result, or review of an appraisal report, (2) requiring only the lender or any third party specifically authorized by the lender to select, retain, and provide for payment of all compensation to the appraiser, (3) requiring absolute independence between the loan production function and the appraisal function within a lender’s organization, (4) limiting communications between loan pro- duction staff and appraisers, and (5) requiring lenders to ensure that borrowers receive a copy of the appraisal report within a certain period before closing. See http://www.fhfa.gov/Media/PublicAffairs/Documents/HVCCFinalCODE122308_N 508.pdf for more details about the HVCC. 8GAO (2012) suggests that some practitioners reported that the HVCC led some lenders to outsource appraisal functions to AMCs because they thought using AMCs would allow them to demonstrate compliance with these requirements easily. 9One major concern is that more appraisals are being done by AMC appraisers, who likely had received more business because of the HVCC. These appraisers may lack the knowledge of the local market because AMCs operate nationally but do not have appraisers in all local areas. In addition, AMC appraisers are usually paid less, which may induce them to invest less time and introduce more bias (GAO, 2012).
The Impact of the Home Valuation Code of Conduct 661
deteriorate and the share of low appraisals could become artificially high after the HVCC. This, in turn, could cause real estate deals to fall apart.
Using a unique transaction-level appraisal data set that contains both approved and nonapproved mortgage applications, this study examined the effect of the HVCC using a difference-in-differences approach. Because not all mortgage applications were subject to the HVCC, we can use statistical models to iso- late the effects of the HVCC by comparing changes in appraisal and mortgage outcomes pre- and post-HVCC for the HVCC-covered loans, relative to those of transactions that were not subject to the HVCC. We found that the HVCC has led to a significantly increased incidence of low appraisals and a reduc- tion in the probability of inflated valuations: The probability of low appraisals among HVCC-covered transactions was at least 2.1 percentage points higher than those transactions not covered by the HVCC while the share of signif- icantly high appraisals (5% or higher than contract prices) also decreased. The results are robust when different evaluation periods or different control groups are used.
A higher incidence of low appraisals also induces higher rates of mortgage denials. The overall denial rates in the purchase market started to decline after 2009; however, the decrease in denial rates, especially in the collateral denial rates, was significantly lower for the HVCC-covered applications. The probability of denials due to insufficient collateral increase by 1.2 percentage points post-HVCC, relative to the control group; the control group has a 5.6% probability of denial due to insufficient collateral.
The empirical results suggest that the HVCC has done some of what it was supposed to do by partially reducing inflated valuations that were more prevalent during the subprime boom. However, this well-intentioned rule also increases the likelihood of low appraisals and made the origination of purchase mortgages more difficult. Because access to mortgage credit has been tight since the housing crisis, more limited credit availability may have more severe consequences in the long term for certain populations and neighborhoods.
Background: Home Mortgage Appraisal and the HVCC
Appraisal and Appraisal Bias
Lending institutions compare the loan amount with the market value of the home in making loan decisions. Such a comparison is important because lenders need to know the property’s market value in order to provide informa- tion for assessing the risk of the mortgage and their potential loss exposure if the borrower defaults. Appraisals, which provide an estimate of market value
662 Ding and Nakamura
based on market research and analysis as of a specific date, have been the most commonly used valuation method for residential mortgage originations (GAO 2012).10 The appraised value and the difference between the appraisal and the contract price influence both the likelihood that the mortgage will default and the options that the mortgage lender has if the borrower defaults on the mortgage.11
In theory, an appraisal should provide an objective valuation of the true market value of a property; however, appraisals are often biased and can be significantly different from a home’s true market value. Recent studies of the accuracy of home mortgage appraisals in the United States started with an article by Cho and Megbolugbe (1996), who compared purchase prices with appraised values to determine whether there were systematic differences based on the 1993 Fannie Mae loan acquisition file. They found that appraisals may be biased since too many mortgage appraisals were exactly the same as the transaction price, and the distribution was highly asymmetric. More than 65% of appraised values were above the purchase prices; about 30% had appraisals that were exactly the same as transaction prices; and only 5% had appraisals that were lower than the transaction prices. Appraisers only assign different value estimates when differences between perceived values and transaction prices are substantial. In more than 80% of the cases, the appraisal was between 0% and 5% above the transaction purchase price. Chinloy, Cho and Megbolugbe (1997) expanded on the earlier research and continued to argue that appraisal bias was present. They estimated an upward bias of 2% and found that appraisals exceeded purchase prices in approximately 60% of the cases.
Agarwal, Ben-David and Yao (forthcoming) documented the appraisal bias for residential refinance transactions. They used the difference in the ini- tial appraisal of the refinance transaction and the subsequent purchase price compared with changes in the prices of pairs of consecutive purchase trans- actions, as a proxy for valuation bias. They found that the appraisal bias
10Methods other than appraisals, such as broker price opinions, automated valuation models, or other mixed methods usually take less time and are less expensive but are often less reliable. When performing appraisals, appraisers can use one or several approaches to determine value, including sales comparison, cost, and income. Of these, the sales comparison approach is most widely used, which compares and contrasts the property under appraisal with recent offerings and sales of similar properties. 11The precise value of the home on the market provides crucial information to the mort- gage lender because the equity stake of a mortgage at origination, usually measured by the loan-to-value (LTV) ratio, reflects the credit risk of a mortgage application. In practice, lenders usually use the lesser of sales price and appraisal as the value of the property in calculating LTV ratios (Nakamura 2010).
The Impact of the Home Valuation Code of Conduct 663
for residential refinance transactions was above 5% for a national sample of conforming loans. The bias was found to be larger for highly leveraged trans- actions (high loan-to-value or LTV), around critical leverage thresholds, and for transactions through a broker. However, in a study focusing on the behav- ior of appraisal professionals, Tzioumis (forthcoming) found only a minority of residential real estate appraisers systematically inflated appraisal values for home purchase loan applications.
When appraisals are biased upward, they provide documentation for loans larger than what the collateral’s market value justified. This makes mortgages riskier, and the risk of mortgage default increases. Unfortunately, there has been very little academic work examining the impact of biased appraisals despite the importance of the subject. LaCour-Little and Malpezzi (2003) used a small data set from Alaska in the 1980s to illustrate that for a single thrift institution in that state, appraisal bias was positively associated with more frequent defaults. Agarwal, Ben-David and Yao (forthcoming) also found that refinance mortgages with inflated appraisals default more often; however, lenders account for the appraisal bias through pricing by charging higher rates for mortgages that have higher appraisal bias.
Determining Factors of Appraisal Bias and the HVCC
The conflict-of-interest issues related to appraisals have been cited as a po- tential explanation for the upward bias in several empirical studies (e.g., Cho and Megbolugbe 1996; Chinloy, Cho and Megbolugbe 1997). Appraisers face asymmetric costs from overstating versus understating: While an above- contract price appraisal will have no direct impact, deals could be threatened by appraisals that fall below the prices that buyers and sellers had agreed to previously. Buyers, sellers and real estate agents, as well as lenders who do not bear the risk of originated loans, all have a vested interest in getting an ap- praisal that is not less than the contract price and completing the sale. The way to ensure the deal is for the appraisers to assess slightly higher than (or equal to) contract prices. Much anecdotal evidence suggests that such bias exists, such as the well-known legal case involving Washington Mutual in which the lender was found to put pressure on eAppraiseIT (an AMC) to generate sys- tematically high appraisals between July 2006 and April 2007.12 The HVCC, together with a set of other regulations and policies, was developed to govern
12According to Agarwal, Ben-David and Yao (forthcoming), Washington Mutual threatened to discontinue its contract with eAppraiseIT and actually did so in a number of cases. With the pressure from Washington Mutual, eAppraiseIT produced a list of “proven accepted” (by Washington Mutual) appraisers. In November 2007, the New York Attorney General filed a lawsuit against Washington Mutual, resulting in the HVCC.
664 Ding and Nakamura
the selection, communication, and possible coercion of appraisers in an effort to address the conflict-of-interest issues related to appraisal practices.
Cho and Megbolugbe (1996) found that appraisal outcomes are different for loans with different characteristics: Approved loans by Fannie Mae with low LTV ratios and/or high house prices are more likely to have negative appraisal gaps (low appraisals). They suspect that these loans are more likely to be approved despite negative appraisal gaps. LaCour-Little and Green (1998) conducted the only known empirical study that examines the role of appraisals in the residential mortgage lending process, though the study sample is quite small (fewer than 3,000 observations). They found that low appraised value is related to proxies for neighborhood quality instead of census tract racial composition. Properties securing adjustable rate mortgages, condominiums, and properties purchased by African American buyers are also found to have an increased probability of low appraisals.
Based on a theoretical model and empirical evidence, Calem, Lambie-Hanson and Nakamura (2014) demonstrated that the mortgage practice that requires the use of the lesser of the transaction price and the appraised value in the calculation of LTV ratios results in upward bias of appraisals, especially the extremely high incidence of appraisals which are exactly the same as or slightly higher than contract prices. They consider the proportion of appraisals that are set at the accepted offer price or very slightly above as “informa- tion loss,” since no precise information is conveyed by these appraisals. Except for this study, the only rigorous empirical study on the impact of HVCC was Agarwal, Ambrose and Yao (2014), which found the magnitude of the observed appraisal bias in the refinance market was reduced after HVCC.
It needs to be noted that the housing market was experiencing significant changes when the HVCC was first introduced. The lack of market sales, especially mortgage-financed sales, may lead to high degrees of uncertainty in appraisals (Lang and Nakamura 1993) and could lead to more mortgage denials (e.g., Blackburn and Vermilyea 2007). The sharp increase in distressed property sales, which could be recorded and used as comparables in the appraisals of nondistressed properties,13 may cause a downward drag on house value estimates.
13According to the Appraisal Institute (2008), an appraiser should not ignore foreclo- sure sales if the consideration of such sales is necessary to develop a credible value opinion. Only sales that might have involved atypical seller motivations (e.g., a highly motivated seller), such as a short sale, could be ignored.
The Impact of the Home Valuation Code of Conduct 665
Appraisal and Mortgage Lending Decisions
Appraised values and the difference between appraisals and contract prices have a direct effect on mortgage outcomes. A low appraisal may force a seller to sell the property at a price lower than the agreed-upon amount. If a seller is not willing to take a loss, the sale could be canceled. Second, low appraisals may cause lenders to seek larger down payments. Low appraised value may simply push the loan applicant to get a higher LTV loan. When the borrower is capital constrained, however, this may cause the lender to reject the loan application. So while an above-contract price appraisal will have no direct impact, a low appraisal may require buyers to come up with an extra down payment or pay a higher price (a higher interest rate or mortgage insurance that otherwise may not be needed), or it may result in a buyer withdrawing or a lender rejecting the application. LaCour-Little and Green (1998) confirmed that a low appraised value significantly increases the probability of mortgage loan application rejection. A low appraisal raises the likelihood of denial by 1.8 percentage points, while an appraisal that is the same as the offer price also raises the probability of denial by 0.6 percentage point.
This study is related to studies on lending disparities in the mortgage market, which tested the associations between neighborhood income, racial compo- nent, or center city location and mortgage lending (see review in Ladd 1998). Other studies, which are more relevant to this analysis, have investigated the impact of various government regulations on mortgage lending decisions. Such examples include studies on the impact of state antipredatory lending laws on mortgage lending (e.g., Harvey and Nigro 2004; Bostic et al. 2008) or on the impact of state foreclosure laws on mortgage lending (Pence 2006). This study contributes to the literature by providing new evidence of the im- pact of the HVCC, a major appraisal rule enacted in the housing crisis, on appraisal and mortgage outcomes.
Data
This analysis used two primary data sets. The first one is the FNC, Inc.’s collateral database (FNC data), which provides a national sample of appraisal records, regardless of whether they end up with mortgage originations. The FNC data have been built from the data aggregated from major mortgage lenders that agreed to share their nonconfidential appraisal data with FNC. The FNC data have information on property type, contract date, appraisal date, rounded sales price (rounded up to the next $50,000), appraisal-price percent difference, zip code and county code of the property. The second data set is the expanded Home Mortgage Disclosure Act (HMDA) data with information on mortgage application action dates (approval dates, denial dates
666 Ding and Nakamura
or other action dates) compiled by the staff of the Board of Governors of the Federal Reserve System. Compared with the publicly available HMDA data, this data set allows us to identify the timing of mortgage applications much more precisely.
The FNC data have some unique features compared with the data sets used in prior studies and can provide insights about the appraisal practices during the housing crisis. Prior studies using approved loans only suffer from a selec- tion bias: Appraisals in the approved samples are a conditional distribution- conditional on the loan being made. Since applications with appraised values lower than contract prices are more likely to be denied, the focus on the approved loans induces an underestimate in the incidence of low appraisals. Second, data sets with approved mortgages usually allow only for a compari- son of appraised values with transaction prices, instead of the initial contract prices, which are not always the same as the final transaction prices. If the seller has been forced to renegotiate the asking price when the appraised value of the property is below the contract price, the observed transaction price could actually be lower than the contract price. Of course, this data set has limitations, such as the sparse information on the borrower and mortgage characteristics and the underrepresentation in certain markets.14
Figure 1 based on the FNC data shows the change in the share of low appraisals over time. In 2006 and 2007, the share of low appraisals was between 4% and 6% nationally. After increasing slightly in 2008, the share of low appraisals started to increase sharply after the enactment of HVCC (from 8.3% in the fourth quarter of 2008 to 14% and 15.2% in the second and third quarters of 2009, respectively), with a peak in the third quarter of 2009. Of course, the decline in housing prices and increase in mortgage defaults during this period may also help explain the dynamics of low appraisal rates: Housing prices bottomed out in the first quarter of 2009 while the mortgage serious delinquency rate peaked in the fourth quarter of 2009 (Figure 1). Figure 2 further compares the distribution of the appraisal ratio pre- and post- HVCC. The share of low appraisals increased from 9.1% in the six months before the HVCC to 15.0% in the six months after the HVCC. The share of appraisals slightly higher than (or equal to) contract prices (0% to 1%) also increased slightly, while the share of significantly high appraisals decreased significantly, from 22.3% pre-HVCC to 14.6% post-HVCC. Overall, at the
14Due to privacy considerations, geographical information in FNC data is only avail- able at the zip code level. Information on individual borrowers, mortgage applications, property condition, and property address is generally unavailable. Some sand states including Arizona, California, Florida and Nevada are overrepresented, while the Midwest areas are slightly underrepresented (see Table 9).
The Impact of the Home Valuation Code of Conduct 667
Figure 1 � Share of low appraisals for the United States first quarter 2006 to third quarter 2012.
Note: Share of low appraisals represents the share of appraisals with appraised values below the contract price. Source: FNC data, LPS data, and CoreLogic HPI.
aggregate level, the distribution curve became more leptokurtic and shifted to the left after the HVCC: More appraisals came in below, equal to, or slightly higher than the contract prices, while there were fewer appraisals that were significantly higher than contract prices.
Descriptive Analysis: A Difference-in-Differences Approach
The changes in the appraisal ratio after the HVCC at the aggregate level do not necessarily reflect the independent effect of the HVCC on appraisal out- comes. As a source of plausibly exogenous variation, we exploit the fact that by regulation, only mortgages below the Conforming Loan Limits (CLL)15
15The national conforming loan limit for mortgages that finance single-family one- unit properties was $417,000 for 2006–2008, with higher limits for certain statu- torily designated high-cost areas and mortgages secured by multifamily dwellings. The Economic Stimulus Act (ESA) of 2008 temporarily raised the CLLs in des- ignated high-cost areas in the contiguous United States to up to $729,750. These higher temporary CLLs were then extended several times, finally expiring on September 30, 2011. Data for the CLLs at the county level are available at http://www.fhfa.gov/DataTools/Downloads/Pages/Conforming-Loan-Limits.aspx.
668 Ding and Nakamura
Figure 2 � Distribution of appraisal ratios pre- and post-HVCC.
Note: Appraisal ratio is defined as appraised value less contract price as a percent of contract price. Pre- and post-HVCC periods are defined here as the six months before and after the HVCC (October 1, 2008, to March 31, 2009, versus June 1, 2009, to November 30, 2009). All appraisals are included. Source: FNC data.
are eligible for GSE purchase and thus subject to the HVCC. By regulation, the CLL is a key determinant of whether a loan application is eligible to be purchased/securitized by GSEs and subject to the HVCC: Mortgages below the CLL are eligible to be purchased by the two GSEs, which either hold the mortgages or package them into securities and sell the securities to investors, while applications for mortgages above the CLL (jumbo loans) are ineligi- ble to be purchased by GSEs, and thus are not subject to the HVCC.16 So appraisals for loans under the CLLs can be roughly treated as the treatment group, while appraisals for loans above the CLLs can be considered as the control group.17 By comparing the changes in the appraisal and loan applica- tion outcomes pre- and post-HVCC between the treatment and control groups
16Jumbo mortgages are only a subset of the nonconforming market because loan characteristics other than size can also make a loan nonconforming. But these other underwriting criteria are not as clearly defined as the size limit. 17When the CLL changed during the study period in a number of areas, the highest level CLL was used to determine the control group, while the lowest level CLL was used to identify the treatment group. For example, if the CLL of one county was
The Impact of the Home Valuation Code of Conduct 669
(difference-in-differences), we attribute any such differences to the treatment itself, namely, the enactment of the HVCC, by factoring out the time and market trends during the study period.
The HVCC only applies to single-family mortgages sold to GSEs, so the ap- praisal analysis focuses on appraisals for relatively high-priced single-family properties during the study periods.18 The mortgage denial analysis primarily used applications for first-lien, single-family conventional loans19 with rel- atively large loan amounts (above 50% of the “conforming loan limit,” or CLL).20 Nonconventional loan applications that were initially not subject to the HVCC—primarily government-guaranteed loans such as Federal Housing Administration (FHA) and U.S. Department of Veteran Affairs (VA) loans —were excluded from the mortgage denial analysis.
The pre- and post-HVCC periods were classified according to appraisal dates or action dates on mortgage applications. Our preferred specification considers the six months before and after the HVCC (October 1, 2008, to March 31, 2009, versus June 1, 2009, to November 30, 2009) as the pre- and post-HVCC periods (Table 1). The months immediately before and after the enactment of the HVCC (April 1 to May 31, 2009) are considered to be a transitional period and are excluded from the analysis. This doughnut hole sampling technique is designed to remove observations from the data when appraisers and lenders might have been preparing for and adjusting to the new regulation, while still
increased from $417,000 in 2008 to $560,000 in 2009, we used over the higher limit of $560,000 to identify non-HVCC-covered appraisals (priced over $560,000*1.25 for our preferred model), while using the lower limit of $417,000 to identify HVCC- covered appraisals. 18As suggested by Liu, Nowak and Rosenthal (2014), not all housing submarkets experienced the same level of decline during the housing crisis and small-size starter houses were hit harder. To reduce the potential bias due to unobserved characteristics of different housing segments, appraisals for relatively low-priced properties (those with rounded contract prices below $300,000 to $500,000 depending on the area’s GSE loan limit) were excluded from the analysis. 19Only applications that were approved and originated, applications that were denied, and applications that were approved but not accepted by the applicant are included in the analyses. A similar treatment is available in Ding (2014). Eight possible outcomes for a loan application in HMDA include: (1) loan originated, (2) application approved but not accepted, (3) application denied by financial institution, (4) application with- drawn by applicant, (5) file closed for incompleteness, (6) loan purchased by financial institution, (7) preapproved request denied by financial institution and (8) preapproval request approved but not accepted. This study primarily focuses on the first three because it is hard to ascertain whether the application is approved or denied for other outcomes. 20Different thresholds, such as 75% of CLL, were tried and the results were quite consistent so this study focuses on this specification only.
670 Ding and Nakamura
Table 1 � Specifications of study periods and treatment/control groups.
Study Period
Pre-HVCC Post-HVCC Explanations
1/1/2009–3/31/2009 6/1/2009–8/30/2009 3 months before and after HVCC; 4/1/2009–5/31/2009 was considered as transitional period
10/1/2008–3/31/2009 6/1/2009–11/30/2009 6 months before and after HVCC; 4/1/2009–5/31/2009 was considered as transitional period (preferred)
4/1/2008–3/31/2009 6/1/2009–5/30/2010 12 months before and after HVCC; 4/1/2009–5/31/2009 was considered as transitional period
9/1/2007–2/28/2008 6/1/2009–11/30/2009 6 months before the introduction of HVCC and 6 months after the enactment of HVCC; 3/1/2008–5/31/2009 was considered as transitional period
Treatment/Control Group (appraisal analysis only)
Treatment Group Control Group
0.5*conforming price limit<price< conforming price limit (1.25*CLL)
price>conforming price limit (1.25*CLL)
appraisals with contract prices over 1.25 times of conforming loan limit is assumed to use nonconforming loans; they do not qualify for a conforming loan with a down payment up to 20% of the contract price (preferred)
0.5*conforming price limit<price< conforming price limit (1.25*CLL)
price>conforming price limit1 (1.43*CLL)
appraisals with contract prices over 1.43 times of conforming loan limit is assumed to use nonconforming loans; they do not qualify for a conforming loan with a down payment up to 30% of the contract price
0.5*conforming price limit<price< conforming price limit (1.25*CLL)
price>conforming price limit2 (1.67*CLL)
appraisals with contract prices over 1.67 times of conforming loan limit is assumed to use nonconforming loans; they do not qualify for a conforming loan with a down payment up to 40% of the contract price
Note: CLL represents conforming loan limit
The Impact of the Home Valuation Code of Conduct 671
allowing enough observations to provide some precision in the results. Most of our discussions of the empirical results are based on this specification.
Additional assumptions are needed to apply this identification strategy to the appraisal analysis. The FNC data used in this study provide informa- tion only on rounded contract prices, instead of the exact contract prices or loan amounts. Because information on the exact loan amount is unavailable, we used 1.25 times of the CLL, which we call the conforming price limit, (CPL) as a proxy to identify appraisals for conforming loans based on con- tract prices. Assuming a typical buyer needs to secure at least 20% for a down payment, transactions with contract prices below CPLs are eligible for conforming loans and thus subject to the HVCC. Transactions with contract prices above CPLs will either have a higher proportionate downpayment or will be ineligible for treatment as a conforming loan.21 As mentioned earlier, transactions with relatively low contract prices (less than 50% of CPL) were excluded to construct a more homogeneous group.
Table 2 examines the distribution of the appraisal ratio, including shares of low appraisals, significantly low appraisals, and significantly high appraisals, among the treatment and control group over the pre- and post-HVCC periods. The differences between periods and groups are provided. Table 3 provides similar data for denial rates and collateral denial rates, respectively, based on the HMDA data.
Appraisal Ratio
The descriptive statistics in Table 2 show that both the treatment group and the control group experienced a quantitatively significant increase in the share of low appraisals between the pre- and post-HVCC periods. The share of low appraisals increased 4.20% among the treatment group compared with a 2.73% increase in the control group for a difference of 1.47 percentage points. These represent, respectively, a roughly 38% relative increase in low appraisals for the treatment group compared with a 24% increase for the control group. The treatment group experienced a 1.72% increase in the share of significantly low appraisals compared with a 0.13% decrease in the control group. During the same period, the share of significantly high appraisals decreased 4.32% for the treatment group, 0.58 percentage point higher than the 3.74% decline in the control group.
21As high-priced homebuyers may have the incentive to find ways to raise cash to bridge the gap so they will not have to do a jumbo, different cut-off prices (CLL divided by 0.8, 0.7 or 0.6, respectively) are used to construct different control groups to test the sensitivity of the results to different cut-off points.
672 Ding and Nakamura
Table 2 � Appraisal ratios in the pre- and post-HVCC periods (%).
Appraisal Ratio = (Appraised Value-Contract Price)/Contract Price
Treatment Group Low Between Between (Priced between 0.5* Appraisal Below –5 and 0.1 and Above CPL and CPL) <0% –5% –0.1% 0% 5% 5%
Pre-HVCC 10.99 6.39 4.60 34.69 40.69 13.63 Post-HVCC 15.19 8.11 7.08 36.13 39.38 9.31 Change 4.20 1.72 2.48 1.44 –1.31 –4.32 Control Group (Priced above CPL)
Pre-HVCC 11.43 7.76 3.67 38.04 34.88 15.65 Post-HVCC 14.16 8.89 5.27 39.94 33.99 11.91 Change 2.73 1.13 1.60 1.90 –0.89 –3.74 Difference in Changes 1.47 0.59 0.88 –0.46 –0.42 –0.58
Note: CPL represents conforming price limit here, which is 1.25 times of conforming loan limit; the pre- and post-HVCC periods here are defined as six months before and after the HVCC; appraisals for single-family properties with contract prices above 50% of the confirming price limit only; based on FNC data.
Mortgage Denial
We further tracked the changes in both the overall denial rates and the rates of denials due to insufficient collateral for the treatment and the control group. The data in Table 3 show that applications in the control groups experienced a significant decline in denial rates, from 23.94% in the pre-HVCC period to 16.73% post-HVCC, or a decline of 7.21 percentage points. However, the decline was less for the treatment group, which was subject to the HVCC: The denial rate decreased 4.86% for the treatment group, or 2.35 percentage points lower than the 7.21% decline for the control group. A similar pattern can be found for the collateral denial rates: The collateral denial rate decreased by 0.51% for the treatment group after the HVCC, lower than the 1.48% decline for the control group, with a difference of 0.97 percentage point.
Overall, the descriptive analysis suggests, though not conclusively, that there was a general increase in the incidence of low appraisals in the treatment group post-HVCC. While the overall denial rates declined after the HVCC, the decline in the treatment group, which was likely to be subject to the HVCC, was much less. So relative to the control group, the HVCC-covered applications were more likely to be denied, especially in the case of insuffi- cient collateral after the HVCC was enacted. Of course, other market forces that might affect appraisal and mortgage lending outcomes, such as property
The Impact of the Home Valuation Code of Conduct 673
Table 3 � Denial rates in the pre- and post-HVCC periods (%).
Denial Rate Collateral Denial Rate
Treatment Group(Loan amounts between 0.5*CLL and CLL)
Pre-HVCC 16.07 3.19 Post-HVCC 11.21 2.68 Change −4.86 −0.51 Control Group (Loan amounts above CLL)
Pre-HVCC 23.94 5.57 Post-HVCC 16.73 4.09 Change −7.21 −1.48 Difference in Changes 2.35 0.97
Note: The pre- and post-HVCC periods here are defined as six months before and after the HVCC. First-lien, one to four family properties, applications with loan amounts above 50% of the confirming loan limit only. Only applications that were approved and originated, applications that were denied, and applications that were approved but not accepted by the applicant are included in the calculation of denial rates; based on the HMDA data.
type, lender type, and neighborhood risk characteristics, need to be taken into consideration in order to arrive at more solid conclusions.
Empirical Method
This section lays out several tests that examine the robustness of the results from the descriptive analysis. Specifically, the remainder of this section out- lines two separate tests that examine the effect of the implementation of the HVCC on appraisal ratio and mortgage denials.
Probability of Low (High) Appraisals
We used a standard difference-in-differences model, which compares the change in the probability of low or high appraisals pre- and post-HVCC among appraisals likely to be subject to the HVCC, relative to the control group. We estimate linear probability models, where the dependent variable is equal to 100% if the appraisal is below contract price and zero otherwise. So the unit of the estimated coefficients is percentage point (of the probability of low appraisals).
Pr(Low Appraisali) = β0 + β1 ∗ HVCCi + β2 ∗ BCPLi +β3 ∗ (HVCCi ∗ BCPLi ) + γ ∗ CBSAi +δ ∗ LENDERi + ε ∗ X i (1)
674 Ding and Nakamura
where
� HVCC is the time dummy to identify the pre- and post-HVCC time periods.
� BCPL is equal to 1 if the contract price is below the conforming price limit and equal to 0 if the contract price is above the limit.
� HVCC� BCPL is the interaction that combines the HVCC rule and conforming price limit variables, which is used to assess the impact of the HVCC on the incidence of low appraisals.
� CBSA is the metro dummy (major metros only). We restrict metro areas to those with over 5,000 appraisals in the appraisal analysis, and to 10,000 applications in the decision analysis.
� LENDER is the lender dummy (major lenders only). The reference group comprises all AMCs. AMCs are a means for satisfying the requirements of the HVCC and thus are a useful base group.
� X represents the control variables that include major lender dum- mies and a few neighborhood characteristics such as prior mortgage originations, recent house price change,22 prior foreclosure rate, and neighborhood size. The variable definitions are provided in Table 4.
Here appraisals with prices below the CPL became subject to the HVCC after the rule was enacted (HVCC*BCPL is equal to 1). So the coefficient of interest, β3, represents the HVCC effect on the incidence of low appraisal by capturing the change in the probability of a low appraisal of covered applications after the HVCC.
Denial Probabilities
A similar difference-in-differences model was used to compare the change in the incidence of denial before and after the HVCC among conforming loans with that of loans that are likely not to be subject to the HVCC. We estimate linear probability models, where the dependent variable is equal to 100% if the application is denied (or denied due to insufficient collateral) and zero otherwise. So the unit of the estimated coefficients is percentage point (of the probability of denial or collateral denial).
Pr(Deniali ) = β0 + β1 ∗ HVCCi + β2 ∗ BCLLi + β3 ∗ (HVCCi ∗ BCLLi ) +γ ∗ CBSAi + δ ∗ LENDERi + ε ∗ X i (2)
22We used the county-level CoreLogic house price index by segment (jumbo and conforming market) to construct the measure of house price change (change in the 12 months prior to the appraisal date).
The Impact of the Home Valuation Code of Conduct 675
Table 4 � Variable definitions.
Appraisal Model
Parameter Definition
low appraisal Indicator variable = 1 if appraisal is lower than contract price
significantly low appraisal Indicator variable = 1 if appraisal is at least 5% lower than contract price
significantly high appraisal Indicator variable = 1 if appraisal is at least 5% higher than contract price
prior loan origination (in log) Number of purchase loan originations in the zip code in the previous year (in log)
prior foreclosure rate Share of loans in serious delinquency in the zip code in the previous year (from LPS)
house price change in 12 months County-level house price change in the previous 12 months (from CoreLogic)
housing units (in log) Number of owner-occupied housing units in the zip code from 2000 census (in log)
hvcc Indicator variable: Time period post-HVCC = 1; 0 otherwise
bcpl Indicator variable: Contract price below CPL = 1; 0 otherwise
hvcc*bcpl Indicator variable: Post-HVCC period for appraisals with contract price below CPL = 1; 0 otherwise
lender dummies Dummies for major non-AMC lenders (Lender 1- Lender 4, and other small lenders)
MSA dummies Dummies for major MSAs
Mortgage Denial Model
Parameter Definition
denial Indicator variable = 1 if denied; 0 otherwise collateral denial Indicator variable = 1 if denied due to
insufficient collateral; 0 otherwise other minority (vs. white) Indicator variable = 1 if other non-African
American non-Hispanic; 0 otherwise African American (vs. white) Indicator variable = 1 for African American
borrower; 0 otherwise Hispanic (vs. white) Indicator variable = 1 for Hispanic borrower; 0
otherwise Owner-occupied property Indicator variable = 1 for owner-occupied
property; 0 otherwise Borrower income (relative to
area median income, %) Applicant household income relative to area
median income Minority tract (10–30%
minority) Indicator variable = 1 for tracts with 10–30%
minority; 0 otherwise High minority tract (�30%
minority) <10%) Indicator variable = 1 for tracts with 30% or
more minority; 0 otherwise Low income tract Indicator variable = 1 for low income tract
(<50% AMI); 0 otherwise
676 Ding and Nakamura
Table 4 � Continued.
Mortgage Denial Model
Parameter Definition
Moderate income tract Indicator variable = 1 for moderate income tract (50-79% AMI); 0 otherwise
Middle income tract Indicator variable = 1 for middle income tract (80-119% AMI); 0 otherwise
Tract population (in log) tract total population (in log) based on 2000 census
0–10 purchase loans in tract Indicator variable = 1 for tracts with 10 or less purchase loans in the previous year; 0 otherwise
11–20 purchase loans in tract (vs.>80)
Indicator variable = 1 for tracts with 11–20 purchase loans in the previous year; 0 otherwise
21–40 purchase loans in tract (vs.>80)
Indicator variable = 1 for tracts with 21–40 purchase loans in the previous year; 0 otherwise
41–80 purchase loans in tract (vs.>80)
Indicator variable = 1 for tracts with 41–80 purchase loans in the previous year; 0 otherwise
Foreclosure rate 3–7% Indicator variable = 1 for tracts with estimated foreclosure rate of 3.1-7% in 2007–2008; 0 otherwise
Foreclosure rate 7–11% Indicator variable = 1 for tracts with estimated foreclosure rate of 7.1-11% in 2007–2008; 0 otherwise
Foreclosure rate >11% Indicator variable = 1 for tracts with estimated foreclosure rate >11% in 2007–2008; 0 otherwise
Tract denial rate in 2008 Tract purchase loan denial rate in the previous year
HVCC Indicator variable: Time period post-HVCC = 1; 0 otherwise
BCLL Indicator variable: Loan amount below CLL = 1; 0 otherwise
HVCC*BCLL Indicator variable: Post-HVCC period for applications with loan amount below CLL = 1; 0 otherwise
Lender dummies Dummies for major non-AMC lenders MSA dummies Dummies for major MSAs
where
� BCLL is equal to 1 if the loan amount is below the “conforming loan limit” and to 0 if the loan amount is above the CLL.
The Impact of the Home Valuation Code of Conduct 677
� HVCC� BCLL is the two-way interaction of the time and below conforming loan limit variables, which is used to assess the impact of the HVCC on denial probabilities.
� X represents the control variables that may influence mortgage appli- cation decisions including borrower demographic information, neigh- borhood risk characteristics, as well as metropolitan and lender dummies (see Table 4).
Potential Concerns
To implement the identification strategy effectively, the control group needs to have a stable legal environment during the study period. But an unknown share of appraisals in the control group that was assumed to be unaffected by the HVCC was actually subject to the HVCC during the study period. We recognize that the lack of information on the exact loan amounts has serious limitations, which makes it impossible for us to determine the treat- ment/control group precisely. This measurement issue may bias the results. However, our identification strategy to isolate the effect of the HVCC should still be valid for the following two reasons. First, the fact that a significant share of transactions in the control group were actually subject to HVCC would only lead to an underestimate the HVCC effect.23 As long as the share of transactions in the control group that were subject to HVCC remained unchanged or was increasing over the study period, the identification strategy to isolate the effect of HVCC should not overestimate the HVCC effect. In fact, during the study period, the use of jumbo loans had been actually quite stable. Our estimation based on the Lender Processing Services (LPS) data suggests that about 23.7% of mortgage-financed purchases of higher-priced properties (with purchase prices above 1.25 times of CLLs) used jumbo loans during the 12 months before HVCC, compared with 22.8% in the 12 months after HVCC.24
Another reason why homes with prices above the conforming price limit could be a useful comparison group to measure the impact of the HVCC is that such homes are likely to have less appraisal bias. The reason goes to the heart of why appraisals are biased in the first place as discussed in Calem,
23Since only a proportion of appraisals in the control group was immune from HVCC, the results should be interpreted in instrumental variables sense (Fuster and Vickery 2014; Kaufman 2014) and an estimate that eliminates this source of bias should be bigger. 24For transactions with purchase prices higher than 1.67 times of CLL, the ratio of using jumbo loans were 42.2% and 43.9% 12 months pre- and post-HVCC. The estimation is based on eligible first-lien conventional purchase loan originations in LPS.
678 Ding and Nakamura
Lambie-Hanson and Nakamura (2014). In that analysis, appraisal bias arises because an appraisal below the transaction price reduces the price upon which the loan’s LTV is calculated according to GSE rules, and the LTV is forced higher, endangering the loan. But in cases where the LTV is substantially lower than 80%, this concern loses much of its force.
Consider the following stylized case. Suppose the conforming loan limit is $300,000, the conforming price limit is $375,000, and the home price in question is $400,000. Either the borrower seeks a jumbo loan of $320,000 (with an 80% LTV) or seeks a conforming loan of $300,000, with an LTV of 75%. In the former case, the HVCC does not apply; in the latter case an appraisal of $375,000 still leaves an LTV of 80% and does not threaten the loan. As Calem, Lambie-Hanson and Nakamura (2014) point out, the bias of appraisals is primarily due to a concern that the loan is threatened, and thus in these cases there is less likely to be bias. Cho and Megbolugbe (1996) present evidence that as LTVs fall, appraisal bias diminishes.
Thus the conforming price limit may be a valid control group either because the HVCC does not apply or because the LTV is not a binding constraint, and so bias is likely to be lower to begin with. In this latter case, the HVCC may be affecting the appraisal, but to a lesser degree. This represents a case in which the control group may be artificially biasing downward the impact of the HVCC, and so our estimates are lower bounds of the HVCC impact.
Second, the mortgage denial model focuses on both the probability of denials in general as well as the denials due to insufficient collateral, which are based on the denial reasons reported by the applicants. The concern is that not all lenders report the reasons for their denied applications, so the results on collateral denials may be biased due to missing values for some denials.25
However, the results are still informative because denial reasons are available for a vast majority of denied applications: About 81.2% of the first-lien conventional purchase loan applicants have the first reason reported, and 16.6% and 4.4% of the applicants reported their second and third reasons in 2009.
Multivariate Analysis Results
This section presents multivariate results for each of the models outlined in the previous section. The multivariate results in Table 5 through Table 8 strongly
25Recording reasons for denial was not obligatory for financial institutions supervised by the Federal Reserve System (FRS) and the Federal Deposit Insurance Corporation (FDIC), which usually have lower denial rates. And reporting institutions may cite up to three reasons for each denied application, although most of those that provide this information cite only one.
The Impact of the Home Valuation Code of Conduct 679
T ab
le 5
� Im
pa ct
of ho
m e
va lu
at io
n co
de of
co nd
uc t
on th
e pr
ob ab
il it
y of
lo w
ap pr
ai sa
ls .
3 M
on th
s P
re -
6 M
on th
s P
re -
12 M
on th
s P
re -
6 M
on th
s P
re -
6 M
on th
s P
re -
an d
P os
t an
d P
os t-
H V
C C
an d
P os
t- H
V C
C an
d P
os t-
H V
C C
R el
ea se
an d
P os
t- H
V C
C -H
V C
C in
S an
d S
ta te
s
P ar
am et
er C
oe f.
S td
. E
rr .
C oe
f. S
td .
E rr
. C
oe f.
S td
. E
rr .
C oe
f. S
td .
E rr
. C
oe f.
S td
. E
rr .
H V
C C
2. 81
0∗ ∗
0. 91
7 2.
75 7∗
∗∗ 0.
77 6
2. 28
4∗ ∗∗
0. 59
2 3.
23 6∗
∗ 1.
04 9
5. 34
5∗ ∗
1. 56
0 B
el ow
co nf
or m
in g
pr ic
e li
m it
(B C
P L
) −2
.7 23
∗∗ 0.
83 3
−2 .7
55 ∗∗
∗ 0.
51 4
−3 .1
88 ∗∗
∗ 0.
47 3
−4 .1
83 ∗∗
∗ 0.
70 3
−5 .0
18 ∗∗
1. 10
1
H V
C C
∗ B C
P L
3. 00
3∗ ∗∗
0. 86
1 2.
14 2∗
∗ 0.
66 9
1. 89
8∗ ∗
0. 65
6 3.
60 6∗
∗∗ 0.
63 1
4. 66
1∗ ∗∗
1. 23
2 C
o n tr
o ls
pr io
r lo
an or
ig in
at io
n (i
n lo
g) −1
.6 01
∗∗ ∗
0. 39
1 −1
.3 63
∗∗ ∗
0. 29
6 −1
.1 38
∗∗ ∗
0. 24
5 −1
.1 90
∗∗ ∗
0. 29
0 −0
.7 26
0. 56
4
P ri
or fo
re cl
os ur
e ra
te 2.
01 1∗
∗∗ 0.
22 1
2. 14
6∗ ∗∗
0. 16
6 1.
71 8∗
∗∗ 0.
16 6
2. 15
5∗ ∗∗
0. 22
1 1.
87 6∗
∗∗ 0.
21 4
H ou
se pr
ic e
ch an
ge in
pr ev
io us
12 m
on th
s −0
.1 51
∗∗ 0.
05 8
−0 .1
40 ∗∗
0. 04
4 −0
.1 66
∗∗ ∗
0. 04
2 −0
.1 34
∗∗ 0.
05 1
−0 .3
48 ∗∗
∗ 0.
07 5
H ou
si ng
un it
s (i
n lo
g) 1.
04 1∗
0. 50
9 0.
80 0∗
0. 40
0 0.
85 4∗
0. 33
2 0.
83 5∗
0. 40
1 0.
29 9
0. 80
4 L
en de
r 1
(v s.
A M
C s)
−3 .4
98 ∗
1. 45
2 −3
.1 18
∗∗ 1.
08 4
−2 .0
41 ∗∗
0. 63
7 −2
.8 79
∗∗ 0.
95 2
−1 .9
54 1.
29 2
L en
de r
2 (v
s. A
M C
s) −3
.6 57
∗ 1.
46 6
−3 .3
41 ∗∗
1. 17
2 −2
.1 84
∗∗ 0.
66 2
−1 .9
42 ∗
0. 97
8 −2
.3 98
∗ 1.
12 9
L en
de r
3 (v
s. A
M C
s) −1
.2 89
5. 76
3 −7
.4 24
∗∗ ∗
1. 59
3 −4
.9 86
∗∗ ∗
0. 77
2 −6
.8 66
∗∗ ∗
1. 38
4 −6
.9 58
4. 07
9 L
en de
r 4
(v s.
A M
C s)
−6 .8
91 ∗∗
∗ 1.
75 8
−5 .9
88 ∗∗
∗ 1.
36 7
−5 .2
08 ∗∗
∗ 0.
90 2
−5 .4
35 ∗∗
∗ 1.
40 5
−3 .6
99 1.
98 5
O th
er le
nd er
s −4
.0 76
∗ 1.
78 1
−2 .5
46 1.
38 2
−4 .3
69 ∗∗
∗ 0.
76 7
−5 .2
39 ∗∗
∗ 1.
06 2
−0 .7
31 1.
75 1
M S
A D
um m
y (C
B S
A )
ye s
Y es
Y es
Y es
Y es
M od
el fi
t A
d ju
st ed
R 2
0. 04
78 0.
04 83
0. 04
29 0.
05 68
0. 05
23 N
u m
b er
o f
o b se
rv a ti
o n s
42 ,1
93 80
,9 97
16 7,
46 2
82 ,3
97 34
,9 00
∗∗ ∗ S
ig ni
fi ca
nt at
0. 00
1 le
ve l;
∗∗ S
ig ni
fi ca
nt at
0. 01
le ve
l; ∗ S
ig ni
fi ca
nt at
0. 05
le ve
l; er
ro rs
ar e
cl us
te re
d at
M S
A le
ve l.
N o
te :
S ee
T ab
le 1
fo r
de fi
ni ti
on s
of pr
e- an
d po
st -H
V C
C pe
ri od
s. B
as ed
on F
N C
da ta
; si
ng le
-f am
il y
pr op
er ti
es w
it h
co nt
ra ct
pr ic
es ab
ov e
50 %
of th
e co
nfi rm
in g
pr ic
e li
m it
on ly
.
680 Ding and Nakamura
support the descriptive analysis results that the HVCC led to an increased incidence of low appraisals and mortgage denials. The R2 and goodness-of- fit are generally low, which is not surprising for linear probability models. The discussion of the empirical results primarily focuses on the effects of the HVCC, while results of some controls are discussed briefly at the end of this section. Again, the discussion is based primarily on our preferred specification, which focuses on the periods six months before and six months after the HVCC.26
Effects of the HVCC on Low (High) Appraisals
Appraisals generally have a higher probability of low appraisals after May 2009 (about 2.8 percentage points higher in six months; see the HVCC coefficient in Table 5) that is likely due to the general market trend influenced by factors other than the HVCC. Appraisals in the treatment group, which generally have lower contract prices, are generally less likely to have low appraisals (about 2.8 percentage points lower; see the BCPL coefficient) than those in the control group, which is consistent with the expectation the share of low appraisals increases with the property price. Appraisers may have greater difficulty in finding comparables for high-priced properties in the control group, and they may be more conservative for these properties as well given the potentially greater loss for the lenders and investors.
However, the HVCC leads to a significant increase in the likelihood of low appraisals in the treatment group, relative to the control group: The probability of low appraisals for the treatment group increases by 3.0 percentage points in three months, 2.1 percentage points in six months, and 1.9 percentage points in 12 months post-HVCC (the HVCC*BCPL coefficients).27 Put this in context, the share of low appraisals was 11.4% for the control group and 11.0% for the treatment group before HVCC. So the increase in the share of low appraisals was quite significant. The magnitude of the HVCC effect diminishes slightly over time likely because lenders and regulators have gradually adjusted to the new system over time. However, the HVCC effect on low appraisals is even greater when compared with the period before the HVCC proposal was first announced in March 2008: The probability of low
26As Table 1 summarizes, this study tried several different definitions of the pre- and post-HVCC periods for two considerations: 1) whether there are significant differences between the short-term and the long-term effects of the HVCC, and 2) whether the results are sensitive to different treatments of the period from the time when it was first announced to its enactment date (March 3, 2008 to April 30, 2009). 27The effects of the HVCC on low appraisals could be underestimated here since the low-end housing segment, which was hit harder by the housing crisis and more vulnerable to tightened regulations like the HVCC, was not included in the analysis.
The Impact of the Home Valuation Code of Conduct 681
appraisals increases by 3.6 percentage points compared with the pre-release period. The results suggest lenders and appraisers may have started to adjust their practices during the transitional period. We also found that the HVCC has a more pronounced effect in the hardest hit states (California, Arizona, Florida and Nevada): an increase of 4.7 percentage points in the share of low appraisals in the six months after HVCC, higher than the 2.1% increase at the aggregate level.28 The results suggest significant spatial heterogeneity of the HVCC effect.
The HVCC also leads to more significantly low appraisals as well (Table 6): For the treatment group, the probability of significantly low appraisals increases by 1.2 percentage points post-HVCC, while the effect of HVCC on the probability of significantly high appraisals decreases by 1.2% (significant at 0.1 level).29
Because this analysis identified the treatment group based on observed con- tract prices instead of exact loan amounts, the results could be sensitive to the price cut-off points, especially considering high-priced homebuyers may have the incentive to use cash to bridge the gap to do a conforming loan. As a robustness check, a set of different thresholds was used to test the sensitivity of the results to price cut-off points. The results listed in Table 7 suggest that the HVCC effect becomes greater when higher price limits are used: The probability of low appraisals increases by 2.7 percentage points and 3.3 percentage points if the CPL is set as 1.43 times of CLL and 1.67 times of CLL, respectively, both of which are higher than the 2.1 percentage points increase when CPL is equal to 1.25 times of CLL.30 The larger gap between the treatment and control groups is consistent with our contention that the HVCC leads to more low appraisals because appraisals of higher contract prices were increasingly less likely to use conforming loans that were subject to the HVCC.
Effect of the HVCC on Mortgage Denials
The results discussed in the previous subsection confirm that the HVCC increases the incidence of low appraisals. So what would be the effect of the HVCC on mortgage lending decisions? Regression results suggest that
28The magnitude of the HVCC effect is smaller in nonsand states (0.93 for the preferred model; significant at the 0.1 level). 29The lower level of statistical significance of the shift in high appraisals is largely due to the smaller sample size. 30The price limit to identify the control group has been kept the same here (1.25 times of CLL).
682 Ding and Nakamura
Table 6 � Impact of the HVCC on probability of significantly low (high) appraisals.
Significantly Low Significantly High Appraisal Appraisal (5% below Prices) (5% above Prices)
Parameter Coef. Std. Err. Coef. Std. Err.
HVCC 1.377∗ 0.599 −2.884∗ ∗ ∗ 0.593 Below conforming price
limit (bcpl) −3.213∗ ∗ ∗ 0.464 −1.024 0.756
hvcc∗bcpl 1.167∗ 0.532 −1.231 0.671 controls Prior loan origination (in
log) −1.489∗ ∗ ∗ 0.307 −0.913∗ ∗ ∗ 0.260
Prior foreclosure rate 1.758∗ ∗ ∗ 0.113 −0.215∗ 0.101 House price change in
previous 12 months −0.142∗ ∗ ∗ 0.036 0.015 0.036
Housing units (in log) 0.633 0.349 −0.215 0.313 Lender 1 (vs. AMCs) −2.357∗∗ 0.898 −1.057 0.679 Lender 2 (vs. AMCs) −1.893 0.991 −0.836 0.644 Lender 3 (vs. AMCs) −4.645∗ ∗ ∗ 1.154 1.754 0.917 Lender 4 (vs. AMCs) −4.783∗ ∗ ∗ 1.039 3.439∗ 1.637 Other lenders −1.546 1.120 4.048∗ 1.691 MSA Dummy Yes Yes Model fit Adjusted R2 0.0498 0.0395 Number of observations 80,997 80,997
∗∗∗Significant at .001 level; ∗∗Significant at .01 level; ∗Significant at 0.05 level; errors are clustered at MSA level. hvcc*bcpl is significant at 0.1 level for the significantly high appraisal model. Note: Single-family properties with contract prices above 50% of the confirming price limit only; treatment group includes appraisals with contract prices below 1.25 times of conforming loan limits (maximum price to get a conventional loan with 80% LTV).
the overall probability of collateral denial decreases (by about 1.3 percentage points) after the HVCC (Table 8). However, there are significant variations between the declines for the treatment group and the control group; the decline in the control group is much sharper. The probability of mortgage denials due to insufficient collateral is about 1.2 percentage points higher in the treatment group post-HVCC.31 Considering the collateral denial rate was only 5.6 for the control group and 3.2 for the treatment group, an increase of 1.2% is quite significant for the treatment group. The probability of any denial is also 3.3
31Denial for collateral may also be due to technical issues with the collateral but we believe such measurement errors to be small and orthogonal to the HVCC.
The Impact of the Home Valuation Code of Conduct 683
Table 7 � Sensitivity analysis of the cut-off points of contract prices.
Price Cut-Off Points (Control Group), Six Months Pre- and Post-HVCC
1.25∗CLL 1.43∗CLL 1.67∗CLL Variables (loan limit/80%) (loan limit/70%) (loan limit/60%)
Parameter Coef. Std. Err. Coef. Std. Err. Coef. Std. Err.
HVCC 2.757∗ ∗ ∗ 0.776 2.183∗∗ 0.770 1.603* 0.818 Below conforming price
limit (bcpl) −2.755∗ ∗ ∗ 0.514 −3.071∗ ∗ ∗ 0.578 −3.896∗ ∗ ∗ 0.693
HVCC*BCPL 2.142∗∗ 0.669 2.744∗ ∗ ∗ 0.640 3.315∗ ∗ ∗ 0.825 Controls Prior loan origination (in
log) −1.363∗ ∗ ∗ 0.296 −1.285∗ ∗ ∗ 0.309 −1.362∗ ∗ ∗ 0.314
Prior foreclosure rate 2.146∗ ∗ ∗ 0.166 2.087∗ ∗ ∗ 0.165 2.100∗ ∗ ∗ 0.156 House price change in
previous 12 months −0.140∗∗ 0.044 −0.141∗ ∗ ∗ 0.042 −0.142∗∗ 0.046
Housing units (in log) 0.800* 0.400 0.775 0.424 0.869* 0.426 Lender 1 (vs. AMCs) −3.118∗∗ 1.084 −3.417∗∗ 1.080 −3.301∗∗ 1.082 Lender 2 (vs. AMCs) −3.341∗∗ 1.172 −3.696∗∗ 1.152 −3.584∗∗ 1.157 Lender 3 (vs. AMCs) −7.424∗ ∗ ∗ 1.593 −8.047∗ ∗ ∗ 1.572 −7.852∗ ∗ ∗ 1.597 lender 4 (vs. AMCs) −5.988∗ ∗ ∗ 1.367 −6.349∗ ∗ ∗ 1.400 −6.175∗ ∗ ∗ 1.412 Other lenders −2.546 1.382 −2.719 1.448 −2.143 1.512 MSA Dummy Yes Yes Yes Model fit Adjusted R2 0.0483 0.0473 0.0474 Number of observations 80,997 77,076 72,819
∗∗∗Significant at .001 level. ∗∗Significant at .01 level. ∗Significant at 0.05 level; errors are clustered at MSA level. Note: See Table 1 for definitions of different price cut-off points. Single-family prop- erties with contract prices above 50% of the confirming price limit only.
percentage points higher in the treatment group than the control group after HVCC. The HVCC is explicitly designed to address the appraisal issues, so it is expected to have a significant impact on collateral denials and the overall denial rate as well. Of course, note that the measure of collateral denial is imperfect due to missing data on denial reasons for a small share of denied mortgage applications.
For different study periods (three months or 12 months), the results are quite consistent: a 1.2 percentage points increase in the probability of collateral denials in three months and a smaller increase of 0.5 percentage point increase in 12 months, respectively. The pattern for the HVCC effect on overall denial rates is quite similar.
684 Ding and Nakamura
T ab
le 8
� Im
pa ct
of th
e ho
m e
va lu
at io
n co
de of
co nd
uc t
on th
e pr
ob ab
il it
y of
pu rc
ha se
m or
tg ag
e de
ni al
s.
3 M
on th
s P
re -
an d
P os
t- H
V C
C 6
M on
th s
P re
- an
d P
os t-
H V
C C
12 M
on th
s P
re -
an d
P os
t- H
V C
C
C ol
la te
ra l
D en
ia l
D en
ia l
C ol
la te
ra l
D en
ia l
D en
ia l
C ol
la te
ra l
D en
ia l
D en
ia l
P ar
am et
er C
oe f.
S td
. E
rr .C
oe f.
S td
. E
rr .C
oe f.
S td
. E
rr .C
oe f.
S td
. E
rr .C
oe f.
S td
. E
rr .C
oe f.
S td
. E
rr .
hv cc
−1 .4
64 ∗∗
∗ 0 .3
51 −5
.8 48
∗∗ ∗
0. 70
1 −1
.2 60
∗∗ ∗ 0
.1 91
−5 .6
15 ∗∗
∗ 0.
50 8
0. 67
8∗ ∗
0. 21
2 2.
63 3∗
∗∗ 0.
47 3
be lo
w co
nf or
m in
g lo
an li
m it
(b cl
l) −2
.5 40
∗∗ ∗ 0
.4 09
−1 0.
05 3∗
∗∗ 0.
71 0
−2 .6
00 ∗∗
∗ 0 .2
76 −1
0. 03
3∗ ∗∗
0. 54
6 −2
.1 49
∗∗ ∗ 0
.2 04
−8 .9
18 ∗∗
∗ 0 .4
74 hv
cc *b
cl l
1. 16
6∗ ∗
0. 36
5 3.
44 1∗
∗∗ 0.
70 4
1. 15
1∗ ∗∗
0. 19
9 3.
25 4∗
∗∗ 0.
44 5
0. 50
6∗ ∗
0. 20
5 0.
91 0∗
0. 41
1 co
n tr
o ls
ot he
r m
in or
it y
(v s.
W hi
te )
0. 16
5 0.
10 8
2. 60
3∗ ∗∗
0. 28
8 0.
02 3
0. 07
6 2.
50 1∗
∗∗ 0.
28 5
0. 06
1 0.
06 6
2. 91
1∗ ∗∗
0. 28
1 A
fr ic
an A
m er
ic an
(v s.
W hi
te )
1. 35
4∗ ∗∗
0. 32
5 12
.9 95
∗∗ ∗
0. 79
9 1.
03 8∗
∗∗ 0.
25 5
11 .9
74 ∗∗
∗ 0.
60 0
1. 26
9∗ ∗∗
0. 16
0 13
.7 80
∗∗ ∗
0. 64
9 H
is pa
ni c
(v s.
W hi
te )
0. 78
2∗ ∗∗
0. 24
3 6.
62 5∗
∗∗ 0.
41 0
0. 75
4∗ ∗∗
0. 23
3 7.
13 2∗
∗∗ 0.
43 7
0. 97
1∗ ∗∗
0. 17
2 9.
02 9∗
∗∗ 0.
41 8
ow ne
r oc
cu pi
ed pr
op er
ty −1
.9 04
∗∗ ∗ 0
.2 69
−2 .5
86 ∗∗
∗ 0.
51 4
−1 .7
82 ∗∗
∗ 0 .2
13 −2
.6 80
∗∗ ∗
0. 51
7 −2
.0 33
∗∗ ∗ 0
.1 67
−3 .2
37 ∗∗
∗ 0 .4
46 bo
rr ow
er in
co m
e (r
el at
iv e
to A
M I,
% )
0. 15
4 0.
14 5
−3 .6
14 ∗∗
∗ 0.
24 3
0. 24
9* *
0. 08
4 −3
.3 96
∗∗ ∗
0. 22
9 0.
21 3∗
∗∗ 0.
05 7
−3 .9
10 ∗∗
∗ 0 .2
41 m
in or
it y
tr ac
t (1
0− 30
% vs
< 10
% )
−0 .0
19 0.
10 7
−0 .0
05 0.
20 4
−0 .0
09 0.
10 6
0. 00
8 0.
16 8
−0 .0
99 0.
10 8
−0 .1
13 0.
18 6
hi gh
m in
or it
y tr
ac t
(> 30
% vs
< 10
% )
0. 37
1 0.
29 2
0. 48
5 0.
64 7
0. 38
0 0.
32 1
0. 53
6 0.
53 8
0. 02
0 0.
27 2
0. 15
2 0.
43 8
lo w
in co
m e
tr ac
t (v
s up
pe r
in co
m e)
2. 57
9∗ ∗∗
0. 66
6 6.
48 4∗
∗∗ 1.
15 6
3. 27
3∗ ∗∗
0. 76
8 7.
54 9∗
∗∗ 1.
06 2
3. 27
7∗ ∗∗
0. 56
0 7.
34 2∗
∗∗ 0.
73 9
m od
er at
e in
co m
e tr
ac t
(v s
up pe
r in
co m
e) 1.
13 7∗
∗∗ 0.
22 2
2. 84
7∗ ∗∗
0. 58
9 1.
21 7∗
∗∗ 0.
18 3
2. 94
5∗ ∗∗
0. 48
2 1.
55 5∗
∗∗ 0.
19 6
3. 55
3∗ ∗∗
0. 46
0 m
id dl
e in
co m
e tr
ac t
(v s
up pe
r in
co m
e) 0.
67 7∗
∗∗ 0.
10 8
1. 77
7∗ ∗∗
0. 23
8 0.
65 4∗
∗∗ 0.
10 8
1. 62
3∗ ∗∗
0. 20
7 0.
68 8∗
∗∗ 0.
08 8
1. 78
3∗ ∗∗
0. 22
1 tr
ac t
po pu
la ti
on (i
n lo
g) −0
.3 57
∗∗ ∗ 0
.1 03
−0 .7
63 ∗∗
∗ 0.
18 1
−0 .3
37 ∗∗
0. 11
3 −0
.7 34
∗∗ ∗
0. 16
0 −0
.1 62
∗ 0.
08 4
−0 .1
66 0.
14 4
0– 10
pu rc
ha se
lo an
s in
tr ac
t (v
s. >
80 )
0. 91
9 0.
57 7
2. 02
6∗ 0.
85 1
0. 97
6 0.
71 4
1. 08
3 0.
66 3
1. 61
8∗ ∗
0. 51
2 3.
42 4∗
∗∗ 0.
59 6
The Impact of the Home Valuation Code of Conduct 685
T ab
le 8
� C
on ti
ue d.
3 M
on th
s P
re -
an d
P os
t- H
V C
C 6
M on
th s
P re
- an
d P
os t-
H V
C C
12 M
on th
s P
re -
an d
P os
t- H
V C
C
C ol
la te
ra l
D en
ia l
D en
ia l
C ol
la te
ra l
D en
ia l
D en
ia l
C ol
la te
ra l
D en
ia l
D en
ia l
P ar
am et
er C
oe f.
S td
. E
rr .
C oe
f. S
td .
E rr
. C
oe f.
S td
. E
rr .
C oe
f. S
td .
E rr
. C
oe f.
S td
. E
rr .
C oe
f. S
td .
E rr
.
11 –2
0 pu
rc ha
se lo
an s
in tr
ac t
(v s.
> 80
) −0
.0 54
0. 26
8 0.
09 4
0. 41
4 0.
13 8
0. 23
6 −0
.0 33
0. 29
0 0.
68 2
∗∗ 0.
22 7
1. 58
2∗ ∗∗
0. 33
4 21
–4 0
pu rc
ha se
lo an
s in
tr ac
t (v
s. >
80 )
0. 01
6 0.
17 3
−0 .6
15 ∗
0. 31
9 0.
07 9
0. 14
6 −0
.5 87
∗ 0.
24 4
0. 43
2∗ ∗∗
0. 12
5 0.
58 7∗
∗ 0.
20 5
41 –8
0 pu
rc ha
se lo
an s
in tr
ac t
(v s. >
80 )
0. 08
3 0.
13 4
0. 06
4 0.
22 6
0. 09
0 0.
11 9
−0 .1
63 0.
19 9
0. 22
3 ∗∗
0. 08
7 0.
29 9
0. 16
2 fo
re cl
os ur
e ra
te 3–
7% (v
s. <
3% )
0. 27
8* 0.
12 5
1. 81
7∗ ∗∗
0. 30
4 0.
42 0∗
∗∗ 0.
11 8
1. 90
9∗ ∗∗
0. 27
6 0.
47 4∗
∗∗ 0.
12 1
2. 42
3∗ ∗∗
0. 24
5 fo
re cl
os ur
e ra
te 7–
11 %
(v s.
< 3%
) 1.
32 1∗
∗∗ 0.
40 6
5. 73
6∗ ∗∗
0. 90
0 1.
82 8∗
∗∗ 0.
35 7
7. 00
4∗ ∗∗
0. 78
7 2.
14 9∗
∗∗ 0.
32 8
8. 34
6∗ ∗∗
0. 82
4 fo
re cl
os ur
e ra
te >
11 %
(v s.
< 3%
) 3.
57 0∗
∗ 1.
25 5
13 .6
82 ∗∗
∗ 2.
86 7
5. 27
3∗ ∗∗
1. 01
8 17
.9 70
∗∗ ∗
3. 13
6 4.
29 2∗
∗∗ 0.
90 5
18 .1
80 ∗∗
∗ 2.
40 4
tr ac
t de
ni al
ra te
in pr
ev io
us ye
ar 0.
09 4∗
∗∗ 0.
00 9
0. 31
1∗ ∗∗
0. 02
0 0.
08 6∗
∗∗ 0.
00 8
0. 30
2∗ ∗∗
0. 01
6 0.
07 0∗
∗∗ 0.
00 8
0. 27
7∗ ∗∗
0. 02
0 L
en de
r du
m m
y ye
s ye
s ye
s Y
es ye
s ye
s M
S A
D um
m y
ye s
ye s
ye s
ye s
ye s
ye s
M od
el fi
t A
d ju
st ed
R 2
0. 01
9 0.
05 4
0. 01
9 0.
06 8
0. 02
1 0.
07 0
N u m
b er
o f
o b se
rv a ti
o n s
25 8,
89 0
25 8,
89 0
54 0,
88 5
54 0,
88 5
1, 10
8, 20
4 1,
10 8,
20 4
∗∗ ∗ s
ig ni
fi ca
nt at
0. 00
1 le
ve l;
∗∗ si
gn ifi
ca nt
at 0.
01 le
ve l;
∗ s ig
ni fi
ca nt
at 0.
05 le
ve l;
er ro
rs ar
e cl
us te
re d
at M
S A
le ve
l. N
o te
: S
ee T
ab le
1 fo
r de
fi ni
ti on
s of
pr e-
an d
po st
-H V
C C
pe ri
od s;
fi rs
t- li
en ,
on e
to fo
ur fa
m il
y pr
op er
ti es
, w
it h
lo an
am ou
nt s
ab ov
e 50
% of
th e
co nfi
rm in
g lo
an li
m it
on ly
; ba
se d
on th
e H
M D
A da
ta .
686 Ding and Nakamura
T ab
le 9
� G
eo gr
ap hi
c re
pr es
en ta
ti ve
ne ss
of th
e F
N C
da ta
, re
la ti
ve to
H M
D A
da ta
.
20 08
F N
C D
at a,
C om
pa re
d w
it h
H M
D A
20 09
F N
C D
at a,
C om
pa re
d w
it h
H M
D A
N um
be r
of %
of T
ot al
% of
T ot
al %
of T
ot al
A pp
ra is
al /
A pp
ra is
al /
N um
be r
of %
of T
ot al
% of
T ot
al %
of T
ot al
A pp
ra is
al /
A pp
ra is
al /
S ta
te A
pp ra
is al
s (A
pp ra
is al
) (A
pp li
ca ti
on )
(O ri
gi na
ti on
) A
pp li
ca ti
on O
ri gi
na ti
on S
ta te
A pp
ra is
al s
(A pp
ra is
al )
(A pp
li ca
ti on
) (O
ri gi
na ti
on )
A pp
li ca
ti on
O ri
gi na
ti on
C al
if or
ni a
50 ,8
21 21
.3 4
11 .8
2 10
.2 9
1. 81
2. 07
C al
if or
ni a
85 ,9
63 23
.3 3
13 .3
4 12
.3 2
1. 75
1. 89
T ex
as 20
,9 83
8. 81
8. 96
9. 17
0. 98
0. 96
T ex
as 23
,3 57
6. 34
8. 39
8. 46
0. 76
0. 75
F lo
ri da
23 ,0
72 9.
69 6.
71 5.
71 1.
44 1.
70 F
lo ri
da 39
,4 45
10 .7
1 6.
22 5.
53 1.
72 1.
94 N
ew Y
or k
5, 21
5 2.
19 4.
50 4.
14 0.
49 0.
53 N
ew Y
or k
8, 34
9 2.
27 4.
14 3.
93 0.
55 0.
58 Il
li no
is 5,
12 6
2. 15
3. 76
3. 83
0. 57
0. 56
Il li
no is
9, 29
1 2.
52 3.
53 3.
59 0.
71 0.
70 P
en ns
yl va
ni a
4, 85
9 2.
04 3.
57 3.
92 0.
57 0.
52 P
en ns
yl va
ni a
8, 52
1 2.
31 3.
59 3.
86 0.
64 0.
60 N
or th
C ar
ol in
a 9,
85 8
4. 14
3. 53
3. 81
1. 17
1. 09
N or
th C
ar ol
in a
12 ,5
71 3.
41 3.
16 3.
28 1.
08 1.
04 G
eo rg
ia 15
,1 76
6. 37
3. 38
3. 40
1. 88
1. 87
G eo
rg ia
18 ,3
52 4.
98 3.
04 2.
98 1.
64 1.
67 O
hi o
2, 09
9 0.
88 3.
13 3.
35 0.
28 0.
26 O
hi o
5, 66
9 1.
54 3.
11 3.
30 0.
50 0.
47 V
ir gi
ni a
12 ,5
46 5.
27 2.
88 3.
15 1.
83 1.
67 V
ir gi
ni a
18 ,1
32 4.
92 3.
00 3.
28 1.
64 1.
50 A
ri zo
na 6,
04 3
2. 54
2. 74
2. 70
0. 93
0. 94
A ri
zo na
10 ,8
47 2.
94 2.
99 3.
02 0.
98 0.
97 N
ew Je
rs ey
4, 57
9 1.
92 2.
65 2.
51 0.
72 0.
76 N
ew Je
rs ey
8, 06
0 2.
19 2.
60 2.
48 0.
84 0.
88 M
ic hi
ga n
5, 31
8 2.
23 2.
57 2.
46 0.
87 0.
91 M
ic hi
ga n
9, 39
6 2.
55 2.
46 2.
36 1.
04 1.
08 C
ol or
ad o
2, 48
7 1.
04 2.
38 2.
55 0.
44 0.
41 C
ol or
ad o
6, 20
8 1.
69 2.
29 2.
46 0.
74 0.
69 W
as hi
ng to
n 7,
59 9
3. 19
2. 37
2. 42
1. 35
1. 32
W as
hi ng
to n
9, 60
1 2.
61 2.
44 2.
46 1.
07 1.
06 T
en ne
ss ee
6, 35
5 2.
67 2.
16 2.
31 1.
23 1.
15 T
en ne
ss ee
8, 56
7 2.
33 2.
11 2.
20 1.
10 1.
06 In
di an
a 3,
03 8
1. 28
2. 05
2. 18
0. 63
0. 59
In di
an a
6, 16
1 1.
67 2.
03 2.
11 0.
83 0.
79 M
is so
ur i
3, 48
4 1.
46 2.
03 2.
21 0.
72 0.
66 M
is so
ur i
4, 04
4 1.
10 2.
01 2.
16 0.
55 0.
51 M
as sa
ch us
et ts
3, 25
8 1.
37 1.
91 1.
98 0.
72 0.
69 M
as sa
ch us
et ts
6, 50
9 1.
77 2.
08 2.
13 0.
85 0.
83 O
th er
S ta
te s
46 ,2
52 19
.4 2
26 .9
1 27
.8 9
0. 72
0. 70
O th
er S
ta te
s 69
,3 58
18 .8
3 27
.4 7
28 .1
3 0.
69 0.
67 U
.S .
(T ot
al )
23 8,
16 8
U .S
. (T
ot al
) 36
8, 40
1
N o
te :
B as
ed on
th e
F N
C da
ta an
d th
e H
M D
A da
ta ;
A p
p ra
is a
l/ A
p p
li ca
ti o
n re
pr es
en ts
th e
ra ti
o be
tw ee
n a
st at
e’ s
sh ar
e of
ap pr
ai sa
ls (o
ut of
th e
na ti
on al
to ta
l) an
d it
s sh
ar e
of ap
pl ic
at io
ns .
If th
e to
ta l
is gr
ea te
r th
an on
e, th
en th
e ap
pr ai
sa l
is ov
er re
pr es
en te
d in
th is
st at
e, re
la ti
ve to
th e
sh ar
e of
H M
D A
lo an
ap pl
ic at
io ns
in th
is st
at e.
If th
e ra
ti o
is le
ss th
an on
e, th
en it
is un
de rr
ep re
se nt
ed .
A p
p ra
is a
l/ O
ri g
in a
ti o
n re
pr es
en ts
th e
ra ti
o be
tw ee
n a
st at
e’ s
sh ar
e of
ap pr
ai sa
ls (o
ut of
th e
na ti
on al
to ta
l) an
d it
s sh
ar e
of pu
rc ha
se lo
an or
ig in
at io
ns .
The Impact of the Home Valuation Code of Conduct 687
Overall, the results suggest that the HVCC helps explain the relatively higher mortgage denial rates in the treatment group, likely due to the higher incidence of low appraisals and the resulting higher rates of collateral denials. The HVCC effects are significant in the purchase market, which could have a negative impact on the access to mortgage credit for potential borrowers.
Other Determining Factors of Low Appraisals and Mortgage Denial
There is extensive literature on mortgage lending decisions, and our findings are generally consistent with those in early studies: Applications for not- owner-occupied properties and those in less desirable neighborhoods (e.g., those with high foreclosure rates, low income, or fewer prior originations) are more likely to be denied. Note also that applications for minority borrowers are also more likely to be denied, but this should not be taken as evidence that minorities are being discriminated against. Rather, because we do not have creditworthiness variables in this regression, it is well-known that minorities tend to have lower credit scores and this typically accounts for much of their higher denial rates (Ross and Yinger 2002). We focus the discussion of the empirical results of other controls on factors influencing appraisals here.
Market Inefficiency: Foreclosures and Prior Market Activities
The regression results confirm that the probability of low appraisals is higher in neighborhoods with higher foreclosure rates and those with a lower level of market activities. The results are generally consistent for different cohorts and when different outcome measures are used. The results confirm that an insufficient number of mortgage-financed sales leads to increased probability of low appraisals. The positive association with the foreclosure rate could be explained by the negative impact of foreclosed properties as comparables on appraisals or simply because foreclosure rates represent neighborhood risk. The impact of prior transactions is also understandable as a sufficient volume of market sales aids in price discovery, allows lenders to distinguish observable risks, and leads to a lower share of low appraisals observed.
Lender Heterogeneity
Due to increased regulations, especially after the enactment of the HVCC and the Dodd-Frank Act, appraisal management companies, which act as intermediaries between lenders and appraisers and manage appraisals for lenders, have taken a greater market share.32 Results from the logit regressions
32The FNC data suggest that the share of appraisals conducted by appraisers through AMCs increased from about 4.5% in the second quarter of 2009 to more than 47.6%
688 Ding and Nakamura
demonstrate that the incidence of low appraisals for those ordered by lenders directly is much lower compared with that of AMC appraisals (about 2.5 percentage points to 7.4 percentage points lower in our preferred model). While the incidence of low appraisal does not necessarily represent the quality of appraisals, the results provide some evidence of significant heterogeneity in the appraisal approaches or the enforcement of existing regulations by different lenders.
Conclusion and Policy Implications
Appraisers are expected to provide unbiased opinions about the value of as- sets. However, appraisers have been criticized for the often inflated home values that occurred during the housing boom, as well as overly conser- vative valuations during the housing bust. As market conditions and legal environments have changed radically during the housing crisis, this study demonstrates how one important rule adopted during the crisis, the HVCC, impacts appraisal and mortgage outcomes.
The HVCC was designed to make appraisals more independent and objective primarily by addressing issues related to the conflicts of interest in the ap- praisal process. The HVCC has done part of what it was supposed to do by reducing inflated valuations that were prevalent during the subprime boom. At the same time, the HVCC rule led to a significant increase in low appraisals and caused the origination of purchase mortgages to be more difficult.
The results demonstrate that in an effort to prevent appraisal frauds and artificially inflated home valuations, the HVCC, which attempted to change the entire appraisal industry fundamentally when the market was in turmoil, had quickly induced significant changes in the housing market. It seems lenders and appraisers generally responded to the rule by becoming more conservative, especially when market conditions were changing rapidly during the housing crisis. The results should shed light on how to design intervention programs in the future, especially during a crisis that is often characterized by a high level of uncertainty and panic.
The authors thank Crocker Liu, Andreas Fuster, Walter Torous, Robert Hunt, Theresa Y. Singleton, an anonymous referee, and participants of the American Real Estate and Urban Economics Association (AREUEA) 2014 National
in the fourth quarter of 2011. The estimate from another source was that the share of AMC appraisals increased from less than half (15% to 50%) before the HVCC to between 60% and 80% in 2011, according to some appraisal industry participants (GAO 2011).
The Impact of the Home Valuation Code of Conduct 689
Conference and of the Federal Reserve System Meeting on Financial Structure and Regulation for their helpful comments. Special thanks to Barbara Brynko for her editing and Naakorkoi Pappoe for her assistance. An early version of this paper received the Best Paper Award in Real Estate Valuation at the 2014 AREUEA National Conference. The views expressed in this paper are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of Philadelphia or the Federal Reserve System.
References
Abernethy, A.M. and H. Hollans. 2010. The Home Valuation Code of Conduct and Its Potential Impacts. The Appraisal Journal 78(1): 71–93. Agarwal, S., B.W. Ambrose and V. Yao. 2014. The Limits of Regulation: Appraisal Bias in the Mortgage Market. Available at http://papers.ssrn.com/sol3/papers.cfm? abstract_id=2487393. ———, I. Ben-David and ———. Forthcoming. Collateral Valuation and Borrower Financial Constraints: Evidence from the Residential Real-Estate Market. Management Science. Appraisal Institute. 2008. Making Sense of the Nonsense, Providing Cred- ible Appraisals in a Declining Market. Available at http://www.appraisal institute.org/assets/1/7/AI_AppraisalsInDecliningMarkets_(1).pdf. Blackburn, M. and T. Vermilyea. 2007. The Role of Information Externalities and Scale Economies in Home Mortgage Lending Decisions. Journal of Urban Economics 61(1): 71–85. Bostic, R.W., K.C. Engel, P.A. McCoy, A. Pennington-Cross and S.M. Wachter. 2008. State and Local Antipredatory Lending Laws: The Effect of Legal Enforcement Mechanisms. Journal of Economics and Business 60(1–2): 47–66. Calem, P.S., L. Lambie-Hanson and L.I. Nakamura. 2014. Information Losses in Home Purchase Appraisals. Working Paper, Federal Reserve Bank of Philadelphia. Chinloy, P., M. Cho and I.F. Megbolugbe. 1997. Appraisals, Transaction Incentives, and Smoothing. Journal of Real Estate Finance and Economics 14(1): 89–111. Cho, M. and I.F. Megbolugbe. 1996. An Empirical Analysis of Property Appraisal and Mortgage Redlining. Journal of Real Estate Finance and Economics 13(1): 45–55. Fuster, A. and J. Vickery. 2014. Securitization and the Fixed-Rate Mort- gage. Federal Reserve Bank of New York Staff Report, Available at http://www.newyorkfed.org/research/staff_reports/sr594.html. Harvey, K.D. and P.J. Nigro. 2004. Do predatory lending laws influence mortgage lending? An analysis of the North Carolina predatory lending law. The Journal of Real Estate Finance and Economics 29(4): 435–456. Inside Mortgage Finance. 2013. The 2013 Mortgage Market Statistical Annual (Vol. I). Bethesda, MD: The Primary Market. Kaufman, A. 2014. The Influence of Fannie and Freddie on Mortgage Loan Terms. Real Estate Economics 42(2): 472–496. LaCour-Little, M. and R. Green. 1998. Are Minorities or Minority Neighborhoods More Likely to Get Low Appraisals? Journal of Real Estate Finance and Economics 16(3): 301–315. Ladd, H.F. 1998. Evidence on Discrimination in Mortgage Lending. The Journal of Economic Perspectives 12(2): 41–62.
690 Ding and Nakamura
Lang, W.M. and L.I. Nakamura. 1993. A Model of Redlining. Journal of Urban Economics 33(2): 223–234. LaCour-Little, M. and S. Malpezzi. 2003. Appraisal Quality and Residential Mortgage Default: Evidence from Alaska. Journal of Real Estate Finance and Economics 27(2): 211–233. Liu, H.C., A. Nowak and S. Rosenthal, 2014. Bubbles, Post-Crash Dynam- ics, and the Housing Market, Working Paper, Available at http://www.business. uconn.edu/finance/seminars/papers/Stuart%20Rosenthal%20Paper.pdf. Nakamura, L.I. 2010. How Much Is That Home Really Worth? Appraisal Bias and House-Price Uncertainty. Federal Reserve Bank of Philadelphia Business Review First Quarter, 11–22. Pence, K.M. 2006. Foreclosing on Opportunity: State Laws and Mortgage Credit. Review of Economics and Statistics 88(1): 177–182. Ross, S. and J. Yinger. 2002. The Color of Credit: Mortgage Discrimination, Research Methodology, and Fair-Lending Enforcement. Cambridge, M A: MIT Press. Tzioumis, K. Forthcoming. Appraisers and Valuation Bias: An Empirical Analysis. Real Estate Economics. U.S. Government Accountability Office (GAO). 2011. Residential Appraisals: Op- portunities to Enhance Oversight of an Evolving Industry. July 2011. Available at http://www.gao.gov/new.items/d11653.pdf. ———. 2012. Residential Appraisals: Regulators Should Take Ac- tions to Strengthen Appraisal Oversight. June 28, 2012. Available at http://www.gao.gov/assets/600/592000.pdf.
Copyright of Real Estate Economics is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
