Sales comparison adjustments for FHA and VA financing after deregulation.
The objective of this study is to establish the adjustments needed for FHA- and VA-financed homes relative to conventionally financed homes. The effects of FHA and VA financing on home prices are determined using a hedonic framework and a database of approximately 10,000 home sales. Analyses of these sales indicate both types of government-backed financing are associated with reductions in selling prices. Specifically, FHA and VA financing result in price discounts of 5.8% and 4.2%. The appraisal implications of these findings are that comparables with FHA and VA financing are to be adjusted upward relative to a property with conventional financing.
This study establishes the form of adjustments necessary for FHA- and VA-financed comparables after deregulation. (1) The findings of this study are based on a hedonic study of the impacts of FHA and VA financing on home prices. Non-conventional loan borrowers (those using FHA and VA default protection) borrow at higher loan-to-value ratios (LVRs) and, under FHA underwriting criteria, qualify with lower credit scores relative to conventional underwriting standards. One might then expect that these lenient standards would lead to higher default rates. (2) In fact, recent data shows that FHA default rates are approximately four times the default rates of conventional loans (for 2003, 12.2% versus 3.5%), as shown in Table 1. Since higher-default-risk borrowers are required to pay risk premiums one way or another, one would expect that the reservation prices (or bids) for FHA and VA borrowers would likely be lower than would otherwise be the case.
Prior to deregulation, borrowers were prohibited from paying interest rate differentials when specified rates on FHA and VA loans were lower than market rates. As a result, sellers had to subsidize these costs in the form of seller-paid points. Sellers in turn shifted the points back to borrowers in the form of higher prices. (3) In sum, both FHA- and VA-financed sales were associated with higher prices prior to deregulation due to the cost-shilling behavior on the part of sellers. The appraisal implication then was that comparables with FHA and VA financing were to be adjusted downward relative to a subject property that was conventionally financed. (4)
Subsequent to deregulation, seller subsidies continued but they took various forms, including seller interest rate subsidies and seller concessions to buyers. Asabere and Huffman (5) in their 1997 study fend a 3% discount associated with seller concessions in the form of buyer interest rate buy-downs for conventional loans. The current study reports evidence showing that there still exist seller subsidies in the marketplace. However, given that the costs of the FHA insurance or VA guarantee are borne by buyers after deregulation, the cost-shifting behavior is now on the part of buyers. Specifically, the current study finds that sales involving FHA and VA financing are associated with 5.8% and 4.2% price discounts.
Of course, the appraisal implication now is that comparables with FHA and VA financing are to be upward adjusted relative to a subject property that is conventionally financed. Thus, what sets this study apart from the previously cited studies is the fact that this is a post-deregulation study. Unlike the previous studies that found FHA and VA financing to be associated with premium prices, this post-deregulation study has established that FHA and VA financing are associated with discount prices. The study also uses a rich database of close to 10,000 observations on San Antonio, Texas, a housing market with significant percentages of FHA (32%) and VA (15%) financing. The next section presents the empirical framework for establishing the prevailing price discounts.
An important risk associated with a mortgage is default risk. While all mortgages carry some default risk, FHA and VA mortgages have been found to be more prone to default relative to conventional mortgages. Thus, FHA and VA mortgages would carry higher default risk. The hypothesis here is that the higher default risk has implications for the pricing of FHA and VA mortgages.
Assume that the effective interest rate ([i.sup.*]) that the lender would charge on any mortgage is given by:
[i.sup.*] = i + DRP (1)
[i.sup.*] = the required interest rate
i = the market interest rate
DRP= default risk premium
Thus, the greater the DRP, the higher the required interest rate; and the smaller the DRP, the lower the required interest rate. In a competitive market with no default insurance, the lender would account for default risk by requiring the higher rates ([i.sup.*]). However, with the government-backed FHA insurance or the VA default-guarantee protection, the lender would willingly accept the market interest rate (i), provided the perceived default risk is fully covered by the mortgage insurance or the loan guarantee. Of course, the FHA or VA borrower must pay for the insurance protection or the loan guarantee in the form of premiums or funding fees.
The central premise of this study is that the insurance premiums and fees associated with the use of a FHA or VA mortgage would dampen buyer reservation prices (or bids) and ultimately motivate the seller to accept price discounts. As noted by an anonymous reviewer, what is the economic rational for the seller to sell for less to any buyer in an arm's-length transaction? Search theory (6) tells us that both the home seller and home buyer face substantial opportunity costs due to asymmetric information. Potential seller opportunity costs include additional mortgage payments, maintenance expenses, time value of money, and other carrying costs. Such carrying costs increase over time significantly enough to create rapid real discounting behavior especially on the part of desperate sellers.
While the seller initially has no benevolent intentions whatsoever to sell for less to any user, eventually the seller will be motivated to provide a price discount for a buyer with a lower reservation price (or bid) in an attempt to minimize the seller's own opportunity costs. As a rule, the seller will accept a lower bid as long as the price discount does not exceed the seller's own savings in opportunity costs. Of course, such seller behavior would produce unintended, implicit subsidies for FHA and VA users whose reservation prices (or bids) would likely be lower. Indeed prior literature (7) found a reverse form of capitalization effects and cost shifting in the 1970s, when sellers shifted seller-paid points back to buyers in the form of price premiums.
The current study uses hedonic analyses to explore the potential price discounts associated with FHA and VA mortgage financing relative to conventional mortgage financing. The hedonic model employed to detect the partial effects of FHA and VA mortgage financing is as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
SPRICE = sale price of the house
FHA = FHA mortgage underwriting dummy variable
VA = VA mortgage underwriting dummy variable
CON = conventional mortgage underwriting dummy variable
[X.sub.ij] = various housing and other pedestrian control variables
A logarithmic transformation of Equation 1 will be used to measure the potential effects of FHA and VA mortgage financing relative to conventional mortgage financing. It is expected that the estimated coefficients of [[beta].sub.1] and [[beta].sub.2] will be significantly negative. It is also expected that [[beta].sub.1] > [[beta].sub.2] given that, as shown in Table 1, default risk rates for FHA loans are higher than those for VA loans. The next section describes the data study.
The Office of the Comptroller, State of Texas, supplied the database for this study. (8) The database consisted of over 10,000 sales of residential property occurring from April 2001 to March 2002 in and around Bexar County, Texas. The data contains sale price and pedestrian information on various price-determining factors. These factors include housing characteristics, lot characteristics, locational effects, and transaction-related characteristics. The variables also include the specific variables of interest in this analysis-the mortgage underwriting options of FHA-insured, VA-guaranteed, or conventional financing.
The transaction-related variables include sale price (SPRICE) and date of sale in sequential months from April 2001 to March 2002 (MONTH). Specific variables used to measure the effect of housing characteristics are square feet of improvements (SQFT), number of bedrooms (BEDS), number of bathrooms (BATHS), age of structure (AGE), presence of a fireplace (FIREPLACE), and two variables to account for exterior features: brick construction (BRICK) and type of foundation, slab (SLAB) versus all other. Lot characteristics include corner lot location (CORNER), cul-de-sac location (CULDESAC), and two lot improvement variables: the presence of a deck (DECK) and in-ground pool (POOL).
Neighborhood variables include the effect of a neighborhood amenity--the proximity of a park or other recreational area (YESNEIGH). Other locational factors include the school district in which the home is located (from SD1 to SD24) and the distance (RING1 to RING4) from the central city district of San Antonio.
School districts for each property are specified in the data and cross-referenced with a standard map of the area. (9) Distance from the central city is first determined by locating the zip code for each property. Taking advantage of the physical layout of the San Antonio urban area, each location is referenced with respect to the highway system present and uses the area's encircling highway system to define four locational areas bounded by rough concentric circles as separated by the highway system. RING1 consists of all sales in zip codes with a majority of its location within the central district. (10) RING2 consists of those properties with zip codes outside the central downtown district and within the Route 410 outer highway. RING3 consists of those properties with zip codes outside Highway 410 but inside of Highway 1604. RING4 properties are those properties with zip codes outside Highway 1604 but within the boundaries of Bexar County. Classical concentric circle theory would imply that properties in areas closer to the central city would sell at premiums related to accessibility and transportation costs. (11)
Sales observations were eliminated where there was a lack of data or a likelihood that the sale was spurious or contained unreliable information. After these adjustments, 9,317 sales observations remained with the variables listed in Table 2. The specific variables used in the analysis are presented in Table 2 along with basic descriptive statistics.
Using the data described above, the hedonic analysis was performed using the natural logarithm of sale price (lnSPRICE) as the dependent variable. Several regressions were run including models utilizing the entire database, and a model utilizing only data on houses with SQFT less than 3,200 (not reported here). The results of the dummy variables on FHA and VA were not qualitatively different; (12) consequently, only Model 1 utilizing the entire database is reported and shown in Table 3. An examination of the correlation matrix also did not indicate high collinearity between FHA/VA and SQFT. Table 4 reports ANOVA statistics for the key variables used for the hedonic model shown in Table 3.
The adjusted [R.sup.2] for Model 1 is 70%. The [R.sup.2] of 70% is well within reasonably acceptable levels for explaining sale prices in a large geographic area such as the study area. The F-statistics for all models are also high and statistically significant at a 99% level of confidence, indicating an acceptable statistical fit for the model. An examination of multicollinearity and standard regression diagnostics reveal no undue problems of statistical bias usually associated with regression studies.
First, the results of the non-financing control variables are examined. Specifically, SQFT, INPOOL, DECK, YESNEIGH, CORNER, CULDESAC, and MONTH are found to be significantly positive at conventional levels for the model. All variables have expected signs. It must be noted that BEDROOMS and BATHS are not included in the reported equations in Table 3 because they would be collinear with SQFT.
Variables with significantly negative coefficients are AGE and RING1, RING2, and RING3. The three rings (RING1, RING2, and RING3) are relative to the omitted RING4 (the outer ring). These results do not appear to conform to classical concentric circle theory. Rather the results may be picking up the differences associated to higher-priced properties located in the less dense suburban areas of the county.
Several of the school districts are significant with various signs (SD1, SD2, SD4, SD9, SD13-SD17, SD22, and SD23) relative to one omitted district (SD13). These results imply that there is a net school district effect in the Bexar County data. Insignificant variables include SLAB, BRICK, and four of the school districts (SD11, SD16, SD18, and SD20).
Turning to the FHA and VA underwriting variables, the estimated coefficients for FHA and VA are significantly negative at the 99% level of confidence for the model. The magnitudes of the estimated coefficient are -0.060 for FHA sales. The corresponding figures for VA sales are -0.043. These imply price discounts of roughly 5.8% and 4.2% for FHA and VA financing. The results also show that the price discount associated with FHA financing is slightly higher than the one associated with VA, a result that conforms to the hypothesis and to the higher default rates found in Table 1 for FHA-financed mortgages.
Summary and Conclusions
The results support the hypotheses that homes associated with FHA and VA mortgage financing will be associated with reductions in equilibrium selling prices relative to conventional financing. The magnitude of the estimated coefficient on FHA sales in Model 1 is -0.060, indicating a price discount or capitalization effect of about 5.8%. The corresponding coefficient on VA financing is -0.043, implying a price discount of roughly 4.2%. The appraisal implications of these price discounts are that comparables with FHA and VA financing are to be adjusted upward in the magnitude of 5.8% and 4.2%, respectively, relative to a subject property that is conventionally financed and vice versa. However, it must be pointed out that the discounts found are probably dynamic and may change with market conditions and location. (13)
by Paul K. Asabere, PhD, and Forrest E. Huffman, PhD
(1.) Deregulation refers to the federal law "Depository Institutions Deregulation and Monetary Control Act of 1980," (12 U.S.C. [section] 226 et seq.) deregulating deposit interest rates and expanding access to the Federal Reserve Discount Window; it was the first major reform of the U.S. banking system since the 1930s. The act has two main sections: Title 1, the Monetary Control Act, which extends reserve requirements to all U.S. banking institutions and also deals with the banking services furnished by the Federal Reserve System; and Title 2, the Depository Institutions Deregulation Act of 1980, phasing out Federal Reserve Regulation Q deposit interest rate ceilings. Among other things, the law preempted state usury laws limiting rates lenders could charge on residential mortgage loans.
(2.) As correctly noted by an anonymous reviewer, the main reason why buyers use FHA-insured or VA-guaranteed mortgages is not the protection from likelihood of foreclosure. They use this type of financing mainly because it requires a lower downpayment.
(3.) Robert H. Zerbst and William B. Brueggeman, "Adjusting Comparable Sales for FHA and VA Financing," The Appraisal Journal (1979): 374-380; Karl L. Guntermann, "FHA Mortgage Discount Points, House Prices and Consumer Behavior," AREUEA Journal 7 no. 2 (1979): 163-176; and Vinod B. Agarwal and Richard A. Phillips, "The Effect of Mortgage Rate Buydowns on Housing Prices: Recent Evidence from FHA-VA Transactions," AREUEA Journal 11, no. 4 (1983): 491-503.
(4.) Robert H. Zerbst and William B. Brueggeman, "FHA and VA Mortgage Discount Points and Housing Prices," Journal of Finance 32, no. 5 (1977): 1766-1773.
(5.) Paul K. Asabere and Forrest E. Huffman, "Discount Point Concessions and the Value of Homes with Conventional Versus Non-conventional Mortgage Financing," Journal of Real Estate Finance and Economics 15, no. 3 (1997): 261-270.
(6.) See Francis J. Cronin, "The Efficiency of Housing Search," Southern Economic Journal 48, no. 4 (1982): 1016-1030; and Abdullah Yavas, "A Simple Search and the Bargaining Model of Real Estate Markets," AREUEA Journal 20, no. 4 (1992): 533-548.
(7.) See for example Zerbst and Brueggeman, "Adjusting Comparable Sales."
(8.) Paul. K. Asabere and Forrest E. Huffman, "Cost Shifting and the Capitalization of FHA/VA Financing Costs" (working paper, Fox School of Business and Management, Temple University, Philadelphia, 2006).
(9.) Ferguson's Quick-Finder MAPSCO 2000 Street Guide and Directory for San Antonio, 2000, MAPSCO, Inc., Dallas, TX.
(10.) Ibid., 14.
(11.) Concentric circle theory was first developed in the 1836 by Von Thunen as a method to explain urban development and location patterns.
(12.) For example, the magnitude of the estimated coefficient on FHA and VA for the model using the whole database was -0.063 for FHA and -0.045 for VA. The corresponding values for the model with SQFT < 3,200 were -0.059 and -0.045, respectively. These imply that the estimated coefficients on FHA and VA are quite robust.
Paul K. Asabere, PhD, is a professor in the Department of Finance at Temple University, Philadelphia. He received his PhD from the University of Illinois, Urbana-Champaign. He has written numerous articles in the field of real estate and urban economics and is a past contributor to The Appraisal Journal. Contact: email@example.com
Forrest E. Huffman, PhD, is a professor in the Department of Finance at Temple University, Philadelphia. He received his PhD from the University of South Carolina, Columbia. He has written numerous articles on real estate markets and valuation and is a past contributor to The Appraisal Journal.
Table 1 Default (Delinquency) Rates for FHA, VA, and Conventional Mortgage Loans, Selected Years 1990-2003 Mortgage 1990 1995 1997 1998 1999 FHA 6.7 7.6 8.1 8.5 8.6 VA 6.4 6.4 6.9 7.1 6.8 Conventional 3.0 2.8 2.8 3.4 3.2 Mortgage 2000 2001 2002 2003 FHA 9.1 10.8 11.5 12.2 VA 6.8 7.7 7.9 8.0 Conventional 3.2 3.8 3.8 3.5 Source: Adapted from U.S. Census Bureau, Table 1179, Mortgage Delinquency and Foreclosure Rates: 1990 to 2003, Statistical Abstract of the United States, 2004-2005 (Washington, DC: U.S. Department of Commerce, 2004), 745. Table 2 Summary Statistics for Relevant Variables Variable Mean Std. Deviation Minimum Maximum SPRICE 122,252 79,998 14,101 1,100,000 CON .53 .499 0 1 FHA .32 .465 0 1 VA .15 .358 0 1 SQFT 1,914 710 1,000 8,291 AGE 20.94 17.66 0 99 BRICK .09 .281 0 1 INPOOL .08 .278 0 1 SLAB .92 .270 0 1 DECK .18 .388 0 1 YESNEIGH .48 .500 0 1 CORNER .10 .305 0 1 CULDESAC .15 .359 0 1 MONTH 4.99 3.447 0 11 RING1 .04 .205 0 1 RING2 .19 .390 0 1 RING3 .54 .499 0 1 RING4 .23 .422 0 1 SD1 .03 .172 0 1 SD2 .02 .127 0 1 SD3 .00 .018 0 1 SD4 .01 .091 0 1 SD5 .00 .015 0 1 SD6 .01 .099 0 1 SD8 .01 .084 0 1 SD9 .08 .271 0 1 SD10 .00 .031 0 1 SD11 .00 .023 0 1 SD14 .36 .479 0 1 SD16 .01 .117 0 1 SD17 .01 .109 0 1 SD18 .00 .010 0 1 SD19 .00 .018 0 1 SD20 .00 .033 0 1 SD21 .00 .065 0 1 SD22 .00 .033 0 1 SD23 .00 .040 0 1 SD24 .01 .109 0 1 Table 3 Regression Results for Model 1 with Dependent Variable In(SPRICE) Variable [beta] t-Value FHA -0.060 -10.73 *** VA -0.043 -6.21 *** SQFT 4.20E-04 86.01 *** AGE -0.002 -10.10 *** BRICK 0.004 0.44 INPOOL 0.104 10.89 *** SLAB -0.003 -0.21 DECK 0.064 10.21 *** YESNEIGH 0.077 15.26 *** CORNER 0.013 1.65 * CULDESAC 0.012 1.80 * MONTH 0.001 1.95 * RING1 -0.118 -6.16 *** RING2 -0.109 -11.33 *** RINGS -0.052 -7.45 *** SD1 0.562 30.70 *** SD2 0.264 11.77 *** SD3 -0.166 -1.31 SD4 0.170 6.05 *** SD5 -0.265 -1.70 * SD6 -0.033 -1.42 SD8 -0.159 -5.65 *** SD9 -0.200 -21.05 *** SD10 0.135 1.73 * SD11 0.129 1.31 SD14 0.069 -12.19 *** SD16 -0.027 -1.30 SD17 -0.084 -3.82 *** SD18 0.022 0.10 SD19 0.151 2.16 ** SD20 0.066 0.52 SD21 -0.108 -7.72 *** SD22 -0.372 -6.52 *** SD23 -0.270 -3.87 *** SD24 -0.272 -12.61 *** Adjusted [R.sup.2] 0.70 *** F-statistic 583.08 *** Observations 8,829 * Significant at the 90% level. ** Significant at the 95% level. *** Significant at the 99% level. Table 4 ANOVA Results for Model 1 with Dependent Variable In(SPRICE) Sum of Squares df * Mean Square Regression 1693.510 38 44.566 Residual 521.303 9278 5.619E-02 Total 2214.813 9316 F-Value Sig. Regression 793.173 .000 ([dagger]) Residual Total * Degrees of freedom ([dagger]) Predictors: (CONSTANT), MONTH, INPOOL, SD5, Sol, SD10, SD19, SD20, SD22, SD23, SD3, SD18, SD21, SD6, SD7, SDll, SOB, SD4, SD17, SD16, SD24, SD2, CORNER, ALLBRICK, RING1, SD9, DOM, DECK, CULDESAC, VA, YESNEIGH, FHA, SD14, RING3, SQFT, SLAB, RING2, SD15, AGE
|Printer friendly Cite/link Email Feedback|
|Author:||Asabere, Paul K.; Huffman, Forrest E.|
|Date:||Mar 22, 2007|
|Previous Article:||Seller disclosure and buyer knowledge: how they affect market value.|
|Next Article:||Partitioning capitalization rates: operating leases in unitary valuation.|