Professional Documents
Culture Documents
The Consequences of Mortgage Credit Expansion: Evidence From The U.S. Mortgage Default Crisis
The Consequences of Mortgage Credit Expansion: Evidence From The U.S. Mortgage Default Crisis
I. INTRODUCTION
The sharp rise in U.S. mortgage default rates has led to the
most severe financial crisis since the Great Depression. A salient
feature of the mortgage default crisis is that it is concentrated
in subprime ZIP codes throughout the entire country. A compari-
son of subprime and prime ZIP codes, which are defined to be the
highest and lowest quartile ZIP codes in the national distribution
based on the fraction of borrowers with a credit score under 660 as
of 1996, reveals that subprime ZIP codes have experienced an in-
crease in default rates since 2006 that is more than three times as
Technology.
The Quarterly Journal of Economics, November 2009
1449
1450 QUARTERLY JOURNAL OF ECONOMICS
1. All of the statistics mentioned in this paragraph are from the Appendix,
which shows mortgage credit growth, mortgage defaults, and income growth for
subprime and prime ZIP codes in the top forty MSAs in the United States.
2. See, for example, Federal Reserve Chairman Alan Greenspan’s testimony
to the U.S. Congress on June 9, 2005, or Council of Economic Advisors Chairman
Ben Bernanke’s testimony to the U.S. Congress on October 20, 2005.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1451
3. We use MSA-level data in our example because this is the most widely used
level of analysis in studies involving housing markets.
4. One possible explanation for the positive correlation between income
growth and subprime population share at the MSA level might be that MSAs
with a higher fraction of subprime population provide a greater supply of cheap
unskilled labor, which differentially attracts growth opportunities. However, most
of the benefits of these growth opportunities may accrue to prime (skilled) individ-
uals. If there is a contemporaneous expansion in the supply of credit to subprime
areas, subprime populations will have disproportionately stronger credit growth,
therefore creating a spuriously positive correlation between income growth and
credit growth in the between-MSA analysis.
PHO
0.4
0.4
SF ORL
RIV
0.1
PHO TAM ORL RIV TAM ORL
PHO
1452
TAM
MIA
SNT RIV HOU POR SAC DC DC
POR
SAZ SEA SAC
NY ATL NY LA SEA LANY
0.08
LA DAL
FTW OAK BAL VIR OAK BALVIR
DC
0.2
0.2
CHI SD MIA ATL SAZ CHI SAZ
ATL MIA
POR VIR SF
NAS BOS SD
NAS
BOS SF
PHI
SEA PRO
SNT
STL PRO
STL SNT
MIN SD BAL HOU HOU
NAS NEABOSSAC NEA CIN KAN FTW
DAL KANNEA
CIN FTW
DAL
0.06
CIN MIN CLE IND CLE
IND MIN
EDI OAK KAN PIT COL DET DET COL
PIT
PIT PRO
STL DEN DEN
DEN
CHI COL WRE WRE
CLE
0
0
WRE IND
0.04
Credit growth '02 to '05
0
DET
–0.2
PHI PHI
–0.2
0.2 0.25 0.3 0.35 0.4 0.45 0.2 0.25 0.3 0.35 0.4 0.45 0 0.02 0.04 0.06 0.08 0.1 0.12
Fraction of population under 660 in 1996 Fraction of population under 660 in 1996 Income growth '02 to '05
0.5
0.5
0.05
0
0
0
Credit growth '02 to '05
Credit growth '02 to '05
–0.5
–0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 –0.06 –0.04 –0.02 0 0.02 0.04 0.06
QUARTERLY JOURNAL OF ECONOMICS
Fraction of population under 660 in 1996 Fraction of population under 660 in 1996 Income growth '02 to '05
FIGURE I
Income Growth, Mortgage Credit Growth, and Subprime Population: Between- and Within-MSA Correlations
This figure presents correlations between income growth, mortgage credit growth, and the fraction of the population with a credit
score under 660. The top panels present between-MSA correlations for the top forty MSAs by population, and the bottom panels utilize
ZIP code–level data to show the within-MSA correlations for the top forty MSAs. The bottom panels present data that are deviated from
MSA level means.
5. Gerardi, Shapiro, and Willen (2007) also have disaggregated data on all
of these variables, but they focus only on Massachusetts and only on subprime
mortgages.
1456 QUARTERLY JOURNAL OF ECONOMICS
6. HMDA data contain census tract, but not ZIP code, information. We match
census tracts to ZIP codes using a match provided by Geolytics. The match quality
is high: 85% of the matched census tracts in our final sample have over 90% of
their population living in the ZIP code to which they are matched.
7. Because one of our key hypotheses involves house price expectations, our
core sample includes only the 3,014 ZIP codes for which we have ZIP code–level
house price data available. However, all of our results that do not require house
prices are qualitatively similar and only slightly smaller in magnitude if we use the
full sample of 18,408 ZIP codes. In the interest of full transparency, we replicate all
of our regression results that do not use house price data on the full sample in the
Online Appendix. As the Online Appendix shows, the main difference between the
house price and non–house price sample is whether the ZIP code is in an urban
environment. We also collect ZIP code–level price indices for 2,248 ZIP codes from
Zillow.com, an online firm that provides house price data. House price changes for
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1457
FCSW and Zillow have a correlation coefficient of .91, and all of our results using
house prices are robust to the use of Zillow indices.
1458 QUARTERLY JOURNAL OF ECONOMICS
TABLE I
Between- Within-
Mean SD county SD county SD
Equifax data
Mortgage debt annualized growth, 0.089 0.065 0.048 0.058
1996–2002
Non-home debt annualized growth, 0.073 0.092 0.040 0.087
1996–2002
Mortgage debt annualized growth, 0.145 0.081 0.049 0.071
2002–2005
Non-home debt annualized growth, 0.058 0.064 0.034 0.059
2002–2005
Mortgage default rate, 1996 0.030 0.024 0.013 0.021
Non-home default rate, 1996 0.068 0.035 0.019 0.030
Mortgage default rate change, −0.003 0.028 0.018 0.023
1996–2005
Non-home default rate change, −0.008 0.031 0.018 0.026
1996–2005
Mortgage default rate change, 2005–2007 0.035 0.038 0.024 0.032
Non-home default rate change, 0.018 0.026 0.012 0.024
2005–2007
Subprime consumer fraction 0.287 0.113 0.063 0.094
(under 660), 1996
Fiserv Case–Shiller–Weiss data
House price annualized growth, 0.096 0.033 0.034 0.011
1996–2002
House price annualized growth, 0.153 0.078 0.078 0.016
2002–2005
HMDA data
Mortgage origination for home purchase 0.144 0.089 0.068 0.067
ann. growth 1996–2002
Mortgage origination for home purchase 0.194 0.189 0.130 0.150
ann. growth 2002–2005
Fraction sold to nonagency investors, 1996 0.253 0.081 0.078 0.045
Change in fraction sold, 1996–2002 0.044 0.057 0.045 0.043
Change in fraction sold, 2002–2005 0.241 0.051 0.038 0.038
IRS and Census statistics of U.S. business
Income annualized growth, 1991–1998 0.052 0.025 0.012 0.022
Income annualized growth, 1998–2002 0.023 0.020 0.011 0.017
Income annualized growth, 2002–2005 0.048 0.035 0.021 0.031
Income annualized growth, 2005–2006 0.041 0.044 0.023 0.040
Employment annualized growth, 0.016 0.058 0.030 0.053
2002–2005
Establishment annualized growth, 0.017 0.029 0.019 0.024
2002–2005
Note. This table presents summary statistics for the 3,014 ZIP codes in our sample.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1459
8. See http://www.freddiemac.com/corporate/reports/moseley/chap6.htm.
9. See congressional testimony by Staten at http://financialservices.house.gov/
media/pdf/033004ms.pdf#page=3.
10. In contrast to other research, our analysis focuses on all mortgages to sub-
prime borrowers rather than on mortgages deemed to be “subprime mortgages”
by alternative definitions (see Mayer and Pence [2008] for a review of these def-
initions). There is a distinction: Subprime borrowers can obtain non-subprime
mortgages and prime borrowers can obtain subprime mortgages. Further, this
distinction is important. For example, Mayer and Pence (2008) argue that the
northeastern United States “did not see especially high rates of subprime usage.”
Our evidence suggests otherwise: As the Appendix shows, subprime ZIP codes in
Boston, Nassau, New York, Newark, and Providence all experience a growth in
mortgage originations that is more than twice as large as for prime ZIP codes in the
same MSA from 2002 to 2005. We are more interested in subprime households’
increased access to credit, whether or not it comes from a mortgage defined to be
“subprime.”
1460 QUARTERLY JOURNAL OF ECONOMICS
TABLE II
Prime Subprime
ZIP codes ZIP codes
Notes. This table presents characteristics of prime and subprime ZIP codes in our sample. Prime and
subprime ZIP codes are determined by splitting ZIP codes into four quartiles based on the national distribution
of the fraction of subprime borrowers (credit score less than 660) as of 1996. Prime ZIP codes are the lowest
quartile and subprime ZIP codes are the highest quartile.
∗∗ ,∗ Difference between prime and subprime statistically distinct from 0 at the 1% and 5% levels, respec-
tively.
in the nation and subprime ZIP codes are in the highest quartile
in the nation.
Subprime ZIP codes have reduced access to mortgage lend-
ing before the subprime mortgage expansion. A higher fraction
of mortgages in subprime ZIP codes as of 1996 are backed by
the Federal Housing Administration, and mortgage application
denial rates as of 1996 are significantly higher. Homeownership
data from the 2000 census show a 25% lower homeownership rate
in subprime ZIP codes. As of 2000, subprime ZIP codes have much
lower median household income, much higher poverty rates, much
lower levels of education, and much higher unemployment rates.
They also have a significantly larger fraction of the population
that is nonwhite.
0.6
0.4
0.4
0.2
0.2
0
–0.2
–0.2
1993 1995 1997 1999 2001 2003 2005 2007 1993 1995 1997 1999 2001 2003 2005 2007
Year Year
(Growth is cumulative since 1991) (Growth is cumulative since 1991)
0.08
0.06
0.04
0.02
FIGURE II
Mortgage Credit Growth and Default Rates: Subprime Relative to Prime ZIP Codes
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION
This figure plots the growth in the number (top left-hand panel) and amount (top right-hand panel) of originated mortgages and the
mortgage default rate (bottom panel) for subprime relative to prime ZIP codes in the same county. Subprime and prime ZIP codes are
defined to be the highest and lowest quartile ZIP codes in the national distribution based on the fraction of residents with a credit score
less than 660 as of 1991.
1461
11. The choice of base year for categorizing ZIP codes as “subprime” is not
important for our results, because of a high level of persistence in the rank of ZIP
codes by subprime population share. For example, the correlation between shares
of subprime population in 1991 and 1996 is .8.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1463
12. One can think of prime customers as those who are qualified to be guar-
anteed by GSEs under the regulatory guidelines.
1464 QUARTERLY JOURNAL OF ECONOMICS
Pzte is equal to (Pzt+1 /Pzt ) and denotes expected house price ap-
preciation in a borrower’s ZIP code. The constant r̄ reflects the
interest rate ceiling beyond which no lender is willing to lend.
We do not model explicitly the underlying friction that leads to
an interest rate ceiling above which originators are unwilling to
lend—borrower moral hazard (Diamond 1991; Holmstrom and Ti-
role 1997) and adverse selection (Stiglitz and Weiss 1981) are po-
tential reasons.14
Let gzt be the fraction of subprime customers in a ZIP code
that are able to get a mortgage for home purchase. These are
the customers who are not credit-rationed, that is, customers for
whom the right-hand side of (1) is less than r̄.
Income Profile (I jzt ). An upward shift in the income distribu-
tion of the subprime population reduces the likelihood of default
(δ) for subprime customers, and hence leads to a higher acceptance
rate for mortgages in (1). More generally, an upward shift in I jzt
captures creditworthiness factors such as higher wages and better
expected employment that increase the ability to repay debt.
Credit Supply Factor (θt ). A reduction in the risk premium
makes mortgages affordable for a greater fraction of the popula-
tion. The risk premium in the mortgage market may go down for a
number of reasons, including greater diversification of mortgage
risk across financial institutions, lax lending standards on be-
half of originators, government programs that subsidize the risk
13. We assume that the lender always loses some principal in the event of de-
fault; that is, αPzt+1 < (1 − γ )Pzt . This assumption is justified by Pence (2006),
who shows that average mortgage losses on foreclosures range from 30% to
60%.
14. Gabriel and Rosenthal (2007) explicitly model how a supply expansion
affects borrowers with a Stiglitz and Weiss (1981) adverse selection problem. Their
conclusions are similar to ours.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1465
Lzt = fz + gzt ∗ (1 − fz ) .
Equation (2) shows that loan originations in a given ZIP code will
be a function of the fraction of subprime borrowers and of the
fraction of subprime borrowers that receive a mortgage. Because
equation (2) includes county interacted with time fixed effects,
county-level changes in income, home prices, or other variables
that uniformly affect ZIP codes in a given county are removed.
First differencing (2) gives us
where βt = gzt .
Equation (3) represents our primary regression specification,
where (1 − fz ) represents the fraction of subprime borrowers in
a ZIP code in the initial period. As we show in Figures II and
III, there is a rapid relative expansion in mortgage lending to
subprime borrowers from 2002 to 2005, which implies that the
estimate of βt in equation (4) is statistically significantly positive
and economically meaningful. As equation (1) shows, the posi-
tive estimate of βt could be due to one of three potential factors:
improved income prospects of subprime borrowers (income-based
hypothesis), a decline in the risk premium charged by lenders
(supply-based hypothesis), or increased house price expectations
1
1466
0.5
0
with household income growth
Mortgage credit growth correlation
–0.5
1991 1993 1995 1997 1999 2001 2003 2005
Year
0.15
–0.1
0.1
–0.15
–0.2
–0.05
Income growth:
0
–0.25
Debt to income ratio:
–0.3
–0.05
1991 1993 1995 1997 1999 2001 2003 2005
Year 1991 1993 1995 1997 1999 2001 2003 2005
(Growth is cumulative since 1991) Year
QUARTERLY JOURNAL OF ECONOMICS
FIGURE III
Mortgage Credit Growth and Income Growth over Time
The top panel in this figure plots the correlation across ZIP codes between income growth and mortgage credit growth over time,
after they are deviated from county means. The bottom left-hand panel plots the relative income growth for subprime relative to prime
ZIP codes in the same counties, and the bottom right-hand panel plots the relative debt-to-income ratio for subprime relative to prime
ZIP codes in the same counties. The debt-to-income ratio is defined as total originated mortgages for home purchase in a ZIP code scaled
by the total income of the ZIP code. Subprime and prime ZIP codes are defined to be the highest and lowest quartile ZIP codes in the
national distribution based on the fraction of residents with a credit score less than 660 as of 1991.
TABLE III
Mortgage
origination Income Employment Establishment
growth growth growth growth
2002–2005 2002–2005 2002–2005 2002–2005
(1) (2) (3) (4)
Notes. This table presents the correlation across ZIP codes between measures of income growth, employ-
ment growth, and business growth and the fraction of subprime borrowers. All regressions include county
fixed effects.
∗∗ ,∗ Statistically distinct from 0 at the 1% and 5% levels, respectively.
15. Strictly speaking, our model generates a positive βt (i.e., higher relative
mortgage growth for subprime ZIP codes) under the house price appreciation
hypothesis only if the house price expectations go up differentially for subprime
ZIP codes. However, one can imagine that a level increase in house price growth
expectation helps subprime customers more because they have a higher probability
of default, and hence a reduction in loss given default is more useful to them.
16. The inclusion of county fixed effects means that our measure of subprime
borrowers is deviated from county means in the regressions. An alternative specifi-
cation is to use the absolute measure of the fraction of borrowers that are subprime
while using deviations from county means for all other variables. In unreported re-
sults, we find similar quantitative results when using this alternative specification.
17. Given the presence of county fixed effects in all specifications, we use
within-county standard deviations when discussing economic magnitudes. Within-
county standard deviations are reported in Table I.
1468 QUARTERLY JOURNAL OF ECONOMICS
18. The inclusion of county fixed effects is critical to capture the negative cor-
relation between measures of economic growth and the fraction of subprime bor-
rowers. As we show in the Online Appendix, the correlation between employment
growth/establishment growth and the fraction of subprime borrowers is positive
when county fixed effects are excluded. Likewise, the negative correlation between
mortgage credit growth/house price growth and income growth is positive when
county fixed effects are not included. This is consistent with Figure I and the
corresponding discussion in the Introduction: examining data at a level more ag-
gregated than ZIP codes leads to misleading conclusions concerning income growth
patterns, mortgage credit growth, and subprime borrower share.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1469
19. As noted in the Data section, we only have ZIP code–level income infor-
mation from the IRS for 1991, 1998, 2001, 2002, 2004, 2005, and 2006.
20. The mortgage debt–to–income ratio is measured as total mortgage origi-
nations in a ZIP code (HMDA) divided by total income of residents in a ZIP code
(IRS).
1470
TABLE IV
HISTORICAL MORTGAGE CREDIT GROWTH AND INCOME GROWTH CORRELATIONS
Income growth, −0.662∗∗ 0.537∗∗ 0.517∗∗ 0.425 −0.394∗∗ −0.383∗∗ 0.103 0.716∗∗
annualized (0.089) (0.084) (0.092) (0.368) (0.122) (0.077) (0.078) (0.093)
Notes. This table presents the correlation across ZIP codes between mortgage origination for home purchase growth and income growth for different periods of our sample.
Income growth and mortgage origination growth are measured for the exact same period for all specifications except the specification reported in column (7). We do not have income
data available for 2007; as a result, in column (7) we examine the correlation between mortgage origination growth from 2006 to 2007 and income growth from 2005 to 2006. All
specifications include county fixed effects.
∗∗ ,∗ Coefficient estimate statistically distinct from 0 at the 1% and 5% levels, respectively.
QUARTERLY JOURNAL OF ECONOMICS
Fraction of subprime borrowers, 1996 0.460∗∗ 0.458∗∗ 0.454∗∗ 0.431∗∗ 0.050∗∗ −0.039∗∗
(0.033) (0.033) (0.035) (0.079) (0.015) (0.013)
Income growth, 2002–2005 −0.064 −0.068 −0.084 0.145 0.360∗∗ 0.201∗∗
(0.097) (0.097) (0.104) (0.223) (0.044) (0.037)
Establishment growth, 2002–2005 0.084 0.096 0.118 −1.898∗∗ 0.805∗∗ 0.546∗∗
(0.124) (0.127) (0.142) (0.314) (0.058) (0.049)
Employment growth, 2002–2005 −0.054 −0.053 −0.041 0.001 −0.040 0.001
(0.056) (0.056) (0.056) (0.101) (0.026) (0.021)
Crime growth, 2002–2005 −0.105 −0.012 0.551 0.251∗ 0.240∗
(0.256) (0.278) (0.779) (0.117) (0.098)
House price elasticity with respect to income −0.014∗∗ −0.010
(0.005) (0.009)
QUARTERLY JOURNAL OF ECONOMICS
Fraction of housing stock built 2–5 years ago, 2000 −0.108 0.162
(0.078) (0.246)
N 2,946 2,946 2,782 2,782 2,946 2,946
R2 .42 .42 .45 .94 .33 .26
Notes. This table presents coefficient estimates from specifications relating the growth in mortgage originations for home purchase in a ZIP code from 2002 to 2005 to the fraction
of subprime borrowers in 1996. Column (5) examines the growth in total mortgage debt, and column (6) examines the growth in non-home consumer debt. The measure of house price
elasticity (house quantity elasticity) uses changes in median house value (number of owner-occupied housing units) from 1990 to 2000 from the decennial census, and changes in
household income from 1991 to 2001 from the IRS. All growth rates are annualized. All specifications include county fixed effects except for the specification reported in column (4),
which includes three–square mile fixed effects.
∗∗ ,∗ Coefficient estimate statistically distinct from 0 at the 1% and 5% levels, respectively.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION
1473
21. It should be kept in mind that because we have county fixed effects in a
first-differenced specification, all macro shocks—even those that are at the level of
counties—that impact everyone in the economy equally are automatically absorbed
away. It is only the differential reaction to a given macro shock that can potentially
bias our coefficient of interest.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1475
22. The thirty-year fixed rate mortgage rate also falls from 2001 to 2005, but
it is important to emphasize that decline in the thirty-year fixed rate mortgage
is consistent with all three of our hypotheses. Either an increase in borrower
income, a shift in the supply of mortgage credit, or increases in house price growth
expectations would lead to a reduction in the thirty-year fixed rate mortgage.
8
8
7
7
1476
6
6
5
5
4
4
3
3
3 month T-bill yield
1
1
0
0
01jan2001 01jan2002 01jan2003 01jan2004 01jan2005 01jan2006 01jan1990 01jan1991 01jan1992 01jan1993 01jan1994 01jan1995
0.4
0.2
2001-2005
0
1990-1994
FIGURE IV
Relative Mortgage Origination Growth for Subprime ZIP Codes: Falling Interest Rate Periods
The top left- and right-hand panels show the evolution of the three-month Treasury bill yield during the five-year periods from 1990
to 1994 and 2001 to 2005, respectively. The bottom panel shows the growth in the amount of originated mortgages for subprime relative
to prime ZIP codes in the same counties for these two five-year periods. Subprime and prime ZIP codes are defined to be the highest- and
lowest-quartile ZIP codes in the national distribution based on the fraction of residents with a credit score below 660 as of the first years
of the respective five-year periods.
0.18
0.1
0.16
0
0.14
–0.1
subprime minus prime ZIP codes
0.12
0.6
0.12
0.1
0.5
0.08
0.4
0.06
0.3
0.04
0.2
0.02
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year Year
QUARTERLY JOURNAL OF ECONOMICS
FIGURE V
Relaxation in Borrower Credit Constraints
The top left-hand panel shows the correlation across ZIP codes between the mortgage application denial rate and the fraction of
residents with a credit score below 660, both as of 1996. The data are deviated from county means. The top right-hand panel shows the
denial rate for mortgage applications for prime relative to subprime ZIP codes in the same county. The bottom left-hand panel shows the
fraction of all originated mortgages for home purchase that are sold to non-GSE investors, and the bottom right-hand panel shows the
relative fraction sold to non-GSE investors for subprime versus prime ZIP codes in the same county. Subprime and prime ZIP codes are
defined to be the highest- and lowest-quartile ZIP codes in the national distribution based on the fraction of residents with a credit score
below 660 as of 1996.
Fraction of subprime borrowers, 1996 −0.094∗∗ 0.048∗∗ −0.055∗∗ −0.007∗ 0.104∗∗ 0.077∗∗
(0.006) (0.009) (0.005) (0.003) (0.004) (0.004)
N 2,946 2,946 2,946 2,946 2,946 2,946
R2 .58 .46 .56 .46 .68 .61
Panel B: Mortgage sales and changes in default rate
Change in mortgage default rate from 2005 to 2007
(1) (2) (3) (4) (5)
Notes. Panel A presents coefficient estimates relating the change in the fraction of originated mortgages sold in a ZIP code from 2002 to 2005 to the share of subprime borrowers
as of 1996. Panel B presents estimates relating default rates from 2005 to 2007 to the fraction of loans sold by originators to investors from 2002 to 2005. All specifications include
county fixed effects and control variables for income, wage, employment, business establishments, and crime growth.
∗∗ ,∗ Coefficient estimate statistically distinct from 0 at the 1% and 5% levels, respectively.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION
1481
23. The mean change in fraction sold to non-GSE investors over 2002–2005 is
24.7%. This change is divided across the four subcategories as 5.4%, 3.9%, 5.2%,
and 10.8%, respectively.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1483
24. The fact that some commercial banks took major capital hits during the
mortgage default crisis partially contradicts this argument. However, the evidence
thus far suggests that the only commercial banks that experienced major losses
were those heavily involved in securitization through special investment vehicles
off balance sheet.
1484 QUARTERLY JOURNAL OF ECONOMICS
0.15
0.1
0.1
00.05
0.05
–0.1 –0.05
1997 1999 2001 2003 2005 2007
1997 1999 2001 2003 2005 2007 Year
Year (Growth is cumulative since 1996)
0.2
0.15
Inelastic housing
supply MSAs
Inelastic housing
0.15
supply MSAs
0.1
0.1
0.05
Elastic housing
0.05
supply MSAs
Elastic housing
0
Annual house price growth
subprime minus prime ZIP codes
–0.05
–0.05
1997 1999 2001 2003 2005 2007
1997 1999 2001 2003 2005 2007 Year
Year (Growth is cumulative since 1996)
FIGURE VI
House Price Growth
The top left-hand panel shows average annual house price growth across ZIP codes. The top right-hand panel shows relative house
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION
price growth for subprime versus prime ZIP codes in the same county. The bottom left-hand panel shows average annual house price
growth for the top-decile and bottom-decile housing supply–elasticity MSAs based on the elasticity measures of Saiz (2008). The bottom
right-hand panel shows relative house price growth for subprime versus prime ZIP codes in the same MSA for elastic and inelastic MSAs.
Subprime and prime ZIP codes are defined to be the highest- and lowest-quartile ZIP codes in the national distribution based on the
fraction of residents with a credit score below 660 as of 1996.
1485
25. In order to select the most extreme MSAs in terms of elasticity, we pick
MSAs from our full sample, not the sample that is restricted to Case–Shiller home
prices. MSA-level housing price indices are from OFHEO. However, when a test
uses ZIP-level house prices (i.e., the top right- and bottom right-hand panels in
Figure VI), we have to limit ourselves to ZIP codes present in the Case–Shiller
home price data set, which represents 84 MSAs. Correspondingly, we have to
redefine the most elastic and inelastic MSAs for the top right- and bottom right-
hand panels in Figure VI. The most elastic MSAs in the house price sample are
Akron, OH; Atlanta, GA; Charlotte, NC; Cincinnati, OH; Greensboro, NC; Toledo,
OH; and Youngstown, OH.
26. This figure highlights that our assumption that lenders understand that
house price appreciation is bounded in highly elastic MSA is a relatively weak
assumption. Even some irrational house price expectations by lenders (such as
adaptive expectations) would still recognize the limits of house price appreciation
in highly elastic MSAs. There is a possible scenario (under irrational beliefs)
where an expansion in house prices in the inelastic housing supply areas leads to
an expansion in credit supply in elastic areas. This could be the case if lenders
irrationally extrapolated the increase in home prices in inelastic areas to house
price patterns in elastic areas.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1487
elastic MSAs. The relative house price growth for subprime ZIP
27. It is interesting to note that the relative house price declines for subprime
ZIP codes in elastic areas beyond 2006 after the subprime mortgage market col-
lapses. This is entirely in line with the predictions of Glaeser, Gyourko, and Saiz
(2008), which show that an expansion in the quantity of housing in preceding years
forces house prices to go below construction costs when housing demand pressures
recede.
1488
TABLE VII
MORTGAGE ORIGINATION GROWTH AND MORTGAGE DEFAULT RATE CHANGES FOR HIGH–HOUSING SUPPLY ELASTICITY MSAS
Change in Change in
fraction sold in fraction to
Income growth securitizations other financial Mortgage origination Change in mortgage
2002–2005 2002–2005 firms 2002–2005 growth 2002–2005 default rate 2002–2005
With controls With controls
(1) (2) (3) (4) (5) (6) (7)
Fraction subprime borrowers, 1996 −0.069∗∗ 0.100∗∗ 0.061∗∗ 0.305∗∗ 0.413∗∗ 0.057∗∗ 0.056∗∗
(0.010) (0.009) (0.014) (0.061) (0.069) (0.015) (0.018)
N 655 655 655 655 655 655 655
R2 .17 .28 .43 .10 .12 .04 .05
Notes. This table examines relative income growth, securitization patterns, and mortgage origination growth from 2002 to 2005 and the relative change in mortgage default
rates from 2005 to 2007 in high–subprime borrower share ZIP codes. The sample is limited to the top–decile MSAs by housing supply elasticity as measured by Saiz (2008). The ten
high–housing supply elasticity MSAs are Dayton, OH; Fort Wayne, IN; Greenville, SC; Indianapolis, IN; Kansas City, MO; Little Rock, AR; McAllen, TX; Omaha, NE; Tulsa, OK;
QUARTERLY JOURNAL OF ECONOMICS
and Wichita, KS. The control variables in columns (5) and (7) are income growth, employment growth, and business establishment growth from 2002 to 2005. All growth rates are
annualized, and all specifications include MSA fixed effects.
∗∗ ,∗ Coefficient estimate statistically distinct from 0 at the 1% and 5% levels, respectively.
Income growth, annualized −0.118∗∗ 0.138∗∗ 0.070∗∗ 0.130∗∗ −0.074∗∗ −0.082∗∗ −0.024∗ 0.195∗∗
(0.010) (0.008) (0.013) (0.011) (0.010) (0.008) (0.010) (0.017)
N 3,014 2,875 3,014 3,014 3,014 3,014 3,014 2,721
R2 .96 .90 .87 .89 .95 .95 .78 .87
Notes. This table presents the correlation across ZIP codes between house price growth and income growth for different periods of our sample. Income growth and house price
growth are measured for the exact same period for all specifications except the specification reported in column (7). We do not have income data available for 2007; as a result, in
column (7) we examine the correlation between house price growth from 2006 to 2007 and income growth from 2005 to 2006. All growth rates are annualized, and all specifications
include county fixed effects.
∗∗ ,∗ Coefficient estimate statistically distinct from 0 at the 1% and 5% levels, respectively.
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION
1489
0.2
2001–2005
House price growth:
1990–1994
–0.05
FIGURE VII
House Price Growth: Historical Evidence
The top panel in this figure plots the correlation across ZIP codes between
income growth and house price growth over time, after they are deviated from
county means. The bottom panel shows house price growth for subprime relative
to prime ZIP codes in the same county during the five-year periods from 1990 to
1994 and 2001 to 2005. Subprime and prime ZIP codes are defined to be the highest-
and lowest-quartile ZIP codes in the national distribution based on the fraction
of residents with a credit score below 660 as of the first year of the respective
five-year period.
1492 QUARTERLY JOURNAL OF ECONOMICS
VII. CONCLUSIONS
We find evidence of sharp relative growth in mortgage credit
followed by a sharp relative increase in defaults in high subprime
share ZIP codes. This pattern is prevalent in almost every city in
the United States (Appendix). These facts demonstrate that any
study seeking to understand the origins of the mortgage default
crisis must explain the expansion of mortgage credit to subprime
neighborhoods across the entire country.
Our ZIP code–level data set enables us to uncover several new
facts regarding the rise in mortgage credit to subprime neighbor-
hoods. We show that subprime ZIP codes experience strong rela-
tive growth in mortgage credit from 2002 to 2005 despite negative
relative, and in some cases absolute, income growth. In fact, this
is the only period in the past eighteen years in which the corre-
lation between mortgage credit growth and income growth at the
ZIP code level is negative. This latter fact suggests a historic shift
in the supply of mortgage credit. There is a possibility that histor-
ically low risk-free rates from 2001 to 2005 are responsible for the
subprime mortgage credit expansion; however, there is no such ex-
pansion when risk-free rates drop sharply from 1990 to 1994, and
there is no corresponding shift in non-mortgage consumer credit
from 2001 to 2005. Additionally, the robustness of the results to
the inclusion of very refined three–square mile block fixed effects
shows that mortgage credit growth in subprime ZIP codes is un-
likely to be explained by local housing factors such as housing
supply elasticity.
Concurrent with these patterns are a rapid growth in the
securitization of subprime mortgages, a disproportionate de-
cline in denial rates for subprime ZIP codes, and a drop in
the subprime–prime interest spread. All of these patterns hold
within very elastic housing supply MSAs in which an increase
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION 1493
Anaheim, CA 7.1 2.2 226.6 20.8 2.9 619.2 7.1 11.0 −35.7
Atlanta, GA 2.9 1.1 166.1 19.6 16.4 19.5 6.4 10.2 −37.3
Baltimore, MD 1.5 1.1 38.9 32.9 10.7 207.7 4.6 8.0 −43.1
Boston, MA 7.0 1.4 414.3 24.5 11.4 113.6 2.6 7.5 −65.7
Chicago, IL 4.2 1.7 149.8 28.4 15.3 85.9 2.7 5.4 −51.2
Cincinnati, OH −0.1 0.5 −114.8 13.5 10.3 30.8 1.8 6.7 −73.0
Cleveland, OH 3.6 1.5 139.8 19.1 7.2 164.0 0.8 5.4 −84.7
Columbus, OH 1.9 −0.1 2,368.9 4.7 3.7 27.5 0.5 6.5 −92.4
Dallas, TX 1.4 0.8 76.6 12.9 25.9 −50.2 5.6 12.2 −54.0
Denver, CO 1.7 1.0 70.5 14.1 5.3 164.8 4.5 6.3 −28.3
Detroit, MI 6.6 1.4 384.8 23.1 3.4 580.1 −1.8 1.3 −234.9
Edison, NJ 7.0 1.4 397.7 −7.3 −21.0 65.0 3.2 6.1 −47.9
Fort Worth, TX 1.8 0.9 87.8 12.2 20.5 −40.6 4.7 11.8 −60.4
Houston, TX 2.9 0.7 301.7 15.8 −1.2 1,413.1 4.5 13.7 −67.3
Indianapolis, IN 2.6 0.6 314.4 10.8 13.8 −21.8 0.7 6.4 −89.9
QUARTERLY JOURNAL OF ECONOMICS
Kansas City, MO 1.1 0.9 23.0 29.7 −1.0 2,986.4 0.9 8.5 −89.5
Los Angeles, CA 7.9 1.8 346.5 34.9 10.2 241.7 4.6 12.1 −62.3
Miami, FL 9.7 5.7 70.1 12.1 18.2 −33.7 5.8 10.2 −43.3
Minneapolis, MN 4.0 1.5 159.1 24.4 6.8 257.7 2.2 7.2 −69.8
Nassau, NY 9.5 1.8 432.3 34.9 13.8 153.2 1.2 6.7 −81.8
New York, NY 6.6 0.9 624.4 37.1 14.5 155.1 4.0 9.4 −57.9
Newark, NJ 8.6 1.4 504.5 41.8 4.8 779.6 1.7 7.1 −76.5
Increase in mortgage default rate Annualized mortgage origination for Annualized income growth
(percentage points) home purchase growth (percentage (percentage points)
2005 to 2007 points) 2002 to 2005 2002 to 2005
Subprime Prime Subprime Subprime Prime Subprime Subprime Prime Subprime
ZIP codes ZIP codes multiple ZIP codes ZIP codes multiple ZIP codes ZIP codes multiple
Oakland, CA 9.2 1.3 591.7 39.1 11.8 230.9 3.0 7.3 −59.0
Orlando, FL 6.2 1.5 306.2 33.6 −2.4 1,475.4 7.3 22.4 −67.4
Philadelphia, PA 2.0 0.9 135.8 −30.1 −16.3 −84.4 2.7 7.6 −64.7
Phoenix, AZ 8.1 2.6 210.9 37.4 26.8 39.8 8.4 15.6 −45.9
Pittsburgh, PA −0.9 1.3 −171.9 4.7 6.8 −31.1 1.6 6.2 −73.4
Portland, OR 0.0 1.1 −98.2 32.3 18.4 75.0 9.5 8.7 8.5
Providence, RI 7.8 2.5 212.9 34.4 7.3 374.5 4.4 6.7 −34.9
Riverside, CA 10.8 6.0 79.6 37.6 25.6 47.0 6.4 13.3 −51.8
Sacramento, CA 13.2 2.7 385.4 44.7 15.9 182.0 2.9 7.3 −60.7
San Antonio, TX 1.5 0.7 130.6 18.0 21.3 −15.4 5.2 15.4 −66.4
San Diego, CA 7.8 1.1 639.1 30.2 11.8 157.2 5.0 9.0 −45.0
San Francisco, CA 3.3 1.0 230.7 25.7 15.3 67.7 7.8 11.8 −33.7
Seattle, WA 2.8 0.7 313.8 21.0 20.3 3.3 3.3 8.2 −59.4
St. Louis, MO 4.6 0.6 691.0 31.5 7.1 341.3 0.2 6.2 −96.9
Tampa, FL 4.8 3.1 53.1 44.9 35.1 28.0 7.9 12.7 −37.9
Troy, MI 3.1 2.1 49.4 5.7 2.9 96.6 1.6 4.6 −65.1
Virginia Beach, VA 1.7 −0.2 858.7 27.2 10.2 166.9 5.3 7.3 −27.4
Washington, DC 6.2 1.5 308.2 32.8 17.9 83.2 4.5 9.6 −53.6
CONSEQUENCES OF MORTGAGE CREDIT EXPANSION
Averages 4.8 1.5 302.8 23.3 10.7 278.9 3.9 9.0 −62.0
Notes. This table presents information on the change in the mortgage default rate from 2005 to 2007, the annualized growth in mortgage originations for home purchase from
2002 to 2005, and the annualized growth rate in income from 2002 to 2005 for the entire United States and broken down by top 40 MSAs by population. Subprime and prime zip
1495
codes are defined to be the highest and lowest quartile zip codes in the national distribution based on the fraction of residents with a credit score less than 660 as of 1996.
REFERENCES
Ashcraft, Adam, and Til Schuermann, “Understanding the Securitization of Sub-
prime Mortgage Credit,” Federal Reserve Bank of New York Staff Report No.
318, 2008.
Chomsisengphet, Souphala, and Anthony Pennington-Cross, “The Evolution of the
Subprime Mortgage Market,” Federal Reserve Bank of St. Louis Review, 88
(2006), 31–56.
Dell’Ariccia, Giovanni, Deniz Igan, and Luc Laeven, “Credit Booms and Lend-
ing Standard: Evidence from the Subprime Mortgage Market,” IMF Working
Paper, 2008.
Demyanyk, Yuliya, and Otto Van Hemert, “Understanding the Subprime Mortgage
Crisis,” Review of Financial Studies, forthcoming.
Diamond, Douglas, “Monitoring and Reputation: The Choice between Bank Loans
and Privately Placed Debt,” Journal of Political Economy, 99 (1991), 689–721.
Doms, Mark, Fred Furlong, and John Krainer, “Subprime Mortgage Delinquency
Rates,” Federal Reserve Bank of San Francisco Working Paper, 2007.
Gabriel, Stuart, and Stuart Rosenthal, “Secondary Markets, Risk, and Access to
Credit: Evidence from the Mortgage Market,” Syracuse University Working
Paper, 2007.
Gerardi, Kristopher, Adam Shapiro, and Paul Willen, “Subprime Outcomes: Risky
Mortgages, Homeownership Experiences, and Foreclosures,” Federal Reserve
Bank of Boston Working Paper, 2007.
Glaeser, Edward, Joseph Gyourko and Albert Saiz, “Housing Supply and Housing
Bubbles,” Harvard University Working Paper, 2008.
Holmstrom, B., and J. Tirole, “Financial Intermediation, Loanable Funds, and the
Real Sector,” Quarterly Journal of Economics, 112 (1997), 663–691.
Keys, Benjamin, Tanmoy Mukherjee, Amit Seru, and Vikrant Vig, “Did Securiti-
zation Lead to Lax Screening: Evidence from Subprime Loans,” University of
Chicago Working Paper, 2008.
Kiyotaki, Nobuhiro, and John Moore, “Credit Cycles,” Journal of Political Economy,
105 (1997), 211–248.
Mayer, Christopher, and Karen Pence, “Subprime Mortgages: What, Where, and
to Whom?” Federal Reserve Board Working Paper, 2008.
Pence, Karen, “Foreclosing on Opportunity: State Laws and Mortgage Credit,”
Review of Economics and Statistics, 88 (2006), 177–182.
Saiz, Albert, “On Local Housing Supply Elasticity,” Wharton Working Paper, 2008.
Stiglitz, Joseph, and Andrew Weiss, “Credit Rationing in Markets with Imperfect
Information,” American Economic Review, 71 (1981), 393–410.