Document Sample
10ASR10_629-651_massey-2 Powered By Docstoc
					                                                                      American Sociological Review
                                                                      75(5) 629–651
Racial Segregation and the                                            Ó American Sociological
                                                                      Association 2010
American Foreclosure Crisis                                           DOI: 10.1177/0003122410380868

Jacob S. Rugha and Douglas S. Masseya

The rise in subprime lending and the ensuing wave of foreclosures was partly a result of mar-
ket forces that have been well-identified in the literature, but it was also a highly racialized
process. We argue that residential segregation created a unique niche of minority clients who
were differentially marketed risky subprime loans that were in great demand for use in
mortgage-backed securities that could be sold on secondary markets. We test this argument
by regressing foreclosure actions in the top 100 U.S. metropolitan areas on measures of black,
Hispanic, and Asian segregation while controlling for a variety of housing market conditions,
including average creditworthiness, the extent of coverage under the Community Reinvest-
ment Act, the degree of zoning regulation, and the overall rate of subprime lending. We
find that black residential dissimilarity and spatial isolation are powerful predictors of fore-
closures across U.S. metropolitan areas. To isolate subprime lending as the causal mecha-
nism through which segregation influences foreclosures, we estimate a two-stage least
squares model that confirms the causal effect of black segregation on the number and rate
of foreclosures across metropolitan areas. We thus conclude that segregation was an impor-
tant contributing cause of the foreclosure crisis, along with overbuilding, risky lending prac-
tices, lax regulation, and the bursting of the housing price bubble.

segregation, foreclosures, race, discrimination

Four decades after passage of the Fair Hous-      been rising; during the 1990s, Latinos in
ing Act, residential segregation remains          New York and Los Angeles joined African
a key feature of America’s urban landscape.       Americans among the ranks of the hyper-
Levels of black segregation have moderated        segregated (Wilkes and Iceland 2004).
since the civil rights era, but declines are      Although much of the increase in Hispanic
concentrated in metropolitan areas with small     segregation stems from rapid popula-
black populations (Charles 2003). In areas        tion growth during a period of mass immigra-
with large African American communities—          tion, levels of anti-Latino prejudice and
places such as New York, Chicago, Detroit,
Atlanta, Houston, and Washington—declines         a
                                                   Princeton University, Office of Population
have been minimal or nonexistent (Iceland,
Weinberg, and Steinmetz 2002). As a result,
in 2000 a majority of black urban dwellers        Corresponding Author:
                                                  Douglas S. Massey, Princeton University, Office
continued to live under conditions of hyper-      of Population Research, Wallace Hall, Princeton,
segregation (Massey 2004). At the same            NJ 08544
time, levels of Hispanic segregation have         E-mail:
630                                                       American Sociological Review 75(5)

discrimination have also risen in recent years     scholarship on segregation and mortgage
(Charles 2003; Massey 2009; Ross and               lending reveals, however, that racial discrimi-
Turner 2005). In addition, much research           nation occurred at each step in the complex
shows that dark-skinned Latinos experience         chain of events leading from loan origination
higher levels of segregation than do their         to foreclosure (Bond and Williams 2007; Im-
light-skinned counterparts (Denton and Mas-        mergluck 2009; Stuart 2003; Williams, Ne-
sey 1989; Massey and Bitterman 1985; Mas-          siba, and McConnell 2005; Wyly et al.
sey and Denton 1992).                              2006). Specifically, ongoing residential segre-
   During the 1990s, rates of subprime mort-       gation and a historical dearth of access to
gage lending, home equity borrowing, and           mortgage credit in U.S. urban areas combined
home ownership increased among minorities;         to create ideal conditions for predatory lend-
in the context of high segregation, many           ing to poor minority group members in poor
new borrowers were necessarily located in          minority neighborhoods (Been et al. 2009;
minority neighborhoods (Been, Ellen, and           Squires et al. 2009). This racialized the ensu-
Madar 2009; Squires, Hyra, and Renner              ing foreclosure crisis and focused its negative
2009). Williams and colleagues (2005) esti-        consequences disproportionately on black bor-
mate, for example, that subprime lending ac-                                          ´
                                                   rowers and home owners (Hernandez 2009;
counted for 43 percent of the increase in          Oliver and Shapiro 2006; Shapiro, Meschede,
black home ownership during the 1990s and          and Sullivan 2010; Wyly et al. 2006, 2009).
33 percent of the growth in ownership within
minority neighborhoods. As a result, when the
housing bubble burst in 2007 and deflated in       SEGREGATION AND THE
2008 and 2009, the economic fallout was
unevenly spread over the urban landscape
                                                   FORECLOSURE CRISIS
(Immergluck 2008). Given that segregation          High levels of segregation create a natural
concentrates the effects of any economic           market for subprime lending and cause
downturn spatially (Massey and Denton              riskier mortgages, and thus foreclosures, to
1993), the rise in foreclosures hit black and      accumulate disproportionately in racially
Hispanic neighborhoods with particular force       segregated cities’ minority neighborhoods.
(Bromley et al. 2008; Hernandez 2009; Im-          By definition, segregation creates minority-
mergluck 2008; Schuetz, Been, and Ellen            dominant neighborhoods, which, given the
2008).                                             legacy of redlining and institutional discrim-
   Economic studies generally conclude that        ination, continue to be underserved by main-
leveraged refinancing, overbuilding, the col-      stream financial institutions (Renuart 2004;
lapse of home prices, and a poorly regulated       Ross and Yinger 2002). Moreover, the finan-
mortgage market were primarily responsible         cial institutions that do exist in minority
for the rise in foreclosures across metropolitan   areas are likely to be predatory—for exam-
areas (Doms, Furlong, and Krainer 2007; Ger-       ple, pawn shops, payday lenders, and check
ardi, Shapiro, and Willen 2009; Glaeser,           cashing services that charge high fees and
Gyourko, and Saiz 2008; Haughwout, Peach,          usurious rates of interest—so that minority
and Tracy 2008; Khandani, Lo, and Merton           group members are accustomed to exploita-
2009). We argue that the foreclosure crisis        tion and frequently unaware that better serv-
also had significant racial dimensions.            ices are available elsewhere (Immergluck
Although prior research has considered race        and Wiles 1999). Segregation also spatially
as a factor, it was mainly to attribute inter-     concentrates the disadvantages associated
group disparities in defaults and foreclosures     with minority status, such as poverty and job-
to minority group members’ weaker econo-           lessness (Massey and Fischer 2000). When
mic position. A careful reading of recent          the economy stagnated, families in minority
Rugh and Massey                                                                               631

neighborhoods were more likely than others        however, changed the calculus of mortgage
to turn to home equity loans as a means of        lending and made minority households very
maintaining consumption, thereby creating         desirable as clients. Indeed, the spread of
a ready demand for unscrupulous brokers to        mortgage-backed securities during the 1980s
exploit (Sullivan, Warren, and Westbrook          transformed home lending throughout the
2000).                                            United States by splitting apart the origination,
    Under conditions of high residential segre-   servicing, and selling of mortgages into dis-
gation, individual disadvantages associated       crete transactions that made it possible for
with minority status are compounded in space      banks to earn more money quickly by originat-
and amplified in markets that are necessarily     ing and selling loans than by lending money
organized geographically (Dymski and Veitch       and collecting interest payments over time
1992; Immergluck 2008). By concentrating          (Raynes and Rutledge 2003; Sowell 2009).
underserved, financially unsophisticated, and        The advent of securitized mortgages
needy minority group members who are              transformed what had been a bank-based
accustomed to exploitation in certain well-       intermediary credit system into a securities-
defined neighborhoods, segregation made it        based market system (Dymski 2002). In so
easy for brokers to target them when market-      doing, the new financial instruments vastly
ing subprime loans (Stuart 2003). Avery, Bre-     expanded the pool of money available for
voort, and Canner (2008) found that among         lending. Under traditional systems of lend-
mortgage lenders who went bankrupt in             ing, the number of mortgages was limited
2007, black borrowers who received loans in       by the amount of deposits a bank had on
2006 were three times more likely to receive      hand to lend. Under the new system, the vol-
a subprime than a prime loan (74 versus 26        ume of mortgages was no longer limited by
percent) and Hispanics were twice as likely       deposits, but by the number of potential bor-
to receive a subprime than a prime loan (63       rowers and investors’ willingness to pur-
versus 37 percent). By contrast, whites were      chase mortgage-backed securities. The new
slightly more likely to get a prime than a sub-   arrangements thus created a demand on the
prime loan from the same lenders (46 versus       part of banks to expand the pool of
54 percent). Among institutions that did not      borrowers.
go bankrupt in 2007, blacks who borrowed in          Securitized mortgages are not sold whole
2006 were just as likely to receive prime as      but are pooled together and divided into dif-
subprime loans (51 versus 49 percent), under-     ferent shares, or tranches, on the basis of
scoring the discriminatory nature of predatory    risk (Raynes and Rutledge 2003). High inter-
lending practices in the United States.           est mortgages pay more to investors, of
                                                  course, but they also carry more risk; to man-
                                                  age the risk, financial engineers combined dif-
Securitization and Rewards to                     ferent risk tranches into diversified bonds that
Risky Lending                                     could be sold on secondary markets. By mix-
                                                  ing different tranches together, financiers
Residential segregation has always created        could create a salable security with almost
dense concentrations of potentially exploitable   any risk rating and interest rate they wished.
clients in need of capital, but in the 1990s,     In theory, the risk of default by borrowers in
these borrowers’ attractiveness to mortgage       high-risk tranches was offset by the surety
lenders changed. Before the 1980s, lenders        of payments within low-risk tranches, thereby
avoided inner-city minority neighborhoods         yielding a relatively safe investment that rat-
through a combination of fear, prejudice,         ing services beholden to the financiers were
and institutional discrimination (Squires         happy to affirm for a generous fee (Raynes
1994). The invention of securitized mortgages,    and Rutledge 2003).
632                                                        American Sociological Review 75(5)

   Because virtually any mortgage, however         1992; Holloway 1998; Stuart 2003). Discrim-
shaky, could be sold and repackaged as part        inatory real estate practices (e.g., steering)
of a collateralized debt obligation, risky bor-    prevented black and Latino homebuyers
rowers who were formerly shunned by lenders        from accessing better housing and sounder
suddenly became quite attractive. The resul-       loan products in affluent suburbs (Friedman
tant wave of predatory lending was spear-          and Squires 2005; Hanlon 2010) and chan-
headed by independent mortgage brokers             neled them into depressed inner-ring suburbs
who did not bear the risk of their reckless        that were undergoing sustained disinvest-
lending practices. They simply generated           ment (Hackworth 2007).
mortgages and immediately sold them to                 In a very real way, as Williams and col-
banks and other financial institutions, which      leagues (2005) show, the old inequality in
in turn capitalized the shaky subprime instru-     home lending made the new inequality possi-
ments as securities and sold them to third-        ble by creating geographic concentrations of
party investors who ended up assuming the          underserved, unsophisticated consumers that
risk, typically in ways they neither appreciated   unscrupulous mortgage brokers could easily
nor understood (Engel and McCoy 2007;                                                           ´
                                                   target and efficiently exploit (see also Hernan-
Lewis 2010; Peterson 2007). These lucrative        dez 2009; Lee 1999). A study of subprime
subprime lending and securitization practices      borrowers in Los Angeles, Oakland, Sacra-
did not suddenly appear ‘‘at the fringes of        mento, and San Diego found that African
finance,’’ but were produced and legitimated       Americans were significantly more likely
by the financial industry using new, high-         than whites (40 versus 24 percent) to report
tech tools such as credit scoring, risk-based      lender marketing efforts as the impetus for
pricing, securitization, credit default swaps,     taking out a home equity loan (California
and variable rate mortgages that were billed       Reinvestment Committee 2001). In the end,
as rational, scientific, and safe (Langley         subprime lending not only saddled borrowers
2008, 2009; Stuart 2003).                          with onerous terms and unforeseen risks, but
                                                   it also reinforced existing patterns of racial
                                                   segregation and deepened the black-white
How Segregation Shaped Unequal                     wealth gap (Bond and Williams 2007; Fried-
Lending                                            man and Squires 2005; Williams et al. 2005).
                                                       In the new regime of racial inequality,
With the move to securitized lending, dis-         African American and Latino homeowners
crimination in real estate lending shifted         bore a disproportionate share of costs stem-
from the outright denial of home loans to          ming from the bursting of the housing bubble.
the systematic marketing of predatory loans        Compared with whites with similar credit pro-
to poor black and Hispanic households,             files, down payment ratios, personal charac-
which were easily found within segregated          teristics, and residential locations, African
neighborhoods (Engel and McCoy 2008;               Americans were much more likely to receive
Massey 2005a). Before the subprime boom,           subprime loans (Avery, Brevoort, and Canner
black borrowers were more likely to be             2007; Avery, Canner, and Cook 2005; Bocian,
denied loans overall, especially in white          Ernst, and Li 2006; Pennington-Cross, Yezer,
areas, whereas whites were often denied            and Nichols 2000). Moreover, after control-
loans in minority neighborhoods (Holloway          ling for background factors, black and His-
1998). During the boom, minority borrowers’        panic homeowners were significantly more
underserved status made them prime targets         likely than whites to receive loans with unfa-
for subprime lenders who systematically tar-       vorable terms such as prepayment penalties
geted their communities for aggressive             (Bocian et al. 2006; Farris and Richardson
marketing campaigns (Dymski and Veitch             2004; Quercia, Stegman, and Davis 2007;
Rugh and Massey                                                                                633

Squires 2004), higher cost ratios (Elliehausen,     Race and the Housing Bust
Staten, and Steinbuks 2008; LaCour-Little and
Holmes 2008), and higher rate spreads (Bo-          From a lender’s perspective, the new system
cian et al. 2006). The racial gap in subprime       worked well as long as real estate and lend-
lending holds across all income levels (Brom-       ing markets remained liquid and housing pri-
ley et al. 2008; Immergluck and Wiles 1999;         ces continued to rise. For a while, minority
Williams et al. 2005), with perhaps an              ownership rates rose and everyone involved
increase at higher income levels (Institute on      in securitized lending made good money—
Race and Poverty 2009; Williams et al. 2005).       the broker who originated the loan, the lender
    During the 1990s, the United States was         who put up the money, the firm that pack-
increasingly characterized by a dual, racially      aged and underwrote the mortgage-backed
segmented mortgage market, one that was             security, the rating agency that affirmed its
structured by the race of borrowers and the         creditworthiness, and the company that
racial composition of neighborhoods (Apgar          insured the investments through novel instru-
and Calder 2005; Stuart 2003; U.S. Depart-          ments known as credit default swaps. To
ment of Housing and Urban Development               keep the system in operation and profits roll-
2000). Controlling for neighborhood charac-         ing in, however, more borrowers had to be
teristics, the incidence of subprime lending        constantly found and housing prices had to
was significantly greater among black and           continue rising. As the number of buyers
Hispanic borrowers (Calem, Hershaff, and            increased and housing prices inflated, a spec-
Wachter 2004) and among people who had              ulative fever took hold (Andrews 2009; Shil-
not gone to college (Manti, Raca, and Zorn          ler 2008). People began to finance home
2004). As a result, from 1993 to 2000, the          purchases on the assumption that home
share of subprime mortgages going to house-         prices would rise indefinitely (Khandani et
holds in minority neighborhoods rose from 2         al. 2009), buying properties with interest-
to 18 percent (Williams et al. 2005).               only loans, waiting for real estate prices to
    The rise of racially targeted subprime lend-    rise, and then ‘‘flipping’’ the properties to
ing destabilized minority neighborhoods by          realize the capital gain. Research shows
increasing turnover (Gerardi and Willen             that minority-owned properties were more
2008), and the destabilizing effects on home-       likely than others to be involved in loan flip-
ownership did not remain confined to minor-         ping and equity stripping schemes (Immer-
ity neighborhoods, but spilled over into            gluck and Wiles 1999).
adjacent white and mixed neighborhoods                 After 2004, as the market peaked, specula-
(Schuetz et al. 2008). National evidence            tors, dubious lenders, negligent securities
from a longitudinal study of homeowners             dealers, and compromised ratings firms
from 1999 to 2005 shows that the racial gap         increasingly focused on a select few booming
in homeownership exit rates widened in              regional markets, such as Las Vegas, Phoenix,
a way that cannot be explained by social, eco-      and South Florida. These markets appear to
nomic, or financial factors that historically ac-   have played a large role in the final run-up
counted for black-white differentials (Turner       in risky lending. At the height of the bubble,
and Smith 2009). The coincidence of the             fraud may have been the rule rather than the
peak in subprime lending with the inexplica-        exception. When Pendley, Costello, and
ble decline in the stability of black home own-     Kelsch (2007:4) analyzed a sample of delin-
ership and exit rates offers compelling             quent subprime loans made in 2006, for exam-
evidence that segregation and the new face          ple, they found widespread ‘‘appearance of
of unequal lending combined to undermine            fraud or misrepresentation’’; two-thirds of
black residential stability and erode any accu-     stated owner-occupied dwellings were never
mulated wealth (Shapiro et al. 2010).               actually occupied, and nearly half of all
634                                                        American Sociological Review 75(5)

claims to first-time ownerships did not appear     Peterson 2007; Powell 2010; Squires et al.
to be valid. These findings suggest that the in-   2009).
flated housing bubble motivated sophisticated         Two additional lines of research also sug-
lenders, brokers, and buyers to engage in an       gest that predatory lending and subsequent
unparalleled level of deceit.                      asset stripping were structured on the basis
    As the market became fully saturated in        of race as well as class. First, contrary to con-
2006, housing prices ultimately stalled, fore-     ventional wisdom, the housing crisis was not
closures rose, and faith in mortgage-backed        caused primarily by a decline in underwriting
securities evaporated, bringing down the           standards (Haughwout 2009; see also Khan-
entire system of collateralized debt obliga-       dani et al. 2009). Bhardwaj and Sengupta
tions and taking much of the U.S. economy          (2009) show, for example, that average credit
with it (Khandani et al. 2009; Shiller 2008).      scores within the subprime loan sector actu-
Ultimately, brokers, lenders, securitizers, and    ally increased in years prior to the housing
most of all, ratings agencies, failed to foresee   bust and that most of the loans fell into the
the perils of ‘‘default correlation,’’ or the      near-prime, low- or no-documentation ‘‘Alt-
interrelated risks bound up in interconnected      A’’ category, rather than in more speculative
portfolios of troubled loans (Langley 2009).       B and C categories. Second, the crisis cannot
The utter collapse of subprime lending             be attributed to riskier lending engendered by
exposed the extremes of a pricing regime           the Community Reinvestment Act (CRA).
that assessed risks individually but not collec-   Using a regression discontinuity design,
tively, not accounting for the aggregate risk of   Bhutta (2008) examined lending rates just
ever-increasing subprime lending and securiti-     below and above the CRA neighborhood
zation. Less than 1 percent of mortgage loans      income cutoff and found that while CRA
were in foreclosure at the end of 2005; this       oversight did increase lending in targeted
rate more than quadrupled to over 4.6 percent      areas, unregulated lending activity also
by the end of 2009. At the same time, the fore-    increased substantially in the same places.
closure rate on riskier subprime loans went        Only 6 percent of subprime loans were
from 3.3 percent in 2005 to 15.6 percent in        made to low-income borrowers or individuals
2009 (Mortgage Bankers Association 2006,           in neighborhoods subject to CRA oversight,
2010).                                             and less than 2 percent of loans that originated
    The resulting tidal wave of foreclosures       with unregulated independent mortgage
was concentrated in areas that only a few          brokers were CRA credit-eligible (Bhutta
years earlier had been primary targets for the     and Canner 2009).
marketing of subprime loans (Bond and Wil-            Although lending standards often left much
liams 2007; Edmiston 2009), and minority           to be desired, it appears that ongoing racial
neighborhoods often bore the brunt of the          segregation, discriminatory lending, and an
foreclosures (Bromley et al. 2008; Edmiston        overheated housing market combined to leave
2009; Hernandez 2009; Immergluck 2008;             minority group members uniquely vulnerable
Institute on Race and Poverty 2009; Mallach        to the housing bust. As Immergluck (2009)
2009; Marquez 2008). In the end, the housing       wryly notes, financial literacy and creditwor-
boom and the immense profits it generated          thiness did not suddenly plummet on the eve
frequently came at the expense of poor minor-      of the crisis—home prices did. At the same
ities living in central cities and inner suburbs   time, although CRA regulations stimulated
who were targeted by specialized mortgage          lending to minority households in low-income
brokers and affiliates of national banks and       neighborhoods, the increase was not nearly
subjected to discriminatory lending practices      enough to bring about the housing crisis
(Been et al. 2009; Engel and McCoy 2008;           (Bhutta and Canner 2009; Park 2008).
Rugh and Massey                                                                              635

DATA SOURCES                                     members to achieve an even residential distri-
                                                 bution; the latter gives the proportion of own-
From the above research, we conclude that        group members in a tract inhabited by the
residential segregation was significant in       average minority group member. Of the five
structuring how the rise of predatory lending    dimensions of segregation identified by Mas-
and the consequent wave of foreclosures          sey and Denton (1988), evenness and isolation
played out in U.S. housing markets (Bond         are the most important and empirically
and Williams 2007; Engel and McCoy               account for most of the common variation
2008; Hernandez 2009; Stuart 2003; Wil-          (Massey, White, and Phua 1996).
liams et al. 2005; Wyly et al. 2006, 2009).          Our analytic strategy is to regress the num-
We hypothesize, first, that segregation facil-   ber and rate of foreclosures in MSAs on these
itated racially targeted subprime mortgage       two measures of segregation while holding
lending during the boom and, second, that        constant the effect of metropolitan-level fac-
it magnified the consequences of the housing     tors shown to influence the odds of foreclo-
crisis for blacks and Latinos by concentrating   sure. Our controls include standard census-
foreclosures in poor minority neighborhoods      based variables, such as 2008 MSA popula-
during the bust. To test these assertions, we    tion and racial and ethnic composition from
draw on two principal sources of data. We        2000, as well as socioeconomic characteristics
define our dependent variables based on          obtained from the 2005 to 2007 American
data obtained from RealtyTrac, the nation’s      Community Surveys, such as percent of per-
largest provider of foreclosure listings. We     sons holding a college degree, median house-
compute the number of properties with            hold income, and percent of homeowners with
at least one foreclosure action in 2006,         a second mortgage (all defined as of 2006).
2007, and 2008 in the nation’s 100 largest       We also include the median age of MSA
metropolitan statistical areas and divisions     housing stock from the 2000 housing census.
(MSAs), as defined in 2003. We then divide       The 2006 unemployment rate and the 2000
this figure by the number of housing units in    to 2006 change in rate of annual unemploy-
2006 to derive a foreclosure rate.1 Over 77      ment come from the Bureau of Labor Statis-
percent of all foreclosure properties during     tics; the share of the workforce that was
the period under consideration were located      unionized comes from Hirsch and MacPher-
in this set of MSAs. This sample is especially   son (2007). In initial specifications of the
useful for making inferences about minorities    model we also tested for effects of the poverty
in the United States because these 100 MSAs      rate, share of female-headed households, and
were home to over 75 percent of African          per capita income, but these variables add
Americans, nearly 80 percent of Hispanics,       nothing to the explanatory power of the mod-
and 90 percent of Asians in the country in       els and were dropped from further consider-
2000.                                            ation. The final models also include dummy
    Our principal explanatory variables are      variables to control for region, coastal loca-
measures of residential unevenness and spatial   tion, and location along the Texas border
isolation for Hispanics, African Americans,      with Mexico (the Rio Grande River).
and Asians computed across census tracts in          In addition to these standard indicators, we
the top 100 metropolitan areas (see Iceland      include three specialized measures of condi-
et al. 2002). We measured unevenness with        tions in metropolitan real estate markets that
the well-known index of dissimilarity and iso-   prior research finds to be important in explain-
lation using the P* within-group isolation       ing the foreclosure crisis. First, we construct
index. The former gives the relative propor-     a measure of overbuilding by computing the
tion of minority group members who would         ratio of 2000 to 2006 MSA housing starts to
have to exchange tracts with majority group      housing units in 2000, a procedure closely
636                                                       American Sociological Review 75(5)

following that of Glaeser and colleagues          intentionally left blank in fields) and is the
(2008). Second, we include the Wharton Res-       definition of subprime lending used in other
idential Land Use Regulation Index as a mea-      research (e.g., Been et al. 2009; Bocian
sure of local land use planning regulation.       et al. 2006; Squires et al. 2009).2
Gyourko, Summers, and Saiz (2008) devel-             To measure the degree of regulatory over-
oped this index for different municipalities;     sight, we draw on 2004 to 2006 HMDA data
following Rothwell and Massey (2009), we          and compute the weighted share (by dollar
take the weighted average of the index for        amount) of all 2004 to 2006 loans in each
municipalities within each MSA that responds      MSA that were originated by CRA-covered
to Gyourko’s survey. We then center the dis-      lending banking institutions (following Fried-
tribution on the mean for the top 100 areas to    man and Squires 2005). In doing so, we use
yield a measure that controls for housing price   HMDA data on conventional loans (i.e., those
premiums stemming from regulatory con-            not guaranteed by government) for the pur-
straints on housing supply (Fischel 1990;         chase of single-family properties (1 to 4 units)
Glaeser 2009; Glaeser and Gyourko 2003,           and take a 20 percent extract of the nearly 28
2008; Glaeser et al. 2008). Finally, we include   million mortgage loans originated in the top
a measure of the housing price boom in each       100 metropolitan areas and compute the share
MSA by using the Federal Housing Finance          of subprime loans going to borrowers. On
Agency’s quarterly metropolitan area House        average, less than two-thirds of total lending
Price Index (HPI), which is a weighted,           per MSA fell under the ambit of federal
repeat-sales index based on transactions          CRA regulation, but this share is noticeably
involving single-family homes. We bench-          lower, about half or less, in MSAs with ele-
mark the housing price boom for each MSA          vated foreclosures (e.g., Detroit, Michigan;
by dividing annualized change in HPI from         Las Vegas, Nevada; and Bakersfield, Califor-
2000 to 2006 by the annualized change in          nia). This suggests that greater CRA lending
the two decades prior to 2000.                    oversight reduces foreclosures even as it
    We also include MSA-level controls for        makes more loans available to minority group
the degree of subprime lending, the extent of     members (Bond and Williams 2007; Friedman
CRA regulatory oversight, and borrowers’          and Squires 2005).
average creditworthiness. To assess the preva-       Finally, to control for borrowers’ credit-
lence of subprime lending, we compute the         worthiness, we compute the average consumer
aggregate share of all loans originated in        credit score at the MSA level using informa-
2004, 2005, and 2006 that were subprime,          tion obtained from the Experian National
drawing on data from the Federal Financial        Score Index. The index ranges from 672 to
Institutions Examination Council (http://         720 for the top 100 MSAs and constitutes The FFIEC tabulates          the mean FICO (Fair Isaac Corporation) credit
information that lending institutions must pro-   score for all consumers living in all counties
vide to the federal government under the          of each MSA whose credit behavior is re-
Home Mortgage Disclosure Act (HMDA).              ported to Experian, one of the nation’s three
These data cover more than 28 million loans       major credit reporting bureaus. We expect
originated during the peak years of the hous-     that higher overall MSA credit scores will
ing boom. More than 6.9 million loans, or         be associated with lower foreclosure rates
nearly one in four, were subprime, meaning        because of the higher aggregate creditworthi-
that the interest rate at origination exceeded    ness of borrowers, and by extension, mortgage
that for a comparable U.S. Treasury security      holders. To mitigate omitted variables bias,
(e.g., a 30-year bond) by 3 percent or more.      we include this variable as a proxy for differ-
This is the cutoff for reporting a loan in the    ences in credit history and access that co-vary
data files (rates of lower priced loans are       with racial composition and segregation.
Rugh and Massey                                                                                 637

    Table 1 presents means and standard devi-       to create areas of concentrated disadvantage
ations of the foregoing variables. In terms of      that constitute prime targets for subprime
our leading explanatory variable, levels of         lending. For privileged minorities, segregation
segregation are highest for African Ameri-          has the opposite effect—it concentrates
cans, lowest for Asians, and Hispanics are          advantage and its correlates to produce areas
in-between. The index of dissimilarity, for         that are unlikely targets for subprime lending.
example, averages .59 for blacks with a range       As of 2008, the poverty rate for Asians stood
of plus or minus two standard deviations that       at 12 percent, compared with 25 percent
goes from .34 to .83. By convention, dissimi-       among African Americans and 23 percent
larities below .3 are considered low, those         among Hispanics (DeNavas-Walt, Proctor,
from .3 to .6 are considered moderate, and          and Smith 2009). Likewise, the median per
those above .6 are regarded as high, with any-      capita income for Asians was $30,000 in
thing above .75 considered extremely high.          2008, compared with just $18,000 for Afri-
Under these criteria, black segregation runs        can Americans and $16,000 for Hispanics.
from moderate to extremely high. By contrast,       The Asian median income even exceeded
the comparable range for Asians is .25 to .52       that for non-Hispanic whites ($28,500). To
with a mean of .39, and Hispanics have a range      the extent that segregation has an effect
of .24 to .67 with a mean of .46. Asians range      on the economic geography of Asians, it
from low to moderate, and Hispanics range           would be to concentrate advantage and
from low to high.                                   thus diminish the frequency of subprime
    Similar patterns prevail with respect to the    lending and foreclosures.3
isolation index. The black isolation index
averages .45 and ranges from .07 to .84, going
from minus to plus two standard deviations;         SEGREGATION’S EFFECT ON
the Hispanic index averages .32 and ranges
from zero to .75; and the Asian mean stands
at a very low .14 and ranges from zero to           Despite the persistence of residential segre-
.40. If segregation creates a natural niche for     gation and its prevalence in areas with large
subprime lending, then the more extreme             minority populations, racial segregation is
maxima and greater variability of black segre-      no longer as universal across U.S. urban
gation measures hold by far the greatest            areas as it once was. As indicated by the
potential to predict foreclosures, followed by      wide variance just described, there is now
Hispanic segregation measures. The generally        considerable variation across metropolitan
low averages and restricted variation of isola-     areas in the degree of black and Hispanic
tion and dissimilarity among Asians suggest         segregation (Charles 2003; Iceland et al.
a more limited potential to influence inter-        2002; Massey, Rothwell, and Domina
metropolitan variation in foreclosures.             2009). In addition to a small minority popu-
    Segregation’s effect in creating fertile ter-   lation, other characteristics that predict lower
rain for subprime lending also depends on           levels of residential segregation are a small
a group’s socioeconomic status. This general-       urban population, newer housing stock, pres-
ization holds because segregation not only          ence of a college or a university, proximity to
concentrates minority group members spa-            a military base, higher socioeconomic status,
tially within particular neighborhoods, but it      and location in the West or the Southwest
also concentrates any characteristic associated     (Farley and Frey 1994). Inter-urban differen-
with minority group status (Massey and Den-         tials in Hispanic and black segregation thus
ton 1993; Massey and Fischer 1999). For             carry the potential to contribute significantly
underprivileged minorities, segregation spa-        to variation in foreclosure rates across U.S.
tially concentrates poverty and its correlates      metropolitan areas.
638                                                      American Sociological Review 75(5)

Table 1. Summary Statistics for Data Used in Analysis of Foreclosures in Top 100 MSAs

Variables                                                              Mean               SD

Outcome Measures
  Foreclosures 2006 to 2008                                            33, 633           38, 710
  Foreclosure Rate (per 100)                                            4.100             3.031
Index of Dissimilarity
  African Americans                                                      .587               .121
  Hispanics                                                              .455               .108
  Asians                                                                 .385               .068
P* Isolation Index
  African Americans                                                      .453               .193
  Hispanics                                                              .317               .215
  Asians                                                                 .137               .133
Control Variables
  Ratio of 2000 to 2006 Housing Starts to 2000 Units                     .121              .073
  Wharton Land Use Regulation Index                                      .000              .757
  Relative Change in Housing Price Index (HPI) after 2000               4.068             2.534
  Percent Subprime Loans, HMDA (2004 to 2006)                          24.230             5.826
  Percent Loans Made by CRA-Covered Lenders (2004 to 2006)             64.729             6.799
  Experian MSA Credit Score Index                                     691.560            14.259
  Population (2008)                                                 1,891, 540        1,787, 775
  Percent Persons 251 with College Degree (2006)                       29.473             6.783
  Percent African American (2000)                                      13.592             9.618
  Percent Hispanic (2000)                                              13.292            15.861
  Percent Asian (2000)                                                  4.834             7.297
  Median Household Income (2006)                                       54, 697           10, 821
  Percent Homeowners with Second Mortgage (2006)                        7.346             2.153
  Percent Workforce Unionized (2006)                                   12.075             6.593
  Unemployment Rate (2006)                                              4.598             1.022
  Change in Unemployment Rate 2000 to 2006                               .755             1.026
  Median Age of Housing Stock, Years (2000)                            31.310             9.370
    Northeast                                                            .230               .423
    Midwest                                                              .180               .386
    South                                                                .370               .485
    West                                                                 .220               .416
  Costal MSA                                                             .380               .488
  Borders Rio Grande                                                     .020               .141

Note: N = 100. See text for data sources.

   To assess the effect of segregation on fore-   of control variables with associated coeffi-
closures, we estimate a simple OLS multiple       cients b2, and e is the error term.4 Table 2
regression model defined by the equation:         presents estimates of equations that use dis-
                                                  similarity and isolation indices to predict the
            F ¼ a þ b1 S þ b 2 Z þ e        ð1Þ   log of total foreclosures by MSA; Table 3
                                                  shows estimates that predict the log of fore-
where F represents either the log of the total    closure rates by MSA.
number of foreclosure filings or the log of          In both tables the models fit the data
foreclosures per housing unit, a is the inter-    extremely well, with the predictor variables
cept, S is a vector of segregation measures       explaining, on average, 77 percent of the var-
with associated coefficients b1 , Z is a vector   iance in the log of the foreclosure rate and 90
Rugh and Massey                                                                                       639

Table 2. OLS Estimates of Effect of Selected Measures of Residential Segregation on Log of
Total Foreclosures

                                                     Dissimilarity Index                Isolation Index
Variables                                             B                SE               B             SE

Index of Segregation
  African Americans                              3.718**               .725          2.122**          .619
  Hispanics                                      2.773                 .596           .080            .656
  Asians                                        22.080*                .920         22.161           1.636
Control Variables
  Housing Starts Ratio                           2.980**               .960          3.067**         1.077
  Wharton Land Use Index                          .250**               .082           .272**          .096
  Change in Housing Price Index                   .082**               .024           .092**          .029
  CRA-Covered Lending Share                     21.295                 .912          2.810           1.061
  Subprime Loan Share                            3.022*               1.353          4.310**         1.581
  MSA Credit Score Index                         2.015*                .007          2.016*           .007
  Log of Population                              1.008**               .089          1.013**          .093
  Percent with College Degree                   21.341                1.315          2.997           1.459
  Log Median Household Income                     .253                 .509           .340            .515
  Percent with Second Mortgage                    .751                3.687           .225           4.350
  Percent Workforce Unionized                    2.025**               .011          2.022*           .011
  Unemployment Rate                              2.010                 .064           .012            .071
  Change in Unemployment Rate                     .245**               .052           .213**          .063
  Age of Housing Stock                            .004                 .012           .014            .013
    Midwest                                       .434*                .200           .631**          .200
    South                                         .042                 .257           .081            .296
    West                                          .463                 .384           .679            .436
  Coastal MSA                                    2.053                 .123           .070            .133
  Borders Rio Grande                            21.030**               .370         21.054**          .380
Constant                                         1.960                7.557           .979           8.150
R2                                                .91                                 .90
Joint F-Test for Region                          3.35*                               7.97**
Joint F-Test for Segregation                    10.48**                              6.28**

Note: N = 99. Robust standard errors. Model also includes percent black, percent Hispanic, and percent
*p \ .05; **p \ .01 (two-tailed tests).

percent of the variance in the logged absolute            in the segregation of Hispanics, however,
number of foreclosures. Moreover, coeffi-                 does not consistently affect the rate and abso-
cients for the segregation indices closely                lute number of foreclosures. As we discuss in
follow theoretical expectations. Whether mea-             the next section, the effect of Hispanic segre-
sured in terms of residential dissimilarity or            gation may be overwhelmed by the effect of
spatial isolation, segregation of African Amer-           black segregation. These statistical estimates
icans is a powerful and highly significant pre-           are consistent with the findings of Been and
dictor of the number and rate of foreclosures             colleagues (2009) and Squires and colleagues
across U.S. metropolitan areas. For instance,             (2009).
a .1 unit increase in black dissimilarity is asso-           By contrast, Asian-white residential dis-
ciated with 37 percent more foreclosure ac-               similarity significantly reduces the number
tions and a 34 percent increase in the                    and rate of foreclosures across metropolitan
foreclosure rate. Inter-metropolitan variation            areas, and spatial isolation also has a negative,
640                                                            American Sociological Review 75(5)

Table 3. OLS Estimates of Effect of Selected Measures of Residential Segregation on the Log
of the Foreclosure Rate

                                                   Dissimilarity Index               Isolation Index
Variables                                           B                SE              B             SE

Index of Segregation
  African Americans                             3.412**              .725         1.990**          .604
  Hispanics                                     2.686                .580          .074            .642
  Asians                                       21.7611               .906        21.572           1.562
Control Variables
  Housing Starts Ratio                          2.868**              .946         2.973**         1.032
  Wharton Land Use Index                         .236**              .081          .259**          .092
  Change in Housing Price Index                  .072**              .023          .080**          .027
  CRA-Covered Lending Share                    21.331                .901         2.847           1.026
  Subprime Loan Share                           3.269*              1.358         4.560**         1.547
  MSA Credit Score Index                        2.0141               .007         2.015*           .007
  Log of Population                              .012                .087          .010            .089
  Percent with College Degree                  21.746               1.291        21.420           1.410
  Log Median Household Income                    .750                .489          .8581           .500
  Percent with Second Mortgage                  1.468               3.643          .703           4.190
  Percent Workforce Unionized                   2.027*               .010         2.024*           .011
  Unemployment Rate                              .016                .063          .031            .068
  Change in Unemployment Rate                    .224**              .051          .200**          .060
  Age of Housing Stock                           .003                .012          .012            .013
     Midwest                                     .409*               .194          .583**          .192
     South                                       .032                .251          .072            .286
     West                                        .448                .372          .663            .422
  Coastal MSA                                   2.077                .124          .030            .129
  Borders Rio Grande                           21.040*               .392        21.042*           .395
Constant                                        1.121               7.443          .098           8.005
R2                                               .78                               .76
Joint F-Test for Region                         3.17*                             7.44**
Joint F-Test for Segregation                    8.71**                            5.62**

Note: N = 99. Robust standard errors. Model also includes percent black, percent Hispanic, and percent
 p \ .10; *p \ .05; **p \ .01 (two-tailed tests).

although insignificant, effect. Residential seg-           The various control variables generally
regation concentrates group characteristics,            function as hypothesized, lending face validity
whatever they are, in space. In the case of             to the equation estimates. As expected, fore-
blacks, the prevailing characteristics are pov-         closures are positively predicted by greater
erty and socioeconomic deprivation; given               housing start shares, higher rates of subprime
that Asian incomes exceed those of whites,              lending, increasing unemployment, rising
on average, their prevailing characteristics            home prices, lower credit scores, more second
are affluence and socioeconomic privilege.              mortgages, higher median incomes, greater
By concentrating affluence, the segregation             levels of land use regulation, and location in
of Asians creates communities that are inher-           the Midwest or the West.5 If we assume that
ently resistant to the entreaties of subprime           the housing start ratio measures overbuilding
lenders—hence the negative relationship                 and that rising relative home prices capture
between Asian segregation and foreclosures.             the housing bubble, then our results are
Rugh and Massey                                                                                641

consistent with prior work on the economic         and a .66 percentage point increase in the
causes of the U.S. foreclosure crisis (e.g.,       foreclosure rate.
Glaeser et al. 2008; Haughwout et al. 2008;
Immergluck 2009; Kochhar, Gonzalez-Bar-
rera, and Dockterman 2009; Mayer, Pence,           FORECLOSURES AND THE
and Sherlund 2009; National Association of         SEGREGATION-SUBPRIME
Realtors 2004).
   In the attempt to understand the foreclo-
sure crisis, our study adds the important          Taking into account the distribution of effect
and independent role played by racial segre-       sizes estimated in Table 4, we conclude that
gation in structuring the housing bust. Table      the influence of black residential segregation
4 shows segregation’s relative predictive          clearly exceeds that of other factors linked by
power compared with other significant fac-         earlier studies to inter-metropolitan variation
tors by reporting the standardized effect          in foreclosures. Furthermore, racial segrega-
sizes evaluated at the sample mean of the          tion is an important and hitherto unappreci-
foreclosure total and rate (in terms of num-       ated contributing cause of the current
ber of foreclosures and percentage points,         foreclosure crisis. This conclusion rests, of
respectively). In the model using dissimilar-      course, on a cross-sectional ecological
ity indexes, a standard deviation increase in      regression and thus may be subject to certain
the segregation of African Americans in-           methodological criticisms. Because we are
creases the number of foreclosures by              not seeking to infer individual behavior
15,028 actions and the rate of foreclosures        from aggregate data, ecological bias itself is
by 1.68 percentage points. This effect ex-         not an issue—our argument is structural
ceeds the effect of MSA home building,             and specified at the metropolitan, not the
house price booms, and all other important         individual, level.
explanatory variables. Changes in the proxy            As with any cross-sectional analysis, how-
measures of economic conditions, land use          ever, endogeneity or reverse causality is
restrictions, and overbuilding exert a consid-     a potential problem. In this case, it does not
erably smaller impact on foreclosures, with        seem likely that foreclosures could reasonably
absolute standard effect sizes just 40 to 56       cause segregation. Patterns of racial segrega-
percent of that for black segregation.             tion are the cumulative product of decades
   In the isolation model, one standard devia-     of actions in the public and private spheres,
tion in black segregation leads to a large         and high levels of black segregation were
change in foreclosures (13,842) and the fore-      well institutionalized in U.S. urban areas by
closure rate (1.58 percentage points). Stan-       the mid-twentieth century (Massey and Den-
dardized changes in the subprime lending           ton 1993). In addition, we measure segrega-
share lead to an increase of nearly 8,500 fore-    tion in 2000 and foreclosures in 2006 to
closures and an even greater effect on foreclo-    2008, so the independent variable is tempo-
sure rates, at 1.1 percentage points. Relative     rally prior to the dependent variable.
changes in house prices and housing starts,            A more serious threat to causal inference is
average credit scores, and changes in unem-        endogeneity. Perhaps there is a third, unmea-
ployment rates show an increase of 7,000 to        sured variable that influences both segregation
8,000 foreclosures and .8 to .9 percentage         and foreclosures to bring about the observed
points in the foreclosure rate. While the effect   association between them. Although we
size of Hispanic segregation in the dissimilar-    endeavored to apply a rather exhaustive set
ity model is not statistically distinguishable     of controls, it is simply not possible to control
from zero, Hispanic isolation has a standard-      for all potential confounding variables. One
ized effect of more than 3,800 foreclosures        possible confounding variable is the degree
642                                                           American Sociological Review 75(5)

Table 4. Effect of a One Standard Deviation Increase in Selected Variables on the Number
and Rate of Foreclosures

                                                   Number of Foreclosures            Foreclosure Rate
Effect of One SD Increase in:                         (Mean: 33,947)                 (Mean: 4.135%)

Dissimilarity Model
  Black Dissimilarity Index                               15,028                          1.680
  Hispanic Dissimilarity Index                             n.s.                            n.s.
  Asian Dissimilarity Index                               24,830                          2.499
  Housing Starts Ratio                                     7,392                           .867
  Wharton Land Use Index                                   6,208                           .713
  Ratio of post- to pre-2000 HPI                           7,094                           .762
  Subprime Loan Share                                      5,944                           .871
  MSA Credit Score Index                                  27,301                          2.817
  Change in Unemployment Rate                              8,383                           .933
Isolation Model
  Black Isolation Index                                   13,842                          1.581
  Hispanic Isolation Index                                 n.s.                            n.s.
  Asian Isolation Index                                    n.s.                            n.s.
  Housing Starts Ratio                                     7,615                           .899
  Wharton Land Use Index                                   6,767                           .784
  Ratio of post- to pre-2000 HPI                           7,939                           .840
  Subprime Loan Share                                      8,477                          1.092
  MSA Credit Score Index                                  27,817                          2.878
  Change in Unemployment Rate                              7,276                           .830

Note: N = 99. Effect on foreclosure rate shown in percentage points.

of anti-black prejudice and discrimination,           foreclosures. Although simple logic predicts
which could well vary across metropolitan             a strong relationship between the overall prev-
areas and simultaneously increase segregation         alence of subprime lending and foreclosures,
and the extent of racially targeted subprime          there is no a priori reason to believe that the
lending, thus increasing the number and rate          black-white or Hispanic-white gap in the
of foreclosures. Indeed, Galster (1986) and           extent of subprime lending will affect these
Galster and Keeney (1988) show that discrim-          outcomes; according to our argument, how-
ination in lending had a strong effect on racial      ever, the size of the racial gap in subprime
segregation across 40 MSAs in the 1970s and           lending should clearly be causally related to
1980s.                                                the degree of black segregation.
   The relationship between segregation and               In a preliminary examination of the data
foreclosures can be purged of endogeneity             we indeed found that inter-metropolitan varia-
using two-stage least squares, but only if            tion in the size of the racial gap in subprime
a suitable instrument is available. Because           lending is strongly correlated with segregation
we argue that segregation facilitates subprime        but uncorrelated with either the rate or the
lending to African Americans, we should be            number of foreclosures. This confirms its suit-
able to use inter-metropolitan variation in           ability as an instrument. According to Angrist
the size of racial differentials in subprime          and Krueger (2001:73), ‘‘a good instrument is
lending to isolate segregation’s causal effect.       correlated with the endogenous regressor for
Specifically, if our argument is correct, then        reasons the researcher can verify and explain,
intergroup differentials in subprime lending          but uncorrelated with the outcome variable for
offer a suitable instrument to predict segrega-       reasons beyond its effect on the endogenous
tion in a two-stage least squares model of            regressor.’’ We compute intergroup differentials
Rugh and Massey                                                                                643

in subprime lending by metropolitan area           equal to 3 percent. The probit model expresses
using the combined HMDA data from                  the likelihood of receiving a subprime loan as
2004, 2005, and 2006 (described in the             a function of the type (i.e., home purchase,
Data Sources section). If lending discrimi-        refinance, or improvement) and amount of
nation is greater in more segregated MSAs,         the loan, borrower income, first or second
then racial-ethnic differentials in subprime       lien status, occupancy (i.e., investor or
lending permit us to identify the causal           owner), type of loan purchaser (i.e., govern-
effect of residential segregation on MSA           ment agency, private, bank, finance company,
foreclosure rates, enabling us to specify          lender affiliate, or other independent entity),
the following two-stage model:                     median tract income and tract-to-MSA ratio,
                                                   ratio of total tract single-family units to popu-
    S ¼ h þ dRACEDIFF þ W l þ n             ð2Þ    lation, and tract minority percentage.
                                                       We also merge the following extended
                                                   control variables to the foregoing data com-
       F ¼ a þ ðh þ RACEDIFFd                      puted from the HMDA data: tract population
            þ W l þ nÞb1 þ Zb2 þ e:                density in persons per square mile, median
                                                   age of tract housing stock, and the MSA-level
    In this system, Equation 2 expresses the       average credit score index, described earlier.
first-stage relationship between segregation,      The probit estimation clusters errors at the
S, and RACEDIFF, the black-white or                MSA level. Avery and colleagues (2005)
Hispanic-white gap in the likelihood of ob-        show that HMDA data file variables explain
taining a subprime loan in 2006. In this equa-     nearly half (48 percent) of the black-white
tion, d is the coefficient associated with this    gap in subprime lending, whereas credit fac-
variable; W is a vector of controls including      tors such as FICO scores, loan-to-value ratios,
percent black, percent Hispanic, and percent       and interest rate type account for only an addi-
Asian; l is a vector of coefficients associated    tional one-sixth (17 percent) of the observed
with these variables; and n is the error term.     gap. Although we recognize the limitations
Equation 3 simply substitutes the value of seg-    of ecological data at the tract- and MSA-
regation predicted from this first-stage equa-     levels, we believe our proxies for credit fac-
tion into Equation 1 to yield a second-stage       tors in the probit equation adequately reduce
equation that expresses foreclosures as a func-    potential bias in our estimates.
tion of the segregation instrument plus the            For each MSA, we average the group like-
variables in Z. b1 and b2 are then re-estimated    lihood of receiving a subprime loan in 2004 to
in the second-stage equation, along with e.        2006 by summing the predicted probability by
    To generate a more refined measure of          race and ethnicity across all loans and then
lending discrimination to use as our instru-       dividing by the total number of loans to
ment, we estimate black-white and Hispanic-        each borrower race/ethnic group (i.e., non-
white differentials in the likelihood of receiv-   Hispanic white, non-Hispanic black, and His-
ing a subprime loan after adjusting for bor-       panic). We then calculate the black-minus-
rower and neighborhood characteristics             white and Hispanic-minus-white differences
reported in the HMDA data. That is, using          in regression-adjusted predicted subprime
an extract of 5,360,007 HMDA loan-level            lending probabilities for each of the 100
records with non-missing data, we predict          MSAs. The black-white differential has
RACEDIFF for each MSA using a probit               a mean of 11.8 percent (sd 4.3 percent) and
model where the dependent variable is              ranges from 2.3 to 24.0 percent; the
a dichotomous indicator equal to one if the        Hispanic-white differential is also always pos-
loan is flagged as subprime in the data by         itive, with a mean of 8.1 percent (sd 3.8 per-
a non-missing interest rate greater than or        cent) and a range of 1.4 to 17.5 percent. We
644                                                         American Sociological Review 75(5)

merge these two differential variables to the          Likewise, the coefficient for Hispanic seg-
main data file.                                     regation is initially insignificant with a coeffi-
    We use the regression-adjusted black-           cient of .81 when estimated using OLS, but
white and Hispanic-white differentials in           using the instrumental variable estimator the
subprime lending by MSA to predict the              value rises to 1.12, which is nearly statistically
segregation instrument inserted into the sec-       significant ( p = .15 using a two-tailed test and
ond-stage equation. Table 5 reports OLS             p = .08 under a one-tailed test). A .10-point
and 2SLS estimates of the effect of black           increase in Hispanic dissimilarity is estimated
and Hispanic segregation on the rate of fore-       to result in an 11 percent increase under IV
closures for the top 100 MSAs, excluding            estimation, indicating that unexplained His-
Honolulu, Hawaii, as in our main analysis.          panic-white differences in subprime loan
The model includes the same covariates as           usage augment our understanding of the effect
in Table 1 except log of population, level          of Latino segregation on metropolitan-level
of unemployment, the Rio Grande border              foreclosures.7 Note that the OLS and IV mod-
dummy, and age of housing stock.6                   els yield similar coefficient estimates for the
    The estimated OLS coefficient for black         effect of economic trends, housing market
segregation (see the first column in Table 5)       conditions, land use regulation, region, and
is highly significant, and at 3.84 it is compara-   other controls. This suggests that segregation
ble to that in our initial model (see Table 2).     contributes to explaining variation in the fore-
This suggests that a .10-point rise in black        closure rate above and beyond the standard in-
segregation is associated with a 38 percent         dicators heretofore employed in analytic
increase in the foreclosure rate. By contrast,      models.
the instrumental variable estimate of the coef-
ficient is 4.64. This coefficient is estimated
quite precisely and attains significance at the
.001 level. Its higher point estimate implies
that a .10-point increase in black segregation      The analyses provide strong empirical sup-
is associated with a 46 percent increase in         port for the hypothesis that residential segre-
the foreclosure rate. While this effect is not      gation constitutes an important contributing
statistically different from the OLS effect         cause of the current foreclosure crisis, that
due to overlapping confidence intervals, its        segregation’s effect is independent of other
higher value offers more evidence that segre-       economic causes of the crisis, and that segre-
gation indeed has a causal effect on the MSA        gation’s explanatory power exceeds that of
foreclosure rate by producing racial differen-      other factors hitherto identified as key causes
tials in subprime lending.                          (e.g., overbuilding, excessive subprime lend-
    Test statistics for endogeneity indicate that   ing, housing price inflation, and lenders’
the racial differential instrument is indeed        failure to adequately evaluate borrowers’
exogenous, a conclusion corroborated by the         creditworthiness). Simply put, the greater
fact that it is uncorrelated with the residuals     the degree of Hispanic and especially black
of the reduced form model in Equation 3.            segregation a metropolitan area exhibits, the
The percent of the MSA population that is           higher the number and rate of foreclosures
black has no impact whatsoever on our segre-        it experiences. Neither the number nor the
gation estimates and a much smaller offsetting      rate of foreclosures is in any way related to
impact on the rate of foreclosures. This auxil-     expanded lending to minority home owners
iary finding underscores our hypothesis that        as a result of the Community Reinvestment
racial concentration in space, and not race         Act.
alone, is a significant structural cause of the        The confluence of low interest rates, un-
current foreclosure crisis.                         paralleled levels of equity extraction via
Rugh and Massey                                                                                        645

Table 5. Estimates of the Effect of Residential Segregation on Log of 2006 to 2008 Foreclosure
Rate via Black-White and Hispanic-White Adjusted Differentials in the Likelihood of
Obtaining a Subprime Loan in 2004 to 2006

                                               Black Segregation                Hispanic Segregation
                                             OLS                IV            OLS                IV
                                           B       SE       B        SE     B        SE      B         SE

Dissimilarity Index
  African Americans                     3.842** .630 4.638** .888
  Hispanics                                                                .811  .662 1.124         .788
Selected Control Variables
  Housing Starts Ratio                  2.581** .773 2.728** .714 1.874* .939 1.901*                .832
  Ratio of pre- to post-2000 HPI          .086** .022    .083** .020 .101** .032         .101** .029
  Wharton Land Use Index                  .221* .083     .229** .078 .182* .091          .178*      .078
  Subprime Loan Share, 2004 to 2006 2.5681 1.404 2.085             1.292 4.893* 1.887 4.965** 1.654
  MSA Credit Score Index                2.015* .007 2.015*          .006 2.0141 .008 2.0144* .007
  Change in Unemployment Rate             .239** .041    .247** .039 .203** .057         .199** .048
  Percent Black, 2000                 21.649* .685 21.826** .595 2.797           .907 2.881         .805
  Percent Hispanic, 2000                2.752    .538 2.635         .48621.315   .737 21.441*       .698
  Percent College Degree, 2006        21.9571 1.161 22.265* 1.147 2.475 1.700 2.445                1.517
     Midwest                              .386* .188     .297*      .152 .813** .213     .840** .199
     South                              2.007    .216 2.010         .189 .009    .288    .060       .280
     West                                 .501   .326    .525*      .289 .382    .395    .419       .366
  Constant                              1.396 6.839      .608      6.363 5.195 8.035 5.751         7.239
R2                                        .77            .76               .65           .65
F                                      27.4                              10.65
Wald x2                                               534.72                          257.27
[F, 2SLS 1st stage]                                   [38.89]                         [27.08]
Tests of Endogeneity (Null: Instrument is Exogenous)
  Robust x2 (p value)                                   1.18 (.27)                       .45 (.50)
  Robust F (p value)                                     .99 (.32)                       .35 (.56)
  Covariance (Instrument, eIV)                         2.00001                         2.0041

Note: N = 99. Ordinary Least Squares (OLS) and Indirect Least Squares or Instrumental Variables (IV)
estimates with robust standard errors. Additional covariates included in model but not shown here are
listed in Table 2 excluding log of 2008 population, 2006 unemployment rate, borders Rio Grande, and age
of housing stock. See text for detailed description of adjusted racial and ethnic differences in subprime
lending instrument. eIV is the residual error term value from the corresponding IV regression model.
  p \ .10; *p \ .05; **p \ .01 (two-tailed tests).

refinancing, and the bust of the housing bub-           causes of the crisis, as well as the geo-
ble may have combined with overbuilding                 graphic and social distribution of its costs,
and lax regulation to make the foreclosure cri-         on the basis of race. Segregation therefore
sis possible (Glaeser 2009; Khandani et al.             racialized and intensified the consequences
2009). However, we add a crucial addition               of the American housing bubble. Hispanic
to the understanding of the causes and                  and black home owners, not to mention
consequences of the foreclosure crisis by               entire Hispanic and black neighborhoods,
demonstrating the key role of residential seg-          bore the brunt of the foreclosure crisis.
regation in shaping how the crisis played out.          This outcome was not simply a result of
By concentrating foreclosures in metropolitan           neutral market forces but was structured on
areas with large racial differentials in sub-           the basis of race and ethnicity through the
prime lending, segregation structured the               social fact of residential segregation.
646                                                          American Sociological Review 75(5)

    Ultimately, the racialization of America’s      goods and services, generate wealth, and pro-
foreclosure crisis occurred because of a sys-       duce income, then it is incumbent upon gov-
tematic failure to enforce basic civil rights       ernment to ensure that all citizens have the
laws in the United States. Discriminatory sub-      right to compete freely in all markets (Massey
prime lending is simply the latest in a long        2005b). In a market society, lack of access to
line of illegal practices that have been foisted    markets translates directly into a lack of equal
on minorities in the United States (Satter          access to material well-being and ultimately
2009). It is all the more shocking because          into socioeconomic inequality (Massey 2007).
these practices were well-known and docu-              Unfortunately, to secure passage of the
mented long before the housing bubble burst         Civil Rights Acts of 1964, 1968, 1974, and
(e.g., Squires 2004; Stuart 2003; U.S. Depart-      1977 and to avoid a southern filibuster, most
ment of Housing and Urban Development               of the enforcement mechanisms included in
2000; Williams et al. 2005). In addition to         the original legislation were stripped away
tighter regulation of lending, rating, and secu-    and the federal government is largely pro-
ritization practices, greater civil rights          hibited from playing an active role in uncover-
enforcement has an important role to play in        ing discrimination or instigating actions to
cleaning up U.S. markets.                           sanction those who discriminate. The existing
    It is in the nation’s interest for federal      body of civil rights law must be updated to
authorities to take stronger and more energetic     establish within the U.S. Departments of Trea-
steps to rid U.S. real estate and lending mar-      sury, Labor, Commerce, and Housing and
kets of discrimination, not simply to promote       Urban Development permanent offices autho-
a more integrated and just society but to avoid     rized to conduct regular audits in markets for
future catastrophic financial losses. Racial dis-   jobs, goods, services, credit, and housing
crimination is easily detected through a meth-      based on representative samples of market
odology known as the audit study, in which          providers, both for purposes of enforcement
trained testers identifiable as black or white      and to measure progress in the elimination
are sent into markets to seek out proffered         of discrimination from U.S. markets.
goods and services. Black and white testers’
experiences over a number of trials are com-
piled and compared to discern systematic dif-
ferences in treatment (Fix and Struyk 1993;         The authors would like to thank the reviewers for their
Yinger 1986). Numerous audit studies docu-          helpful comments and criticisms that have helped to
                                                    improve the article.
ment the persistence of anti-black discrimina-
tion not only in markets for real estate (Yinger
1995; Zhao, Ondrich, and Yinger 2006) and           Notes
credit (Ross and Yinger 2002; Squires               1. The RealtyTrac database does not include tabulations
1994), but also in markets for jobs (Bertrand          for foreclosures in the Grand Rapids–Wyoming,
and Mullainathan 2004; Pager 2007; Turner,             Michigan MSA; it substitutes the Charleston–North
Fix, and Struyk 1991), goods (Ayres and Sie-           Charleston, South Carolina MSA.
gelman 1995), and services (Feagin and Sikes        2. Subprime loan pricing data were first made available
                                                       in the 2004 HMDA data. Comparing extremes in our
1994; Ridley, Bayton, and Outtz 1989). None-           data, about 40 percent of all 2004 to 2006 loans were
theless, the discrimination continues.                 subprime in the Detroit-Livonia-Dearborn, MI Metro
    An important goal in expanding civil rights        Division and the Miami–Miami Beach–Kendall, FL
enforcement should be the creation of federal          Metro Division, but less than 10 percent of loans
                                                       were subprime in the San Francisco-San Mateo-
programs to monitor levels of discrimination
                                                       Redwood City, CA Metro Division.
in key U.S. markets and to take remedial            3. Asian Americans are also clustered in MSAs with
action on a routine basis. If a society uses           either very high (e.g., coastal California, New
markets to allocate production, distribute             York, and Hawaii) or remarkably affordable (e.g.,
Rugh and Massey                                                                                                      647

     Texas) home prices, which somewhat forestalled the           Ayres, Ian and Peter Siegelman. 1995. ‘‘Race and Gen-
     rise of subprime lending.                                       der Discrimination in Bargaining for a New Car.’’
4.   The error term is specified to be robust to heterosce-          American Economic Review 85:304–321.
     dasticity using the ‘‘robust’’ option in Stata statistical   Been, Vicki, Ingrid Ellen, and Josiah Madar. 2009. ‘‘The
     software version 10.                                            High Cost of Segregation: Exploring Racial Dispar-
5.   The share of lending made by CRA-covered banks is               ities in High-Cost Lending.’’ Fordham Urban Law
     negative, as predicted, and is insignificant only in the        Journal 36:361–93.
     presence of the subprime lending share (the two var-         Bertrand, Marriane and Sendhil Mullainathan. 2004.
     iables are significantly negatively correlated, r =             ‘‘Are Emily and Greg More Employable than Laki-
     2.31, p \ .01).                                                 sha and Jamal? A Field Experiment on Labor Market
6.   Additionally, in the Hispanic segregation models,               Discrimination.’’ American Economic Review
     black segregation is omitted.                                   94:991–1013.
7.   To estimate the potential effects of Hispanic segrega-       Bhardwaj, Geetesh and Rajdeep Sengupta. 2009.
     tion, we undertook a separate analysis of the nation’s          ‘‘Where’s the Smoking Gun? A Study of Underwrit-
     largest state, California, where Hispanics are numer-           ing Standards for US Subprime Mortgages.’’ Federal
     ous and there are far fewer blacks. In the analysis of          Reserve Bank of St. Louis Working Paper 2008-
     California foreclosures at the city- and county-levels          036C. Retrieved June 14, 2010 (http://research
     that control for a much more extensive array of loan  
     underwriting factors, such as weighted loan-to-value         Bhutta, Neil. 2008. ‘‘Giving Credit Where Credit Is
     ratios, average credit scores, and interest rates and           Due? The Community Reinvestment Act and Mort-
     matched city-level home price trends, we estimated              gage Lending in Lower-Income Neighborhoods.’’
     a significant, robust effect of Hispanic segregation.           Finance and Economics Discussion Series Divisions
     Notwithstanding the incredible boom and bust in pla-            of Research & Statistics and Monetary Affairs Fed-
     ces like the Central Valley and Inland Empire, the              eral Reserve Board, Washington, DC. Working
     residential segregation of Latinos matters a great              Paper 2008-61.
     deal to local differences in foreclosure trends. These       Bhutta, Neil and Glenn B. Canner. 2009. ‘‘Did the CRA
     results support our proposition about the primacy of            Cause the Mortgage Market Meltdown?’’
     segregation in structuring the foreclosure crisis and           Community Dividend (Federal Reserve Bank of
     do not bode well for the housing market fortunes of             Minneapolis). Retrieved June 14, 2010 (http://
     Hispanics, who became the largest minority group      
     during the housing boom.                                        display.cfm?id=4136).
                                                                  Bocian, Debbie, Keith Ernst, and Wei Li. 2006. ‘‘Unfair
                                                                     Lending: The Effect of Race and Ethnicity on the
References                                                           Price of Subprime Mortgages.’’ Center for Responsi-
                                                                     ble Lending. Retrieved June 14, 2010 (http://
Andrews, Edmund L. 2009. Busted: Life Inside the           
   Great Mortgage Meltdown. New York: W.W.                           research-analysis/rr011-Unfair_Lending-0506.pdf).
   Norton.                                                        Bond, Carolyn and Richard Williams. 2007. ‘‘Residen-
Angrist, Joshua D. and Alan B. Krueger. 2001. ‘‘Instru-              tial Segregation and the Transformation of Home
   mental Variables and the Search for Identification:               Mortgage Lending.’’ Social Forces 86:671–98.
   From Supply and Demand to Natural Experiments.’’               Bromley, Charles, Jim Campen, Saara Nafici, Adam
   Journal of Economic Perspectives 15:69–85.                        Rust, Geoff Smith, Kevin Stein, and Barbara van
Apgar, William and Allegra Calder. 2005. ‘‘The Dual                  Kerkhove. 2008. ‘‘Paying More for the American
   Mortgage Market: The Persistence of Discrimination                Dream: The Subprime Shakeout and Its Impact on
   in Mortgage Lending.’’ Pp. 101–123 in Geography                   Lower-Income and Minority Communities.’’ A joint
   of Opportunity: Race and Housing Choice in Metro-                 report by the California Reinvestment Coalition,
   politan America, edited by X. De Souza Briggs.                    Community Reinvestment Association of North Car-
   Washington, DC: Brookings Institution Press.                      olina, Empire Justice Center, Massachusetts Afford-
Avery, Robert B., Kenneth P. Brevoort, and Glenn B.                  able Housing Alliance, Neighborhood Economic
   Canner. 2007. ‘‘The 2006 HMDA Data.’’ Federal                     Development Advocacy Project, Ohio Fair Lending
   Reserve Bulletin 93:73–109.                                       Coalition, and Woodstock Institute.
———. 2008. ‘‘The 2007 HMDA Data.’’ Federal                        Calem, Paul S., Jonathan E. Hershaff, and Susan M.
   Reserve Bulletin 94:107–146.                                      Wachter. 2004. ‘‘Neighborhood Patterns of Sub-
Avery, Robert B., Glenn B. Canner, and Robert E. Cook.               prime Lending: Evidence from Disparate Cities.’’
   2005. ‘‘New Information Reported Under HMDA                       Housing Policy Debate 15:603–622.
   and Its Implication in Fair Lending Enforcement.’’             California Reinvestment Committee. 2001. Stolen
   Federal Reserve Bulletin 91:344–94.                               Wealth: Disparities in California’s Subprime
648                                                                 American Sociological Review 75(5)

    Lending Market. San Francisco, CA: California              Penalty Patterns.’’ Housing Policy Debate 15:
    Reinvestment Committee. Retrieved June 14, 2010            687–714.
    (     Feagin, Joe R. and Melvin P. Sikes. 1994. Living with
Charles, Camille Zubrinsky. 2003. ‘‘The Dynamics of            Racism: The Black Middle-Class Experience. Bos-
    Racial Residential Segregation.’’ Annual Review of         ton, MA: Beacon Press.
    Sociology 29:167–207.                                  Fischel, William. 1990. ‘‘Do Growth Controls Matter? A
DeNavas-Walt, Carmen, Bernadette D. Proctor, and Jes-          Review of Empirical Evidence on the Effectiveness
    sica C. Smith. 2009. ‘‘Income, Poverty, and Health         and Efficiency of Local Government Land Use Reg-
    Insurance Coverage in the United States: 2008.’’           ulation.’’ Lincoln Institute of Land Policy Working
    U.S. Census Bureau, Current Population Reports             Paper Series 87:1–66.
    P60-236(RV). Washington, DC: U.S. Government           Fix, Michael and Raymond J. Struyk. 1993. Clear and
    Printing Office. Retrieved June 14, 2010 (http://          Convincing Evidence: Measurement of Discrimination                 in America. Washington, DC: Urban Institute Press.
Denton, Nancy A. and Douglas S. Massey. 1989.              Friedman, Samantha and Gregory D. Squires. 2005.
    ‘‘Racial Identity among Caribbean Hispanics: The           ‘‘Does the Community Reinvestment Act Help
    Effect of Double Minority Status on Residential Seg-       Minorities Access Traditionally Inaccessible Neigh-
    regation.’’ American Sociological Review 54:790–           borhoods?’’ Social Problems 52:209–231.
    808.                                                   Galster, George C. 1986. ‘‘More than Skin Deep: The
Doms, Mark, Fred Furlong, and John Krainer. ‘‘Sub-             Effects of Housing Discrimination on the Extent
    prime Mortgage Delinquency Rates.’’ Federal                and Pattern of Racial Residential Segregation in
    Reserve Bank of San Francisco, Working Paper               the United States.’’ Pp. 119–38 in Housing Discrim-
    2007-33.                                                   ination and Federal Policy, edited by J. M. Goering.
Dymski, Gary A. 2002. ‘‘U.S. Housing as Capital Accu-          Chapel Hill, NC: University of North Carolina.
    mulation: The Transformation of American Housing       Galster, George C. and W. Mark Keeney. 1988. ‘‘Race,
    Finance, Households, and Communities.’’ Pp. 63–96          Residence, Discrimination, and Economic Opportu-
    in Seeking Shelter on the Pacific Rim: Financial           nity: Modeling the Nexus of Urban Racial Phenom-
    Globalization, Social Change, and the Housing Mar-         ena.’’ Urban Affairs Quarterly 24:87–117.
    ket, edited by G. A. Dymski and D. Isenberg. Ar-       Gerardi, Kristopher S., Adam Hale Shapiro, and Paul S.
    monk, NY: M.E. Sharpe.                                     Willen. 2009. ‘‘Decomposing the Foreclosure Crisis:
Dymski, Gary A. and John M. Veitch. 1992. ‘‘Race and           House Price Depreciation versus Bad Underwrit-
    the Financial Dynamics of Urban Growth: L.A. as            ing.’’ Federal Reserve Bank of Atlanta, Working
    Fay Wray.’’ Pp. 131–57 in City of Angels, edited           Paper 2009-25. Retrieved June 14, 2010 (http://
    by G. Riposa and C. G. Dersch. Dubuque, IA: Ken- 
    dall/Hunt.                                             Gerardi, Kristopher S. and Paul S. Willen. 2008. ‘‘Sub-
Edmiston, Kelly D. 2009. ‘‘Characteristics of High-            prime Mortgages, Foreclosures, and Urban Neigh-
    Foreclosure Neighborhoods in the Tenth District.’’         borhoods.’’ Federal Reserve Bank of Boston,
    Federal Reserve Bank of Kansas City Economic               Public Policy Discussion Paper 2008-06. Retrieved
    Review (2nd Quarter):51–75.                                June 14, 2010 (
Elliehausen, Gregory, Michael E. Staten, and Jevgenijs         ppdp/2008/ppdp0806.pdf).
    Steinbuks. 2008. ‘‘The Effect of Prepayment Penal-     Glaeser, Edward L. 2009. ‘‘Foreclosing the Crisis.’’ The
    ties on the Pricing of Subprime Mortgages.’’ Journal       New Republic, February 4. Retrieved June 14, 2010
    of Business and Economics 60:33–46.                        (
Engel, Kathleen C. and Patricia A. McCoy. 2007.                crisis).
    ‘‘Turning a Blind Eye: Wall Street Finance of Pred-    Glaeser, Edward L. and Joseph Gyourko. 2003. ‘‘The
    atory Lending.’’ Fordham Law Review 73:2039–               Impact of Building Restrictions on Housing Afford-
    2103.                                                      ability.’’ Federal Reserve Bank of New York Eco-
———. 2008. ‘‘From Credit Denial to Predatory Lend-             nomic Policy Review 9:21–39.
    ing: The Challenge of Sustaining Minority Home-        ———. 2008. Rethinking Federal Housing Policy.
    ownership.’’ Pp. 81–123 in Segregation: The                Washington, DC: American Enterprise Institute
    Rising Costs for America, edited by J. H. Carr and         Press.
    N. K. Kutty. New York and London: Routledge.           Glaeser, Edward L., Joseph Gyourko, and Albert Saiz.
Farley, Reynolds and William H. Frey. 1994. ‘‘Changes          2008. ‘‘Housing Supply and Housing Bubbles.’’
    in the Segregation of Whites from Blacks during the        Journal of Urban Economics 64:198–217.
    1980s: Small Steps toward a More Integrated Soci-      Gyourko, Joseph, Anita Summers, and Albert Saiz.
    ety.’’ American Sociological Review 59:23–45.              2008. ‘‘A New Measure of the Local Regulatory
Farris, John and Christopher A. Richardson. 2004. ‘‘The        Environment for Housing Markets.’’ Urban Studies
    Geography of Subprime Mortgage Prepayment                  45:693–729.
Rugh and Massey                                                                                                649

Hackworth, Jason R. 2007. The Neoliberal City. Ithaca,        14, 2010 (
    NY: Cornell University Press.                             .pdf).
Hanlon, Bernadette. 2010. Once the American Dream:        LaCour-Little, Michael and Cynthia Holmes. 2008.
    Inner-Ring Suburbs of the Metropolitan United             ‘‘Prepayment Penalties in Residential Mortgage
    States. Philadelphia, PA: Temple University Press.        Contracts: A Cost-Benefit Analysis.’’ Housing Pol-
Haughwout, Andrew, Richard Peach, and Joseph Tracy.           icy Debate 19:631–73.
    2008. ‘‘Juvenile Delinquent Mortgages: Bad Credit     Langley, Paul. 2008. ‘‘Sub-prime Mortgage Lending: A
    or Bad Economy?’’ Journal of Urban Economics              Cultural Economy.’’ Economy and Society 37:469–94.
    64:246–57.                                            ———. 2009. ‘‘Debt, Discipline, and Government:
Hernandez, Jesus. 2009. ‘‘Redlining Revisited: Mort-          Foreclosure and Forbearance in the Subprime
    gage Lending Patterns in Sacramento 1930–2004.’’          Mortgage Crisis.’’ Environment and Planning A 41:
    International Journal of Urban and Regional               1404–19.
    Research 33:291–313.                                  Lee, Matthew. 1999. ‘‘Strategies to Combat Predatory
Hirsch, Barry T. and David A. MacPherson. 2007.               Subprime Lending, Particularly When Targeted at
    ‘‘Metropolitan Area: Union Membership, Coverage,          Communities of Color.’’ National Fair Housing
    Density, and Employment by Metropolitan Area and          Advocate Online, May. Retrieved June 14, 2010
    Sector, 1986–2006,’’ Union Mem-           (
    bership and Coverage Database from the CPS [elec-         .display&pagename=guest_room_lee).
    tronic file].                                         Lewis, Michael. 2010. The Big Short: Inside the Dooms-
Holloway, Steven R. 1998. ‘‘Exploring the Neighbor-           day Machine. New York: W.W. Norton.
    hood Contingency of Race Discrimination in Mort-      Mallach, Alan. 2009. ‘‘Addressing Ohio’s Foreclosure
    gage Lending in Columbus, Ohio.’’ Annals of the           Crisis: Taking the Next Steps.’’ Brookings Institution
    Association of American Geographers 88:252–76.            Metropolitan Policy Program Papers Series, no. 6.
Iceland, John, Daniel H. Weinberg, and Erika Steinmetz.   Manti, Michael, Paul Raca, and Peter Zorn. 2004. ‘‘Sub-
    2002. Racial and Ethnic Residential Segregation in        prime Lending: An Investigation of Economic Effi-
    the United States, 1980–2000. U.S. Census Bureau,         ciency.’’ Housing Policy Debate 15:533–71.
    Special Report Series, CENSR-3. Washington, DC:         ´
                                                          Marquez, Mercedes. 2008. ‘‘Models of Vacant Property
    U.S. Government Printing Office. Retrieved June           Disposition: Los Angeles Housing Department.’’
    14, 2010 (         Presented at the Federal Reserve Board Symposium,
    housing_patterns/pdf/censr-3.pdf).                        ‘‘Confronting the Neighborhood Impacts of Foreclo-
Immergluck, Dan. 2008. ‘‘From the Subprime to the             sure,’’ October 20, Washington, DC.
    Exotic: Excessive Mortgage Market Risk and Impli-     Massey, Douglas S. 2004. ‘‘Segregation and Stratifica-
    cations for Metropolitan Communities and Neigh-           tion: A Biosocial Perspective.’’ The DuBois Review:
    borhoods.’’ Journal of the American Planning              Social Science Research on Race 1:1–19.
    Association 74:59–76.                                 ———. 2005a. ‘‘Racial Discrimination in Housing: A
———. 2009. Foreclosed: High-Risk Lending, Deregu-             Moving Target.’’ Social Problems 52:148–51.
    lation, and the Undermining of America’s Mortgage     ———. 2005b. Return of the L-Word: A Liberal Vision
    Market. Ithaca, NY: Cornell University Press.             for the New Century. Princeton, NJ: Princeton Uni-
Immergluck, Dan and Marti Wiles. 1999. Two Steps              versity Press.
    Back: The Dual Mortgage Market, Predatory Lend-       ———. 2007. Categorically Unequal: The American
    ing, and the Undoing of Community Development.            Stratification System. New York: Russell Sage
    Chicago, IL: Woodstock Institute.                         Foundation.
Institute on Race and Poverty. 2009. Communities in       ———. 2009. ‘‘Racial Formation in Theory and Prac-
    Crisis: Race and Mortgage Lending in the Twin Cit-        tice: The Case of Mexicans in the United States.’’
    ies. University of Minnesota Law School. Retrieved        Race and Social Problems 1:12–26.
    June 14, 2010 (   Massey, Douglas S. and Brooks Bitterman. 1985. ‘‘Ex-
    projects/IRP_mortgage_study_Feb._11th.pdf).               plaining the Paradox of Puerto Rican Segregation.’’
Khandani, Amir E., Andrew W. Lo, and Robert C. Mer-           Social Forces 64:307–31.
    ton. 2009. ‘‘Systemic Risk and the Refinancing        Massey, Douglas S. and Nancy A. Denton. 1988. ‘‘The
    Ratchet Effect.’’ MIT Sloan Research Paper No.            Dimensions of Residential Segregation.’’ Social
    4750-09; Harvard Business School Finance Working          Forces 67:281–315.
    Paper No. 1472892. Retrieved June 14, 2010 (http://   ———. 1992. ‘‘Racial Identity and the Segregation of                    Mexicans in the United States.’’ Social Science
Kochhar, Rakesh, Ana Gonzalez-Barrera, and Daniel             Research 21:235–60.
    Dockterman. 2009. ‘‘Through Boom and Bust:            ———. 1993. American Apartheid: Segregation and the
    Minorities, Immigrants and Homeownership.’’ Pew           Making of the Underclass. Cambridge, MA: Harvard
    Hispanic Center, Washington, DC. Retrieved June           University Press.
650                                                                 American Sociological Review 75(5)

Massey, Douglas S. and Mary J. Fischer. 1999. ‘‘Does       Quercia, Roberto G., Michael A. Stegman, and Walter
   Rising Income Bring Integration? New Results for            R. Davis. 2007. ‘‘The Impact of Predatory Loan
   Blacks, Hispanics, and Asians in 1990.’’ Social Sci-        Terms on Subprime Foreclosures: The Special
   ence Research 28:316–26.                                    Case of Prepayment Penalties and Balloon Pay-
———. 2000. ‘‘How Segregation Concentrates Pov-                 ments.’’ Housing Policy Debate 18:311–46.
   erty.’’ Ethnic and Racial Studies 23:670–91.            Raynes, Sylvain and Ann Rutledge. 2003. The Analysis
Massey, Douglas S., Jonathan Rothwell, and Thurston            of Structured Securities: Precise Risk Measurement
   Domina. 2009. ‘‘Changing Bases of Segregation in            and Capital Allocation. Oxford and New York:
   the United States.’’ Annals of the American Academy         Oxford University Press.
   of Political and Social Science 626:74–90.              Renuart, Elizabeth. 2004. ‘‘An Overview of the Preda-
Massey, Douglas S., Michael J. White, and Voon-Chin            tory Lending Mortgage Lending Process.’’ Housing
   Phua. 1996. ‘‘The Dimensions of Segregation Revis-          Policy Debate 15:467–502.
   ited.’’ Sociological Methods and Research 25:172–206.   Ridley, Stanley E., James A. Bayton, and Janice H.
Mayer, Christopher, Karen Pence, and Shane M. Sher-            Outtz. 1989. ‘‘Taxi Service in the District of Colum-
   lund. 2009. ‘‘The Rise in Mortgage Defaults.’’ Jour-        bia: Is it Influenced by Patron’s Race and Destina-
   nal of Economic Perspectives 23:27–50.                      tion?’’ Paper prepared for the Washington, DC
Mortgage Bankers Association. 2006. ‘‘Residential              Lawyer’s Committee for Civil Rights Under the Law.
   Mortgage Delinquencies Increase, According to           Ross, Stephen L. and Margery A. Turner. 2005. ‘‘Hous-
   MBA National Delinquency Survey [NDS].’’ Press              ing Discrimination in Metropolitan America: Ex-
   Release, NDS, March 16. Retrieved June 14, 2010             plaining Changes between 1989 and 2000.’’ Social
   (          Problems 52:152–80.
   Center/40245.htm).                                      Ross, Stephen L. and John Yinger. 2002. The Color of
———. 2010. ‘‘Delinquencies, Foreclosure Starts Fall            Credit: Mortgage Discrimination, Research Method-
   in Latest MBA National Delinquency Survey                   ology, and Fair-Lending Enforcement. Cambridge,
   [NDS].’’ Press Release, NDS, February 19.                   MA and London: MIT Press.
   Retrieved June 14, 2010 (http://www.mortgageban         Rothwell, Jonathan and Douglas S. Massey. 2009. ‘‘The               Effect of Density Zoning on Racial Segregation in
National Association of Realtors, Research Division.           U.S. Urban Areas.’’ Urban Affairs Review 44:779–
   2004. ‘‘Rising Foreclosure Rates in Indiana: An             806.
   Explanatory Analysis of Contributing Factors.’’         Satter, Beryl. 2009. Race, Real Estate and the Exploita-
   Retrieved June 14, 2010 (              tion of Black Urban America. New York: Metropol-
   _pdfs/ForeclosureStudy2004.pdf).                            itan Books.
Oliver, Melvin L. and Thomas M. Shapiro. 2006. Black       Schuetz, Jenny, Vicki Been, and Ingrid Gould Ellen.
   Wealth/White Wealth: A New Perspective on Racial            2008. ‘‘Neighborhood Effects of Concentrated Mort-
   Inequality, 2nd ed. New York and London:                    gage Foreclosures.’’ Journal of Housing Economics
   Routledge.                                                  17:306–319.
Pager, Devah. 2007. Marked: Race, Crime, and Finding       Shapiro, Thomas M., Tatjana Meschede, and Laura Sul-
   Work in an Era of Mass Incarceration. Chicago, IL:          livan. 2010. ‘‘The Racial Wealth Gap Increases
   University of Chicago Press.                                Fourfold.’’ Institute on Assets and Social Policy
Park, Kevin. 2008. ‘‘Subprime Lending and the Commu-           Research and Policy Brief. Retrieved June 14,
   nity Reinvestment Act.’’ Joint Center for Housing           2010 (
   Studies at Harvard University, Cambridge, MA.               Gap-Brief.pdf).
Pendley, M. Diane, Glenn Costello, and Mary Kelsch.        Shiller, Robert J. 2008. The Subprime Solution: How To-
   2007. ‘‘The Impact of Poor Underwriting Practices           day’s Global Financial Crisis Happened, and What
   and Fraud in Subprime RMBS Performance.’’ Fitch             to Do about It. Princeton, NJ: Princeton University
   Ratings Structured Finance U.S. Mortgage Special            Press.
   Report. Retrieved June 14, 2010 (http://www.fitch       Sowell, Thomas. 2009. The Housing Boom and Bust.        New York: Basic Books.
   id=356624).                                             Squires, Gregory D. 1994. Capital and Communities in
Pennington-Cross, Anthony, Anthony Yezer, and Joseph           Black and White: The Intersections of Race, Class,
   Nichols. 2000. ‘‘Credit Risk and Mortgage Lending:          and Uneven Development. Albany, NY: State Uni-
   Who Uses Subprime and Why?’’ Research Institute             versity of New York Press.
   for Housing America Working Paper 00–03.                ———, ed. 2004. Why the Poor Pay More. Westport,
Peterson, Christopher L. 2007. ‘‘Predatory Structured          CT: Praeger.
   Finance.’’ Cardozo Law Review 28:2185–282.              Squires, Gregory D., Derek S. Hyra, and Robert N. Ren-
Powell, Michael. 2010. ‘‘Blacks in Memphis Lose Decades        ner. 2009. ‘‘Segregation and the Subprime Lending
   of Economic Gains.’’ New York Times, May 30, p. A1.         Crisis.’’ Presented at the 2009 Federal Reserve
Rugh and Massey                                                                                               651

    System Community Affairs Research Conference.        Wyly, Elvin, Markus Moos, Daniel Hammel, and Ema-
    Retrieved June 14, 2010 (http://www.kansascityf         nuel Kabahizi. 2009. ‘‘Cartographies of Race and             Class: Mapping the Class-Monopoly Rents of Amer-
    .pdf).                                                  ican Subprime Mortgage Capital.’’ International
Stuart, Guy. 2003. Discriminating Risk: The U.S. Mort-      Journal of Urban and Regional Research 33:332–54.
    gage Lending Industry in the Twentieth Century.      Yinger, John. 1986. ‘‘Measuring Racial Discrimination
    Ithaca, NY: Cornell University Press.                   with Fair Housing Audits: Caught in the Act.’’ Amer-
Sullivan, Teresa A., Elizabeth Warren, and Jay Law-         ican Economic Review 76:991–93.
    rence Westbrook. 2000. The Fragile Middle Class:     ———. 1995. Closed Doors, Opportunities Lost: The
    Americans in Debt. New Haven, CT: Yale Univer-          Continuing Costs of Housing Discrimination. New
    sity Press.                                             York: Russell Sage.
Turner, Margery A., Michael Fix, and Raymond J.          Zhao, Bo, Jan Ondrich, and John Yinger. 2006. ‘‘Why
    Struyk. 1991. Opportunities Denied, Opportunities       Do Real Estate Brokers Continue to Discriminate?
    Diminished: Racial Discrimination in Hiring. Wash-      Evidence from the 2000 Housing Discrimination
    ington, DC: Urban Institute Press.                      Study.’’ Journal of Urban Economics 59:394–419.
Turner, Tracy M. and Marc T. Smith. 2009. ‘‘Exits from
    Homeownership: The Effects of Race, Ethnicity, and
    Income.’’ Journal of Regional Science 49:1–32.
U.S. Department of Housing and Urban Development         Jacob S. Rugh is a PhD candidate in Public Affairs at
    (HUD). 2000. Unequal Burden: Income and Racial       the Woodrow Wilson School of Public and International
    Disparities in Subprime Lending in America.          Affairs at Princeton University. His research focuses on
    Retrieved June 14, 2010 (     urban policy and the intersection of housing markets,
    publications/fairhsg/unequal.html).                  land use regulation, and local politics. His forthcoming
Williams, Richard, Reynold Nesiba, and Eileen Diaz       dissertation will focus on the social, economic, and local
    McConnell. 2005. ‘‘The Changing Face of Inequality   regulatory roots of the recent U.S. housing crisis and
    in Home Mortgage Lending.’’ Social Problems          their implications for public policy.
Wilkes, Rima and John Iceland. 2004. ‘‘Hypersegrega-     Douglas S. Massey is the Henry G. Bryant Professor of
    tion in the Twenty-First Century: An Update and      Sociology and Public Affairs at Princeton University. He
    Analysis.’’ Demography 41:23–36.                     currently serves as President of the American Academy
Wyly, Elvin K., Mona Atia, Holly Foxcroft, Daniel J.     of Political and Social Science and is past-President of
    Hamme, and Kelly Phillips-Watts. 2006. ‘‘Ameri-      the American Sociological Association and the Popula-
    can Home: Predatory Mortgage Capital and Neigh-      tion Association of America. His latest book is Brokered
    bourhood Spaces of Race and Class Exploitation in    Boundaries: Creating Immigrant Identity in Anti-Immi-
    the United States.’’ Geografiska Annaler 88:         grant Times, coauthored with Magaly Sanchez and pub-
    105–132.                                             lished by the Russell Sage Foundation.

Shared By: