the impact of homeowners’ housing wealth misestimation on
TRANSCRIPT
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
2007 V35 2: pp. 135–154
REAL ESTATE
ECONOMICS
The Impact of Homeowners’ HousingWealth Misestimation on Consumptionand Saving DecisionsSumit Agarwal∗
Using a unique data set of 81,943 house value estimates by the homeown-ers and their financial institution, I find that homeowners overestimate theirhouse value by 3.1%. After controlling for homeowners’ socioeconomic char-acteristics, I find that ex ante homeowners who rate (cash-out) refinance anexisting loan to increase savings (consumption) are significantly more likelyto underestimate (overestimate) their house value. Moreover, overestimators(underestimators) are more likely to increase (reduce) their spending ex post.Finally, I also find that underestimators are more likely to prepay their loans andoverestimators are more likely to default on their loans.
There is general agreement in the literature that homeowners significantly mis-estimate their house value.1 The average absolute misestimation ranges between14% and 25%. Kish and Lansing (1954) and Kain and Quigley (1972) also findthat homeowners’ misestimation is systematically correlated to their socioeco-nomic characteristics. Goodman and Ittner (1992) do not find any such correla-tion but argue that if socioeconomic characteristics are systematically related tohomeowners’ misestimation of the house value, then it would lead to errors inhousehold consumption and savings decisions because of their perceived (vs.actual) housing wealth.
The literature has studied the impact of the homeowners’ housing wealth es-timation on their consumption and savings decision using the Panel Study ofIncome Dynamics (PSID). Skinner (1989) finds that housing wealth increasedconsumption, while Engelhardt (1996) finds that households experiencing cap-ital losses reduced consumption. Hoynes and McFadden (1997) do not find any
∗Federal Reserve Bank of Chicago, Chicago, IL 60604 or [email protected].
1 See Kish and Lansing (1954), Kain and Quigley (1972), Robins and West (1977),Follain and Malpezzi (1981), Ihlanfeldt and Martinez-Vazquez (1986), DiPasquale andSomerville (1995) and Bucks and Pence 2005. Specifically, Kain and Quigley (1972) andFollain and Malpezzi (1981) find that homeowners underestimate their house value by2%, while Kish and Lansing (1954) and Goodman and Ittner (1992) find that homeownersoverestimate their house value by about 4%.
C© 2007 American Real Estate and Urban Economics Association
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
136 Agarwal
correlation between expectations about capital gains in housing wealth and sav-ings. Finally, Case, Quigley and Shiller (2005) using a panel data set find thathousing wealth does impact consumption. Given these inconsistent findings,any strong conclusion is still difficult to make. Hence, in this article I use aunique micro-loan-level data to empirically examine the differential impact ofthe homeowners’ housing wealth underestimation and overestimation on theirconsumption and saving behaviors.
Recently, low mortgage rates fueled many households to “rate refinance” theirmortgage and lower their stream of mortgage payments and increase lifetimewealth.2 In addition, about 45% of households “cash-out refinanced” in 2001–2002 to extract the equity they accumulated in their homes.3 In fact, homeequity grew by $2 trillion between 2001–2003 reaching $7.7 trillion, allowinghomeowners to convert this equity into cash by taking out home equity lines ofcredit (Nothaft 2004). In 2002, homeowners were able to cash out over $100billion; over 61% of the families indicated that they would use the money towardhome improvement or to pay down debt.4
The recent increase in households’ ex ante willingness to either cash-out hous-ing wealth to smooth current consumption or to lower mortgage payments(increase lifetime wealth) provides us with an ideal economic setting to studythe role of rate and cash-out refinancing on homeowners’ misestimation oftheir house values and the impact of such house price misestimations on ex postconsumption and saving behaviors. One way homeowners’ ex ante reveal theirconsumption and saving preferences is through the reason for refinancing theloan (e.g., to lower interest payment, to finance home improvements or to financegeneral consumption).5 Observing this information in the data set, I compare Q1the underestimation and overestimation behaviors of households who cash-outrefinance the perceived additional housing equity in order to increase currentconsumption (i.e., ex ante spenders) to those households who rate refinance to
2 According to Canner, Dynan and Passmore (2002), 52% of households who raterefinanced in 2001 and early 2002 were able to lower their monthly payment due tochanges in interest rates, loan maturities and amounts owned. All else being equal, theaverage rate refinancing households were able to save about $98 in monthly paymentdue to lower interest rates and $135 a month due to increase in maturity.
3 A Wall Street Journal article (Barta 2001) cites several examples including a consumerwho says, “I just didn’t want to let $70,000 sit in my home.” While homeowners couldalso tap the home equity by refinancing the first mortgage, Agarwal, Driscoll and Laibson(2004) point out that there are significant costs to refinancing a first mortgage; on theother hand, there are no costs to originate a home equity line of credit.
4 See Nothaft (2004).
5 Hurst and Stafford (2004) develop a theoretical model of refinancing behavior ofhomeowners who refinance to lower payments and thereby increase their lifetime wealthposition versus those who refinance to extract home equity in order to smooth currentconsumption.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 137
lower interest rates in order to increase lifetime savings (i.e., ex ante savers).6
Homeowners who have lived in the house for a shorter time period are likelyto be overconfident (Schrag and Rabin 1999, Koszegi 2005,Yariv 2005) about
Q2
their house value estimate. And homeowners who have lived in the house for alonger time period are likely to have imperfect knowledge of the true value ofthe house (Agarwal et al. 2006, Gabaix et al. 2006).
Next, I assess the impact of house price misestimation by households on their expost consumption and saving behaviors. I measure the ex post consumption andsaving patterns vis-a-vis changes in the credit line utilization ex post. Finally, themisestimation of the house value by the homeowners may also affect the risksof prepayment and default on their loans. Hence, I also estimate a competingrisk model of home equity credit prepayment and default risks to assess whetherhouse price misestimation by households ex ante can also provide informationabout their prepayment and default behaviors ex post.
I use a unique panel data set of more than 81,000 home equity lines of creditissued to homeowners in 2002 and followed each account’s utilization and per-formance on a monthly basis through 2005. Some of the critical information ob-served in the data set is as follows. At loan origination, homeowners provide thebank with the following information: (1) the reason for loan origination—raterefinance, home improvement or cash-out refinance (e.g., automobile purchase,vacation, etc.), (2) their own estimate of the house value and (3) credit lineamount requested. Other important information observed is the bank’s estimateof the house value, which is based on the Case–Shiller weighted repeat salesindex (see Case and Shiller 1987, 1989, 1990), 7 and the loan amount approved.
6 The behavioral economics literature may classify the homeowners who have lived inthe house for a shorter time period as overconfident (Schrag and Rabin 1999) in theirhouse value estimate, and homeowners who have lived in the house for a longer timeperiod as having imperfect knowledge of the true value of the house (Gabaix, Laibson,Moloche and Weinberg 2006). Also see Koszegi (2005) Yariv (2005) and Agarwal,Driscoll, Gabaix and Laibson (2006).
7 The weighted repeat sales index method was originally proposed by Bailey, Muth andNourse (1963). The Case–Shiller indices control for the changes in property characteris-tics and can pick up turns in price direction. Additionally, the index segment’s price gains Q3by house price tier (low, middle and high). Case and Shiller (2003) use the indexes as ameasure of house price appreciation in comparison to the homeowners’ estimate. Kainand Quigley (1972) also compare the homeowners’ estimate to the appraisals based onthe repeat sales of comparable properties. The bank validated the Case–Shiller indexesto in-person house appraisal and actual house sale prices (for a subset of the houses) andfound, on average, the differences between the indexes, in-person appraisals and the saleprice are statistically insignificant. However, the validation data set is not available forthe purposes of this study. Loebs (2005) also finds an “absence of statistically significantbias” in the Case–Shiller index when compared to the actual sale price. He comparedthe Case–Shiller index to the sale price for over 77,708 properties and found a 0.8%difference between the two. Alternatively, he also compared the Case–Shiller to the fullappraisal for 15,524 properties and found a difference of –3.6%.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
138 Agarwal
For the purposes of this study, I define house price misestimation as a 10%or more difference between the house value estimate of the homeowner andthe homeowner’s financial institution (based on the Case–Shiller index). The10% difference was also used by the financial institution to trigger an in-personappraisal with no cost to the borrower for the appraisal. Hence, I define anunderestimator as a homeowner whose house price estimate is below 10% ofthe bank’s estimate and an overestimator as a homeowner whose house priceestimate is above 10% of the bank’s estimate.
In addition, I also observe in the data set a very rich set of demographic and creditrisk characteristics of the homeowners. Thus, I am able to control for manysocioeconomic factors such as owner’s age, employment type, employmenttenure, income, housing tenure, debt-to-income ratio and credit risk (FICOscore). Finally, I also control for whether or not the owner has a first mortgageas well as the mortgage balance and whether the owner also owns a secondhome or a condo.
Previewing the results, I find that homeowners on average significantly overes-timate their house value by 3.1%, with mean absolute misestimation of 13.1%.8
Consistent with previous studies (e.g., Kish and Lansing 1954, Kain and Quigley1972), I find house price misestimation to be significantly correlated with housetenure, income, borrower credit quality, borrower age, years on the job and em-ployment status. Specifically, homeowners who are less credit worthy, own thehouse more recently, have been at their job longer, are self-employed or arehomemakers are more likely to overvalue their houses, while those who areolder, with higher income, have lower debt-to-income ratio, are more creditworthy, have more years on the job or own the home longer are more likely tounderestimate their house values.
Moreover, I also find that homeowners’ misestimation of their home values ishighly correlated with their ex ante savings and consumption decisions (i.e., thereason for refinancing—rate vs. cash-out). The results show that homeownerswho rate refinance their existing loans (the ex ante savers) are 13.9% more likelyto underestimate their house values, while homeowners who cash-out refinance(the ex ante spenders) are 17.9% more likely to overestimate their house values.
Both underestimators and overestimators of housing wealth requested and re-ceived a walk-in appraisal. Underestimators with an in-person appraisal tend
8 The overestimation is 4.64% if I also include homeowners who were either rejectedby the bank or who turned down the loan. Following Kain and Quigley (1972), I alsoanalyze the homeowners of multi-family houses (condominiums); the results confirmtheir findings that condo owners overestimate their house value by as much as 4.5%.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 139
to be those who have higher income or higher loan-to-value (LTV) ratio, whileoverestimators with an in-person appraisal tend to be those who are relativelyolder or face a higher bank–market APR differential. Equally important, under-estimators with a walk-in appraisal tend to be ex ante savers, perhaps hopingto lower the LTV and thereby to lower APR and current mortgage payments(to increase their lifetime wealth). On the other hand, overestimators with awalk-in appraisal tend to be ex ante spenders, perhaps cashing out additionalhousing wealth to smooth current consumption.9
To study the ex post spending and saving behaviors, I model the credit lineusage behaviors of underestimators and overestimators. My objective is to testwhether underestimators are indeed lowering their credit line usage (i.e., savingex post) and overestimators are increasing their credit line usage (i.e., spendingex post).10 The regression results show that underestimators are 14.9% morelikely to increase their savings ex post, while overestimators are 14.4% morelikely to increase their spending ex post.11
Finally, I find that overestimators, especially those who requested an in-personvaluation from the bank, have a 14% higher risk of defaulting on their loans.It is possible for the default option to be “in-the-money” after the home equityhas been cashed out to smooth current consumption. On the other hand, I findthat underestimators, especially those who requested the bank for an in-personvaluation, have a 10.2% higher risk of prepayment. It may be the case that theinterest rate did not maximize their lifetime savings, leaving the prepaymentoption still “in-the-money.”
I describe the data in the next section and present empirical results in the thirdsection. The fourth section concludes.
Data
The data come from a large financial institution (proprietary in nature) thatoriginates home equity lines of credit. The sample consists of 81,943 credit
9 To test for selectivity bias, I also analyze the behavior of homeowners who rejectedthe bank’s offer for the loan. I find that 29% of the homeowners who rejected the loanoffer were underestimators and 38% were overestimators.10 We know from the previous section that refinancing with no cash out is the primarymotive of the underestimators, and refinancing with cash out is the primary motive ofthe overestimators. However, homeowners could have changed their mind subsequentto originating the line of credit; these results will confirm if underestimators do actuallysave and overestimators do actually spend over the two-year period.
11 Though not reported, I also find that the initial utilization for underestimators washigher and for overestimators was lower.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
140 Agarwal
Table 1 � Distribution by state.
State Percentage
NJ 27.54%MA 21.30%NY 20.36%CT 9.40%PA 5.29%ME 3.42%RI 2.93%NH 2.81%FL 1.53%CA 1.38%Others∗ 4.04%
∗Others include the following states: AZ, CO, DC, DE, GA, IA, ID, IL, IN, KS, KY,LA, MD, MI, MN, MO, NC, NM, NV, OH, OK, OR, SC, TN, UT, VA, WA and WI.
lines issued to owner-occupants from March 2002 to December 2002; eachaccount’s utilization and performance (default and prepayment) was observedthrough January 2005. These loans are typical credit lines that are open for thefirst five years, during which time the borrower is only required to make interestpayments on the utilized line balance. After the fifth year, the line is closed andis converted to a fixed-rate term loan with a remaining term of five to 15 years.At this point, the borrower is required to make fixed monthly payments ofprincipal and interest for the remaining period of the line. Consistent with othermortgage loans, the borrower may prepay or default on the line at any time. Forour purposes, all credit lines have at least 24 months of performance data.
The majority of the credit lines in my sample has originated in eight Northeasternstates (see Table 1); however, 1.53% originated in Florida, 1.38% in Californiaand 4.04% in 28 other states. Table 2 reports the descriptive statistics for thelines at origination. The descriptive statistics are segmented into five categories:(1) overall sample, (2) underestimators, (3) underestimators with additionalwalk-in appraisal, (4) overestimators and (5) overestimators with additionalwalk-in appraisal. Once again, underestimators (overestimators) are homeown-ers with house price estimate below (above) 10% of the bank’s estimate.12
Below I describe the summary statistics in Table 2 for our entire sample as wellas separately for the underestimators and overestimators.
12 I choose the 10% differential because the bank uses the same criterion to approvethe loan without review. I tried alternative segments at 5% and 15%; the results arequalitatively the same. This is also consistent with Kain and Quigley (1972).
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 141
All Accounts
In Table 2, the summary statistics for the overall sample (all accounts) indicatethat homeowners overestimate their house value by 3.1% on average. The bankprovides a loan of 1% less than the borrower-requested amount. The average
Table 2 � Descriptive statistics by underestimation and overestimators and appraisal type forhome equity lines of credit.
Underestimators Overestimators
All Walk-in Walk-in WithinAccounts All Appraisal All Appraisal 10%
Homeowner estimate $346,065 $302,143 $306,000 $387,363 $398,492 $333,322Bank estimate $335,797 $344,289 $352,800 $347,758 $356,340 $326,105Homeowner price 3.1% −12.2% −13.3% 11.4% 11.8% 2.2%
misestimationLoan requested by $61,347 $52,700 $54,564 $68,718 $71,168 $59,140
borrowerLoan approved by bank $60,725 $56,425 $56,892 $64,019 $69,160 $59,262Borrower loan request 1.0% −6.6% −4.1% 7.3% 2.9% −0.2%
misestimationFirst mortgage balance $154,444 $116,406 $149,452 $177,914 $198,841 $146,871Customer LTV 62 55 66 63 67 61Appraised LTV 64 50 59 69 74 63Months at address 99 140 123 97 90 93No first mortgage 15% 26% 28% 16% 18% 12%Second home 3% 1% 3% 4% 5% 3%Condo 6% 2% 14% 1% 10% 8%Rate refinancing 39% 41% 44% 31% 29% 39%Home improvement 25% 25% 25% 25% 26% 26%Cash-out refinance 35% 34% 31% 44% 45% 35%Borrower age 46 47 46 41 41 47Self-employed 7.76% 5.90% 6.90% 8.40% 10.00% 7.50%Retired 7.74% 7.00% 6.40% 9.50% 7.30% 6.50%Home maker 1.31% 1.00% 1.00% 1.60% 1.20% 1.30%Employed 83.19% 86.10% 85.70% 80.50% 81.50% 84.70%Years on the last job 7.62 8.93 7.55 8.52 7.32 7.29Income $90,293 $94,452 $95,345 $82,718 $83,480 $93,051DTI 41 38 40 41 44 42APR 4.60% 4.31% 4.22% 4.71% 4.87% 4.59%FICO 733 741 744 739 721 731Account balance $33,848 $50,039 $53,407 $35,543 $38,754 $23,521Account balance $36,727 $40,271 $44,615 $43,485 $45,674 $31,950
avg 2 yrsFrequency 81,943 8,845 2,021 17,125 6,901 47,051Percentage dist 100% 11% 2% 21% 8% 57%Percentage prepayment 26% 35% 21% 19% 18% 27%Percentage default 0.62% 0.40% 0.35% 0.77% 1.20% 0.53%
Notes: The data cover home equity originations from March 2002 to December 2002. All thestatistics are reported at account origination. Except for Account Balance Avg (average accountbalance over the performance period), default and prepayment rates. They are reported over theentire performance window.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
142 Agarwal
borrower FICO score is 733 and the average interest rate is 4.6%.13 The ap-praised average LTV ratio at origination (calculated as total debt (bank approvedcredit line plus first-mortgage debt) divided by appraised house value) is 64%;about 15% of our sample has no first mortgage. Of the total borrowers request-ing for a second mortgage line of credit, about 39% rate refinance their existingloan without cashing out the home equity, 25% cash out the home equity forhome improvement and the remaining 35% cash-out refinance for general con-sumption purposes. These observations are consistent with the survey findingsby Nothaft (2004).
The demographics are as follows. The average age of the homeowners is about46 years, and the average house tenure is slightly over eight years. About 7%are self-employed, 7% are retired and 1.2% are homemakers. The homeownershave on average 7.6 years on the last job and $90,293 family income. Andabout 45% of the homeowners allow automatic payment of the credit line tothe financial institution from their deposit (checking/savings) accounts.
Underestimators versus Overestimators
When I compare the summary statistics for homeowners who underestimatetheir house value and those who overestimate their house value in Table 2,I observe some very interesting differences between the two groups. About11% of homeowners undervalue their house, while about 21% of homeownersovervalue their house. On average, underestimators undervalue their house byabout 12%, while overestimators overvalue their house by about 11%. In turn,underestimators requested for a lower loan amount than the bank is willingto lend by 6.6% on average, while overestimators requested for a higher loanamount than the bank is approving by 7.3%.
About 41% of the underestimators rate refinance without cashing out the equityin the house, while only about 31% of the overestimators rate refinance withoutcashing out the equity. On the other hand, 34% of the underestimators refinanceto cash out the equity to finance general consumption, while 44% of the over-estimators do so. About 25% of both underestimators and overestimators cashout to finance home improvement.
In addition, about 26% (16%) of underestimators (overestimators) do not havea first mortgage. For homeowners with a first mortgage, underestimators (over-estimators) have a first mortgage balance of $166,406 ($177,914). Furthermore,underestimators have lived at the current address for almost 12 years on average,
13 Borrower credit scores are provided by Fair, Isaac and Company (FICO). Higherscores indicate higher credit quality.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 143
compared to eight years for overestimators. Underestimators, on average, tendto be significantly older (47 years old on average) than overestimators (41 yearsold on average). Underestimators earn relatively higher income (almost $12,000higher) and have slightly lower debt-to-income ratio (three percentage pointslower) than overestimators.
Walk-in Appraisals: Underestimators versus Overestimators
If the owner’s estimate is more than 10% below that of the bank’s estimate,the owner has the incentive to and can request the bank to conduct a walk-inappraisal on the house. I define these homeowners as underestimators with ad-ditional walk-in appraisal. If the owner’s estimate is more than 10% abovethe bank’s estimate, the bank conducts a walk-in appraisal. I define thesehomeowners as overestimators with additional walk-in appraisal.
The overall house price misestimation by the underestimators with walk-in ap-praisals is −13.3%, compared to the −12.2% for all underestimators. The mis-estimation by the overestimators with walk-in-appraisals is 11.8%, compared tothe 11.4% for all overestimators. These observations are consistent with under-estimators having a relatively longer tenure at their house than the overestima-tors; therefore the underestimators may not be as informed and knowledgeableabout the housing market. As a result, it is beneficial for the underestimators torequest a walk-in appraisal. Overall, the loan amount approved by the bank ison average 4.1% more than that requested by the underestimators with a walk-inappraisal, and on average 2.9% less than that requested by the overestimatorswith a walk-in appraisal.
Results
The result section is divided into three main parts: the first subsection focuseson the impact of ex ante consumption and saving expectation on the likelihoodof homeowners to underestimate and overestimate their house value; the secondsubsection presents the impact of house price underestimation and overestima-tion on the ex post consumption and saving patterns; and the third subsectionpresents the effect of house price underestimation and overestimation on theprepayment and default risks.
Ex Ante Consumption and Saving Behaviors
I estimate a multinomial logit model to assess the impact of an ex ante consump-tion or saving decision on the likelihood of a homeowner underestimating andoverestimating his house value. The model specification treats underestimationand overestimation as competing options. The results are reported in Table 3.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
144 AgarwalTa
ble
3�
Det
erm
inan
tso
fu
nd
eres
tim
atio
nan
dover
esti
mat
ion
ath
om
eeq
uit
yli
ne
ori
gin
atio
n.
Un
der
esti
mat
ion
Over
esti
mat
ion
Var
iab
leC
oef
f.S
td.E
rr.
t-S
tat.
Mar
gin
alC
oef
f.S
td.E
rr.
t-S
tat.
Mar
gin
al
Inte
rcep
t−0
.87
86
0.1
97
2−4
.46
−0.7
87
00
.14
01
−5.6
2B
orr
ower
age
0.0
09
30
.00
12
7.8
50
.58
%0
.00
72
0.0
00
98
.11
0.8
3%
Ret
ired
−0.2
20
50
.04
38
−5.0
3−1
.87
%0
.01
71
0.0
35
10
.49
1.8
1%
Ho
me
mak
er−0
.23
44
0.1
02
7−2
.28
−1.7
6%
0.1
91
00
.06
60
2.8
93
.76
%S
elf-
emp
loy
ed−0
.07
93
0.0
42
7−1
.86
−2.8
3%
0.1
22
70
.02
85
4.3
14
.07
%Y
ears
on
the
job
−0.0
06
30
.00
13
−4.9
6−0
.54
%0
.00
25
0.0
01
02
.60
0.2
5%
Inco
me
0.0
00
00
.00
00
4.9
40
.01
%0
.00
00
0.0
00
01
.19
0.0
1%
No
firs
tm
ort
gag
e0
.21
44
0.0
29
07
.39
4.2
8%
0.2
29
80
.02
20
10
.45
4.5
5%
LT
V−0
.17
23
0.0
32
3−5
.33
−0.4
6%
0.0
77
60
.01
41
5.5
00
.21
%Y
ears
ho
me
own
ed0
.00
15
0.0
00
11
5.3
00
.60
%−0
.00
07
0.0
00
1−9
.15
−0.2
6%
Sec
on
dh
om
e−0
.27
15
0.0
68
1−3
.99
−2.5
2%
−0.4
37
30
.15
31
−2.8
6−1
.17
%C
on
do
−0.0
57
00
.04
94
−1.1
5−2
.15
%−0
.06
69
0.0
37
8−1
.77
3.8
5%
Rat
ere
fin
anci
ng
0.1
44
10
.02
60
5.5
41
3.9
0%
−0.1
00
90
.01
91
−5.2
8−3
.09
%C
ash
-ou
tre
fin
ance
0.1
19
90
.18
80
0.6
40
.47
%0
.00
62
0.0
01
15
.70
17
.92
%L
oan
amo
un
tre
qu
este
d0
.00
00
0.0
00
0−1
6.4
7−0
.01
%0
.00
00
0.0
00
09
.39
0.0
6%
FIC
O0
.00
09
0.0
00
23
.86
0.2
0%
−0.0
00
50
.00
02
−2.6
9−0
.26
%D
TI
−0.0
05
70
.00
07
−7.8
3−0
.62
%0
.00
04
0.0
00
50
.97
0.3
8%
ZIP
cod
ed
um
mie
sY
esM
on
tho
rig
inat
ion
du
mm
ies
Yes
Nu
mb
ero
fu
nd
er/o
ver
est.
10
,86
62
4,0
26
Nu
mb
ero
fO
bs.
81
,94
3L
og
likel
iho
od
5,6
31
Pse
ud
oR
-sq
uar
e0
.47
Not
es:
Est
imat
ea
mu
ltin
om
ial
log
itm
od
elo
fh
om
eow
ner
s’u
nd
eres
tim
atio
nan
dover
esti
mat
ion
of
the
ho
use
valu
eat
loan
ori
gin
atio
n.U
nd
eres
tim
atio
nan
dover
esti
mat
ion
isd
efin
edas
ag
reat
erth
an1
0%
dif
fere
nce
inth
eh
ou
seva
lue
esti
mat
eo
fth
eh
om
eow
ner
inco
mp
aris
on
toth
eC
ash
–S
hil
ler
ind
exes
.A
fter
con
tro
llin
gfo
rso
cio
eco
no
mic
char
acte
rist
ics,
Ite
stif
exan
teco
nsu
mp
tio
nan
dsa
vin
g(r
efin
ance
vs.
con
sum
pti
on
)ar
ea
det
erm
inan
to
fm
ises
tim
atio
n.T
he
esti
mat
esar
ed
eriv
edw
ith
het
ero
sked
asti
city
-co
rrec
ted
stan
dar
der
rors
.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 145
With respect to socioeconomic characteristics, I find that on average an olderhomeowner is more likely to misestimate (both underestimate and overestimate)his house value. Furthermore, a homeowner who does not have a first mortgageis more likely to both underestimate as well as overestimate his house value,while one who has a second home or a condo is less likely to both underestimateand overestimate the price of his home.
Moreover, I find that a borrower who is a homemaker or is retired is less likelyto underestimate his house value. A homeowner with high LTV or lower creditquality is less likely to underestimate (instead is more likely to overestimate)his house value. A homeowner with longer home tenure or has higher incomeis more likely to underestimate (instead is less likely to overestimate) his housevalue. These results are consistent with the findings of Kain and Quigley (1972).
Finally, a homeowner who indicates that he or she intends to use the funds forgeneral consumption purposes (i.e., ex ante spenders) is almost 17.9% morelikely to overestimate, while a homeowner who indicates that he or she intendsto use the funds to rate refinance the existing mortgage (i.e., ex ante savers)is almost 13.9% more likely to underestimate his house value. Moreover, ahomeowner who requests for a larger loan amount is less likely to underestimate(instead he is more likely to overestimate) the house value. These results indicatethat homeowners who are ex ante spenders are more likely to overestimate theirhouse values, while those who are ex ante savers more likely to underestimatetheir house values.14
The Likelihood of a Walk-in Appraisal
Next, I look at a subset of homeowners who, after reviewing the bank’s loanamount and contract rate offer, request the bank to conduct a walk-in appraisal oftheir house. I estimate a logit model of the likelihood that an underestimator oroverestimator requesting and receiving a walk-in appraisal. I explicitly controlfor the difference in the bank’s interest rate and the average interest rate in themarket for a home equity line of credit (APR differential).15 If a homeowner
14 I also test for selectivity bias and analyze the behavior of homeowners who rejectthe bank’s offer for the loan. I find that 18% of the homeowners who rejected the loanwere underestimators and 25% were overestimators. The regression results confirmthat underestimators who rejected the loan were highly sensitive to the interest rates(i.e., trying to maximize their lifetime savings) and the overestimators who rejected theloan were highly sensitive to the loan amount approved (i.e., trying to maximize theirlifetime or smooth current consumption). These results provide additional evidence thathomeowners’ misestimation of house value does provide valuable information abouttheir savings and consumption behaviors.
15 Current period average home equity line interest rates were obtained from the HeitmanGroup (http://www.heitman.com).
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
146 Agarwal
is sensitive to interest rate, he or she will compare the bank rate to the marketrate.
The results (in Table 4) indicate that higher APR differential increases thelikelihood of an underestimator (but not an overestimator) requesting a walk-in-appraisal by 8.2%. On the other hand, greater loan differential increases thelikelihood of an overestimator (not an underestimator) requesting and receiv-ing a walk-in appraisal by 8.6%. Moreover, an underestimator who is an exante saver (who rate refinances without cashing out in order to lower interestpayment) is 10.2% more likely to receive a walk-in appraisal, while an overesti-mator who is an ex ante spender (who cash-out refinances to fund consumptions)is 12.3% more likely to receive a walk-in appraisal.
Ex Post Savings and Consumption Behaviors
Thus far I have shown that ex ante consumption and saving decisions by home-owners impact their house price misestimations. Now I want to see whetherhouse price misestimations by the homeowners affect their ex post consump-tion and saving behaviors. To quantify a homeowner’s ex post spending andsaving behaviors, I first construct a variable to measure the homeowner’sutilization of the home equity line of credit over the two-year period. Specifi-cally, I define ex post spending as a 10% increase in utilization over a two-yearperiod and ex post saving as a 10% decrease in utilization over a two-yearperiod.
Table 5 presents preliminary evidence of ex post utilization behaviors. Gener-ally, I find that ex ante savers tend to reduce their spending ex post, while ex antespenders tend to increase their spending ex post. Specifically, I find that 14.6%of the ex ante savers actually lowered their credit line utilization ex post, whileonly 7% of them increased their spending ex post. In contrast, only 4.3% of exante spenders lowered their credit line utilization ex post, while about 16.5%of them increased their spending ex post. The small percentage of homeownerswho intended to save (spend) but find themselves spending (saving) may bethose who faced ex post negative (positive) income shocks, respectively (seeAgarwal et al. (2006), who find that a small percentage of credit card borrowersex post revolve higher debt than expected).
Table 6 estimates a multinomial logit model to determine the likelihood of anunderestimator decreasing his or her utilization of the home equity line of credit(i.e., saving ex post) and an overestimator increasing his or her utilization ofthe home equity line of credit (i.e., spending ex post). After controlling for allthe variables in previous estimates, the results show that an underestimator is14.9% more likely to pay down the home equity account balance (i.e., increase
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 147
Tabl
e4
�D
eter
min
ants
of
wal
k-i
nap
pra
isal
.
Un
der
esti
mat
ors
Over
esti
mat
ors
Var
iab
leC
oef
f.S
td.E
rr.
t-S
tat.
Mar
gin
alC
oef
f.S
td.E
rr.
t-S
tat.
Mar
gin
al
Inte
rcep
t−1
.22
94
0.5
02
4−2
.45
−0.3
07
50
.29
26
−1.0
5B
orr
ower
age
−0.0
00
40
.00
28
−0.1
6−0
.25
%0
.00
51
0.0
01
72
.95
0.1
4%
Ret
ired
0.0
64
60
.10
57
0.6
12
.08
%0
.11
21
0.0
68
71
.63
0.4
3%
Ho
me
mak
er0
.18
79
0.2
46
40
.76
1.0
9%
−0.0
72
50
.13
44
−0.5
4−0
.49
%S
elf-
emp
loy
ed0
.43
41
0.0
94
14
.61
2.3
4%
0.2
52
20
.05
37
4.7
01
.91
%Y
ears
on
the
job
0.0
04
00
.00
30
1.3
50
.52
%0
.00
15
0.0
01
90
.80
0.2
9%
Inco
me
0.0
00
00
.00
00
2.7
30
.00
%0
.00
00
0.0
00
00
.69
0.0
1%
No
firs
tm
ort
gag
e0
.31
90
0.0
64
64
.94
4.7
2%
0.2
81
70
.04
18
6.7
44
.06
%L
TV
0.4
06
20
.08
64
4.7
00
.72
%1
.18
90
0.9
38
41
.27
0.6
2%
Yea
rsh
om
eow
ned
0.0
01
50
.00
02
6.1
10
.25
%−0
.00
16
0.0
00
2−1
0.0
0−0
.14
%S
eco
nd
ho
me
0.3
64
70
.13
25
2.7
51
.37
%0
.33
71
0.1
04
33
.23
2.6
9%
Co
nd
o0
.22
17
0.0
95
52
.32
2.2
5%
0.1
95
10
.07
64
2.5
53
.17
%R
ate
refi
nan
cin
g0
.14
13
0.0
63
42
.23
10
.22
%−0
.00
54
0.0
38
2−0
.14
−1.2
4%
Cas
h-o
ut
refi
nan
ce0
.10
95
0.0
70
41
.56
0.6
8%
0.0
57
10
.02
16
2.6
41
2.3
8%
Lo
anam
ou
nt
dif
fere
nce
0.0
00
60
.00
07
0.7
62
.82
%0
.00
39
0.0
00
58
.27
8.5
8%
FIC
O−0
.00
15
0.0
00
6−2
.70
−0.5
1%
−0.0
03
80
.00
03
−11
.02
−0.2
8%
DT
I0
.00
46
0.0
01
53
.01
0.5
1%
0.0
04
80
.00
09
5.5
80
.38
%A
PR
dif
fere
nti
al0
.00
62
0.0
01
54
.11
8.2
3%
0.0
05
50
.02
73
0.2
00
.24
%Z
IPco
de
du
mm
ies
Yes
Yes
Mo
nth
ori
gin
atio
nd
um
mie
sY
esY
esN
um
ber
of
wal
k-i
nap
pra
isal
2,0
21
6,9
01
Nu
mb
ero
fO
bs.
10
,86
62
4,0
26
Lo
gli
kel
iho
od
97
34
,83
6P
seu
do
R-s
qu
are
0.3
40
.58
Not
es:
Est
imat
ea
log
itm
od
elo
fu
nd
eres
tim
atio
nw
ho
req
ues
tan
in-p
erso
nap
pra
isal
(to
tho
sew
ho
do
no
tre
qu
est
anin
-per
son
app
rais
al)
of
the
ho
use
valu
eat
loan
ori
gin
atio
n.
Aft
erco
ntr
oll
ing
for
soci
oec
on
om
icch
arac
teri
stic
s,I
test
ifth
ein
tere
stra
ted
iffe
ren
tial
(in
tere
stra
teo
ffer
ed–
aver
age
mar
ket
inte
rest
rate
)le
adu
nd
eres
tim
ato
rso
rover
esti
mat
ors
tore
qu
esta
nd
rece
ive
aw
alk
-in
app
rais
al.T
he
esti
mat
esar
ed
eriv
edw
ith
het
ero
sked
asti
city
-co
rrec
ted
stan
dar
der
rors
.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
148 Agarwal
Table 5 � Ex ante and ex post distribution of savings and consumption.
Refi. with Refi. withHome Cash Out No Cash Out
Ex Post\Ex Ante Improvement (Consumption) (Saving) Total
Less than 10% � Utilization 10.73 14.7 17.55 42.98Saving 12.39 4.29 14.56 31.24Spending 2.22 16.46 7.1 25.78
Total 25.34 35.45 39.21 100
Notes: Distribution of ex ante reasons for the loan (home improvement, consumptionand refinancing) and ex post utilization of the line of credit. Spending is defined as anincrease in the utilization over a two years period by more than 10%, and saving isdefined as a decrease in utilization by more than 10%.
saving ex post), while an overestimator is almost 14.4% more likely to increasespending ex post.
Prepayment and Default Behaviors
Finally, I want to test whether a homeowner’s house price underestimation andoverestimation influence his or her loan prepayment and default patterns. Tothis end, I estimate a Cox proportional hazard model with competing risks andtime-varying covariates to determine if the underestimation and overestimationof the home value by the owner influence prepayment and default decisions,after controlling for the traditional variables as predicted by the option theory.16
Table 7 presents the results. Most of the signs on the traditional variables areconsistent with past studies of home equity lines prepayment and default be-havior (e.g., Agarwal et al. 2006). Specifically, I find that higher interest ratedifferential significantly increases the likelihood of a homeowner prepayinghome equity.17 A borrower with lower credit score, higher LTV or higher loanbalance is more likely to default.
16 See, Deng, Quigley and Van Order (2000) and Agarwal, Ambrose and Liu (2006) andthe references therein.17 I approximate the interest rate differential (prepayment option) as outlined in Deng,Quigley and Van Order (2000)
OPTIONi,t = Vi,t − V ∗i,t
Vi,t
,
where Vi,t is the market value of loan i at time t (i.e., the present value of the remainingmortgage payments at the current market mortgage rate) and V ∗
i,t is the book value ofloan i at time t (i.e., the present value of the remaining mortgage payments at the contractinterest rate).
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 149
Tabl
e6
�H
ou
seh
old
su
tili
zati
on
beh
avio
ro
fth
eir
cred
itli
ne.
Ex
Post
Sav
ing
sE
xPo
stC
on
sum
pti
on
Var
iab
leC
oef
f.S
td.
Err
.t-
Sta
t.M
arg
inal
Co
eff.
Std
.E
rr.
t-S
tat.
Mar
gin
al
Inte
rcep
t1
.92
88
40
.18
94
61
0.1
8−0
.70
23
80
.28
93
8−2
.43
Bo
rrow
erag
e0
.03
43
60
.00
49
36
.96
0.3
6%
0.0
09
23
0.0
05
35
1.7
30
.30
%B
orr
ower
age(
sq)
−0.0
00
29
0.0
00
04
−7.1
70
.00
%0
.00
02
80
.00
02
21
.27
0.0
0%
Ret
ired
−0.1
34
90
0.0
35
65
−3.7
8−1
.69
%0
.14
35
90
.05
88
02
.44
1.4
1%
Ho
me
mak
er0
.04
32
50
.06
93
30
.62
0.8
3%
0.0
39
85
0.0
68
33
0.5
80
.69
%S
elf-
emp
loy
ed0
.29
23
80
.03
94
47
.41
2.3
0%
0.2
89
05
0.0
59
33
4.8
71
.93
%Y
ears
on
the
job
0.0
00
29
0.0
00
14
2.1
00
.01
%0
.00
50
90
.00
23
52
.17
0.0
1%
Inco
me
0.0
00
00
0.0
00
00
7.1
50
.00
%0
.00
00
00
.00
00
09
.23
0.0
0%
Inco
me(
sq)
0.0
00
00
0.0
00
00
17
.74
0.0
0%
0.0
00
00
0.0
00
00
−10
.00
0.0
0%
No
firs
tm
ort
gag
e0.0
0634
0.0
2944
0.2
23.7
0%
0.4
3890
0.0
5898
7.4
43.1
0%
LT
V0
.38
98
30
.05
09
37
.65
0.4
2%
0.1
58
48
0.0
38
97
4.0
70
.28
%Y
ears
ho
me
own
ed0
.00
00
70
.00
00
90
.79
0.0
5%
−0.0
00
33
0.0
00
18
−1.8
1−0
.01
%S
eco
nd
ho
me
−0.0
40
97
0.0
59
33
−0.6
9−0
.23
%−0
.07
23
90
.08
43
9−0
.86
−0.2
0%
Co
nd
o−0
.12
39
90
.03
38
6−3
.66
−3.3
0%
0.0
03
83
0.0
45
83
0.0
8−0
.25
%U
nd
eres
tim
ato
r0
.01
94
60
.00
29
46
.62
14
.93
%−0
.10
43
80
.06
48
9−1
.61
−0.9
4%
Un
der
esti
mat
or
wit
hw
alk
-in
0.0
20
44
0.0
02
92
6.9
91
5.9
5%
0.1
29
85
0.1
89
35
0.6
90
.79
%O
ver
esti
mat
or
0.0
40
95
0.0
39
33
1.0
41
.18
%0
.00
38
90
.00
18
42
.11
14
.49
%O
ver
esti
mat
or
wit
hw
alk
-in
0.0
13
50
0.0
57
48
0.2
30
.52
%0
.10
34
80
.03
61
02
.87
15
.43
%F
ICO
0.0
01
24
0.0
00
26
4.7
9−0
.29
%0
.00
00
70
.00
02
40
.29
−0.1
1%
AP
R−0
.33
29
00
.09
05
0−3
.68
−4.2
6%
0.2
30
95
0.3
98
35
0.5
8−3
.57
%D
TI
0.0
07
93
0.0
53
41
0.1
50
.15
%0
.01
48
40
.00
42
43
.50
0.1
3%
Au
toP
ay0
.06
87
20
.01
73
53
.96
5.2
8%
−0.0
34
85
0.0
39
83
−0.8
7−0
.24
%Z
IPco
de
du
mm
ies
Yes
Mo
nth
ori
gin
atio
nd
um
mie
sY
esN
um
ber
of
saver
s/co
nsu
mer
s2
5,4
02
20
,48
5N
um
ber
of
Ob
s.8
1,9
43
Lo
gli
kel
iho
od
4,1
63
Pse
ud
oR
-Sq
uar
e0
.48
Not
es:
Est
imat
ea
mu
ltin
om
ial
log
itm
od
elo
fex
post
sav
ing
and
spen
din
g,
wh
ere
expo
stsa
vin
g(s
pen
din
g)
isd
efin
edas
10
%in
crea
se(d
ecre
ase)
incr
edit
lin
eu
tili
zati
on
over
atw
oy
ears
per
iod
.T
he
esti
mat
esar
ed
eriv
edw
ith
het
ero
sked
asti
city
-co
rrec
ted
stan
dar
der
rors
.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
150 AgarwalTa
ble
7�
Det
erm
inan
tsof
def
ault
and
pre
pay
men
tfo
rhom
eeq
uit
yli
nes
of
cred
it.
Def
ault
Pre
pay
men
t
Var
iable
Coef
f.S
td.E
rr.
t-S
tat.
Mar
gin
alC
oef
f.S
td.E
rr.
t-S
tat.
Mar
gin
al
Inte
rcep
t11.8
983
0.8
490
14.0
12.7
439
0.2
200
12.4
7B
orr
ower
age
−0.0
748
0.0
194
−3.8
6−0
.01%
−0.0
456
0.0
049
−9.3
6−0
.69%
Borr
ower
age(
sq)
0.0
008
0.0
002
4.1
90.0
0%
0.0
003
0.0
000
6.9
80.0
1%
Ret
ired
0.3
716
0.2
481
1.5
00.5
6%
0.2
280
0.0
428
5.3
23.4
6%
Hom
em
aker
0.4
328
0.3
792
1.1
40.3
4%
0.0
118
0.0
818
0.1
40.2
1%
Sel
f-em
plo
yed
0.5
263
0.1
594
3.3
00.3
1%
−0.1
448
0.0
363
−3.9
9−1
.89%
Yea
rson
the
job
0.0
044
0.0
069
0.6
30.0
2%
0.0
008
0.0
011
0.6
80.0
1%
Inco
me
0.0
000
0.0
000
−3.0
10.0
0%
0.0
000
0.0
000
−5.8
90.0
0%
Inco
me(
sq)
0.0
000
0.0
000
2.3
30.0
0%
0.0
000
0.0
000
6.0
10.0
0%
No
firs
tm
ort
gag
e−0
.1944
0.1
645
−1.1
8−0
.16%
−0.6
708
0.0
275
−24.3
5−6
.99%
LT
V0.2
586
0.1
197
2.1
60.2
1%
−0.0
872
0.0
237
−3.6
7−0
.31%
Yea
rshom
eow
ned
−0.0
002
0.0
001
−1.9
8−0
.01%
0.0
002
0.0
001
2.3
60.3
2%
Sec
ond
hom
e0.0
938
0.3
043
0.3
10.1
7%
0.0
755
0.0
620
1.2
21.0
8%
Condo
−0.2
911
0.2
306
−1.2
6−0
.62%
0.0
698
0.0
289
2.4
21.9
4%
Under
esti
mat
e−0
.0664
0.1
983
−0.3
3−0
.02%
0.1
015
0.0
325
3.1
27.5
1%
Under
esti
mat
eU
pgra
de
−0.3
854
0.2
411
−1.6
0−0
.04%
0.0
821
0.0
237
3.4
610.2
7%
Over
esti
mat
e0.0
863
0.0
381
2.2
710.0
2%
0.0
409
0.0
396
1.0
31.2
3%
Over
esti
mat
eU
pgra
de
0.1
856
0.0
418
4.4
414.0
1%
0.0
206
0.0
543
0.3
80.5
7%
FIC
O−0
.0258
0.0
010
−25.4
9−0
.90%
−0.0
050
0.0
002
−23.9
5−0
.74%
PP
Opti
on
0.4
482
0.0
768
5.8
40.3
6%
0.1
389
0.0
168
8.2
66.2
1%
DT
I0.0
075
0.0
032
2.3
70.1
1%
−0.0
005
0.0
006
−0.9
4−0
.07%
Auto
Pay
−1.5
012
0.1
406
−10.6
8−7
.07%
−0.2
626
0.0
201
−13.0
4−4
.06%
Acc
ount
bal
ance
0.0
000
0.0
000
4.3
30.0
0%
0.0
000
0.0
000
2.5
10.0
0%
ZIP
code
dum
mie
sY
esM
onth
ori
gin
atio
ndum
mie
sY
esT
ime
dum
mie
sY
esN
um
ber
of
def
ault
/pre
pay
507
20,9
24
Num
ber
of
acco
unts
81,9
43
Log
likel
ihood
15,7
17
Pse
udo
R-S
quar
e0.5
3
Not
es:
This
table
show
sre
sult
sof
apro
port
ional
haz
ard
model
of
pre
pay
men
tan
ddef
ault
usi
ng
month
lydat
efo
rhom
eeq
uit
yli
nes
of
cred
itfr
om
Mar
ch2002
toM
arch
2005.
Pre
pay
men
tis
defi
ned
asac
tual
pay
men
tof
the
loan
amount
pri
or
toco
ntr
act
term
san
ddef
ault
isdefi
ned
as90
day
spas
tdue.
The
indep
enden
tva
riab
les
contr
ol
for
loan
ori
gin
atio
nm
onth
,ca
lendar
tim
e,st
ate
dum
mie
s,cr
edit
risk
,cu
rren
tlo
an-t
o-v
alue
rati
o,
pre
pay
men
topti
on,
vari
ous
dem
ogra
phic
vari
able
s(a
ge,
inco
me,
occ
upat
ion,
etc.
)an
dth
ere
asons
for
the
loan
(refi
nan
cevs.
consu
mpti
on).
All
tim
eva
ryin
gva
riab
les
are
lagged
by
six
month
sto
avoid
any
endogen
iety
.T
he
com
pet
ing
risk
sm
odel
ises
tim
ated
asa
mult
inom
ial
logit
via
max
imum
likel
ihood.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 151
Other socioeconomic variables are also significant determinants of default andprepayment. For example, a homeowner with higher LTV is less likely to prepayand more likely to default, whereas a self-employed homeowner is more likelyto default on the loan and less likely to prepay, while a retired homeowner ismore likely to prepay. These results are also consistent with those of Agarwal,Chomsisengphet and Hassler (2005).
Next, I turn my attention to the variables of interest. I find that on averagean overestimator is 10% more likely to default, while an underestimator is7.5% more likely to prepay on his or her loan. The results suggest that anunderestimator is more likely to prepay, and an overestimator is more likelyto default on his or her loan. These results provide a new perspective to theextensive prepayment and default literature.
Conclusion
A number of studies have pointed out that homeowners either underestimate oroverestimate their house value between 2% and 4%. Furthermore, homeowners’misestimation of the house value could lead to errors in household consumptionand savings decisions because of their perceived (vs. actual) housing wealth.In this article, with the help of a unique proprietary panel data set from a largefinancial institution of more than 81,000 home equity lines of credit issued tohomeowners in 2002 and followed through 2005, I assess how ex ante sav-ing/consumption decisions affect homeowners’ misestimation of their housevalue as well as the impact of such house price misestimation on households’consumption and saving behaviors ex post. In addition, I also look at the impactof house price misestimation on the risks of homeowners to prepay and defaulton their home equity lines of credit.
The results are consistent with the previous studies: homeowners on averageoverestimate their house value by 3.1%, with mean absolute misestimation of13.1%. I find that house price misestimation is highly correlated with home-owners’ ex ante consumption and saving decisions. Specifically, homeownerswho take out a loan to rate refinance their existing loan without cashing out theequity are almost 13.9% more likely to underestimate their house value. On theother hand, homeowners who cash-out refinance to extract their housing equityto fund consumption are almost 17.9% more likely to overestimate the value oftheir homes.
Among the homeowners who underestimate, those that are more likely to re-quest and receive an in-person appraisal tend to be those who have higherincome or higher LTV, while overestimators requesting an in-person appraisaltend to be those who are relatively older or face a higher bank-market APR
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
152 Agarwal
differential. Equally important, underestimators with a walk-in appraisal tendto be ex ante savers, perhaps hoping to lower the LTV and thus to lower APRand current mortgage payments (to increase their lifetime wealth). On the otherhand, overestimators with a walk-in appraisal tend to be ex ante spenders,cashing out additional housing wealth in order to smooth current consump-tion. These results provide additional support that homeowners tend to under-estimate their house value in order to perhaps lower their current mortgagepayments and increase their lifetime wealth, while homeowners tend to over-estimate their house value when cashing out home equity to smooth currentconsumption.
In addition, I model the credit line usage behaviors of underestimators andoverestimators. The objective here is to test whether underestimators are indeedreducing their credit lines (i.e., ex post saving) and overestimators are indeedincreasing their credit lines (i.e., ex post spending). The regression results showthat underestimators are 14.9% more likely to increase saving ex post, whileoverestimators are almost 14.4% more likely to increase spending ex post. Theseresults support the findings of Case, Quigley and Shiller (2005), but they go onestep further to show a differential impact of the homeowners’ housing wealthmisestimations on their consumption and saving decisions.
Finally, I estimate a competing risks model of home equity line default and pre-payment to assess whether house price misestimation by the borrowers can alsoprovide information about their prepayment and default risks. I find that over-estimators, especially those who requested the bank for an in-person valuation,have a 14% higher risk of default. On the other hand, I find that underestimators,especially those who requested the bank for a walk-in appraisal, have a 10.2%higher risk of prepayment.
I would like to thank Brent Ambrose, Souphala Chomsisengphet, Bert Higgins, DavidLaibson, Chunlin Liu, Donna Nicholson, Seow Eng Ong, Nick Souleles, Tony Yezer,Crocker Liu (editor), two anonymous referees and seminar participants at the annualAmerican Real Estate Economics and Financial Association meetings for helpful dis-cussions and comments. The opinions expressed in this research are those of the authorand do not necessarily reflect the opinion of the Federal Reserve Bank of Chicago orthe Federal Reserve System.
References
Agarwal, S., B.W. Ambrose, S. Chomsisengphet and C. Liu. 2006. An Empirical Anal-ysis of Home Equity Loan and Line Performance. Journal of Financial Intermediation15(4): 444–469.
Agarwal, S., B.W. Ambrose and C. Liu. 2006. Credit Lines and Credit Utilization.Journal of Money, Credit and Banking 38(1): 1–22.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
Homeowners’ Housing Wealth Misestimation 153
Agarwal, S., S. Chomsisengphet and O. Hassler. 2005. The Impact of the 2001 FinancialCrisis and the Subsequent Economic Response on the Argentine Mortgage Market.Journal of Housing Economics 14(3): 242–270.
Agarwal, S., S. Chomsisengphet, C. Liu and N. Souleles. 2006. Do Consumers Choosethe Right Credit Contracts? Working Paper. University of Pennsylvania.
Agarwal, S., J. Driscoll and D. Laibson. 2004. Mortgage Refinancing for DistractedConsumers. Working Paper. Harvard University.
Agarwal, S., J. Driscoll, X. Gabaix and D. Laibson. 2006. Financial Mistakes over theLifecycle. Working Paper. Harvard University.
Bailey, M.J., R.F. Muth and H.O. Nourse. 1963. A Regression Method for Real EstatePrice Index Construction. Journal of the American Statistical Association 58: 933–942.
Bucks, B. and K. Pence. 2005. Comparing Homeowner and Lender Estimates of HousingWealth and Mortgage Terms. Working Paper. Federal Reserve Board.
Canner, G.B., K. Dynan and W. Passmore. 2002. Mortgage Refinancing in 2001 andEarly 2002. Federal Reserve Bulletin 88(12): 469–481.
Case, K.E., J.M. Quigley and R.J. Shiller. 2005. Comparing Wealth Effects: The StockMarket versus the Housing Market. Advances in Macroeconomics 5(1): 1–32.
Case, K.E. and R.J. Shiller. 1987. Prices of Single-Family Homes since 1970: NewIndexes for Four Cities. New England Economic Review September/October: 45–56.——. 1989. The Efficiency of the Market for Single-Family Homes. American EconomicReview 79(1): 125–137.——. 1990. Forecasting Prices and Excess Returns in the Housing Market. AmericanReal Estate and Urban Economics Association Journal 18(3): 253–273.——. 2003. Is There a Bubble in the Housing Market? Working Paper. Yale University.
Deng, Y., J.M. Quigley and R. Van Order. 2000. Mortgage Terminations, Heterogeneityand the Exercise of Mortgage Options. Econometrica 68(2): 275–307.
DiPasquale, D. and C.T. Somerville. 1995. Do House Price Indices Based on TransactingUnits Represent the Entire Stock? Evidence from the American Housing Survey. Journalof Housing Economics 4: 195–229.
Engelhardt, G.V. 1996. House Prices and Home Owner Saving Behavior. RegionalScience and Urban Economics 26: 313–336.
Follain, J.R. and S. Malpezzi. 1981. Are Occupants Accurate Appraisers? Review ofPublic Data Use 9: 47–55.
Gabaix, X., D. Laibson, G. Moloche and S. Weinberg. 2006. Costly Information Ac-quisition: Experimental Analysis of a Boundedly Rational Model. American EconomicReview 96(4): 1043–1068.
Goodman, J.L. and J.B. Ittner. 1992. The Accuracy of Home Owners’ Estimates ofHouse Value. Journal of Housing Economics 2(4): 339–357.
Hoynes, H.W. and D.L. McFadden. 1997. The Impact of Demographics on Housingand Non Housing Wealth in the United States. M.D. Hurd and Y. Naohiro, editors. TheEconomic Effects of Aging in the United States and Japan. University of Chicago Press:Chicago, for NBER. 153–194.
Hurst, E. and F. Stafford. 2004. Home is Where the Equity Is: Mortgage Refinancingand Household Consumption. Journal of Money, Credit, and Banking 36(6): 985–1014.
Ihlanfeldt, K.R. and J. Martinez-Vazquez. 1986. Alternative Value Estimates of Owner-Occupied Housing: Evidence on Sample Selection Bias and Systematic Errors. Journalof Urban Economics 20: 356–369.
Kain, J.F. and J. Quigley. 1972. Note on Owners Estimate of Housing Value. Journal ofthe American Statistical Association 67: 803–806.
reec˙185 REEC.cls March 13, 2007 20:38 Char Count=
154 Agarwal
Kish, L. and J.B. Lansing. 1954. Response Errors in Estimating the Value of Homes.Journal of the American Statistical Association 49: 520–538.
Koszegi, B. 2005. Ego Utility, Overconfidence, and Task Choice. Working Paper. Uni-versity of California – Berkeley.
Loebs, T. 2005. Systemic Risks in Residential Property Valuations: Perceptions andReality. Working Paper. Collateral Assessment and Technologies Committee.
Nothaft, F.E. 2004. The Contribution of Home Value Appreciation to US EconomicGrowth. Urban Policy and Research 22(1): 23–34.
Robins, P.K. and R.W. West. 1977. Measurement Errors in the Estimation of HomeValue. Journal of the American Statistical Association 72: 290–294.
Schrag, J. and M. Rabin. 1999. First Impressions Matter: A Model of Confirmatory Bias.Quarterly Journal of Economics 114(1): 37–82.
Skinner, J. 1989. Housing Wealth and Aggregate Saving. Regional Science and UrbanEconomics 19: 305–324.
Yariv, L. 2005. I’ll See It When I Believe It—A Simple Model of Cognitive Consistency.Working Paper. University of California – Los Angeles.