the potential and limitations of mortgage innovation in ...ewyly/research/listokin(2002).pdf ·...
TRANSCRIPT
The Potential and Limitations of Mortgage Innovation in Fostering Homeownership in the United States
David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
ABSTRACT
Homeownership is widely regarded as the foundation of neighborhood stability and long-termwealth accumulation for American families. In turn, home-purchase mortgage lending hasbecome a central policy instrument in efforts to broaden access to homeownership amongunderserved populations.
This study presents a nationwide empirical analysis of the potential and limitations of mortgageinnovation to increase homeownership among underserved populations. We examine the finan-cial and underwriting criteria of a typology of mortgage products, and we develop syntheticunderwriting models calibrated with 1993–95 Survey of Income and Program Participation datato account for all direct purchase costs (including itemized components of closing costs anddown payments). Our analysis provides comprehensive estimates of a large, untapped market,but even the most aggressive mortgage market innovations can play only a limited role in effortsto deliver the material benefits of homeownership to underserved populations.
INTRODUCTION
Homeownership is widely regarded as the foundation of neighborhood stability andlong-term wealth accumulation for American families. Although two-thirds of thehouseholds in the United States have achieved homeownership, the rate is much lowerfor racial and ethnic minority and low- to moderate-income (LMI)1 populations, recentimmigrants, and others referred to as traditionally underserved populations. Recentscholarly research, community activism, and regulatory intervention have focused onways of expanding homeownership opportunities, especially for traditionally under-served populations. However, the limitations of available data and methods have, un-til recently, made it difficult to estimate how many renters could become homeownersif mortgage underwriting requirements were changed. The purpose of this study,conducted by the Center for Urban Policy Research (CUPR), Rutgers University, is tomeasure the effects of more affordable and flexible home financing (mortgage inno-
1 LMI is often defined in terms of a percentage of median income. For example, as used here “low income” refersto a household of four earning less than 50 percent of the median income, and “moderate income” refers to ahousehold of four earning between 50 percent and 80 percent of the median income.
© Fannie Mae Foundation 2002. All Rights Reserved. 1
The Potential and Limitations of Mortgage Innovation
vation) on access to homeownership by renters, especially members of traditionallyunderserved populations.2
The study is organized into the following sections:
1. Background2. Literature Review3. Simulation Framework and Models4. Model Calibration and Analysis5. Results: Homeownership Affordability6. Explanation of the Findings7. Policy Considerations8. Summary: Achievements and Limitations of Mortgage Innovation9. Evaluation of Simulation Results
10. Technical Appendices
2 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
2 Researchers have considered various aspects of the underserved market. Carr et al. (1994a), for example, ana-lyzed the spatial distribution of loans acquired by Fannie Mae and Freddie Mac in order to define underservedand served mortgage markets. (See also Can and Megbolugbe 1995; Carr et al. 1994b.)
BACKGROUND: HOMEOWNERSHIP IN THE UNITED STATES AND
STRATEGIES TO ASSIST THE TRADITIONALLY UNDERSERVED
Homeownership in the United States has long been touted as a preferred form ofhousing tenure, one that offers numerous advantages over renting a home. Allegedbenefits of owning a home include equity accumulation, greater involvement in neigh-borhood activities, greater sense of civic responsibility, improved educational andother achievements by the owner’s children, and other socioeconomic gains (Bunce etal. 1996; Dreier 1996; Green and White 1994; Rohe and Stegman 1994a, 1994b; Roheand Stewart 1995; Rossi and Weber 1996; Scanlon 1996; Shear, Wachter, and Weicher1988). Given this array of supposed benefits, it is not surprising that homeownershipis associated with realizing the “American Dream” (U.S. Department of Housing andUrban Development [HUD] 1995a).
Some researchers have criticized a historical bias by policy makers that has favoredproperty ownership (Krueckeberg 1999). Although the empirical evidence that home-ownership benefits both owners and society is supported, it is far from unequivocal(Rohe and Stegman 1994a, 1994b; Rossi and Weber 1996). Further, some studies in-dicate that many households rent their homes by choice. Varady and Lipman (1994)and Goodman (1999) have identified lifestyle renters (e.g., many elderly people andhouseholds with adult interests and schedules), who rent their homes because of theflexibility, mobility, reduced portfolio risk, locational benefits (e.g., apartments in adult-oriented downtown areas), and other advantages renting offers. There are also transi-tory renters, for whom renting for a selected period of time is preferable to owning ahome (e.g., households temporarily relocated because of a job or for a personal rea-son). Thus, the benefits of homeownership may be overstated, or the benefits maynot always be verifiable. Indeed, homeownership may not be for everyone; further-more, an individual’s preference for homeownership or renting may depend on severalfactors, including stage of life.
While homeownership may not be a universal good, it nonetheless is deeply ingrainedas the American Dream. To realize that dream, billions of dollars in federal tax incen-tives have been extended to homeowners. Tax and other societal support for home-ownership has prompted a dramatic shift to this form of tenure. In 1980, the home-ownership rate was almost 66 percent. It then declined to 64 percent in 1990, causingmuch consternation. Yet homeownership has since rallied to reach a recent historichigh; as of the first quarter of 2000, 67.1 percent of all American households werehomeowners. The most current (2000) State of the Nation’s Housing reports that inthe last five years alone, the number of owner households grew by 6.9 million (JointCenter for Housing Studies 2000, 17).
Realizing the goal of homeownership is harder for some than for others (Gyourko,Linneman, and Wachter 1996). Since housing is such an expensive consumption item,its purchase is heavily influenced by one’s financial position (Gyourko and Linneman1993). Thus, homeownership understandably correlates with income. As of the firstquarter of 2000, 82 percent of all U.S. households with a family income higher than
Background: Homeownership in the United States 3
The Potential and Limitations of Mortgage Innovation
the median were homeowners, compared with only 52 percent of households with afamily income lower than the median (U.S. Bureau of the Census 2000).
Homeownership rates also vary by race and ethnicity, with racial and ethnic minoritiesmuch less likely than the rest of the population to be homeowners (Long and Caudill1992; Wachter and Megbolugbe 1992). As of the first quarter of 2000, the white (non-Hispanic) homeownership rate in the United States was 73.4 percent—much higherthan the rate of 47.4 percent for African-American households and 45.7 percent forHispanic households (U.S. Bureau of the Census 2000). According to the U.S. Bureauof the Census, since 1985, the white (non-Hispanic) homeownership rate has beenbetween 25 and 30 percentage points higher than the African-American or Hispanichomeownership rate (U.S. Bureau of the Census 2000).
These racial and ethnic disparities are the result of a complex set of economic, histor-ical, institutional, and other (e.g., demographic) factors (Gyourko and Linneman 1997).Historically, the nation’s housing finance system was designed to predominantly servethe needs of the housing construction industry and white middle-class nuclear familiesseeking to escape the crowded cities of the interwar period for new opportunities inthe suburbs (Hayden 1984; Jackson 1985; Wright 1981). The housing finance systemhas changed dramatically in recent years, but economic barriers—minorities’ reducedincome and wealth, and lower levels of intergenerational wealth transfers and upwardclass mobility—continue to suppress the homeownership rate of African Americansand Hispanics (Gyourko, Linneman, and Wachter 1997). Moreover, there is wide-spread evidence that racial discrimination persists in various forms in the housingand mortgage markets (Yinger 1995, 1998). The Boston Federal Reserve study pro-vided the most conservative and rigorous measurement of discrimination because itinvolved the analysis of household financial characteristics actually considered byunderwriters. Even after controlling for income, wealth, credit history, and all otherfactors considered in the loan-underwriting decision, African Americans were approx-imately 60 percent more likely to be denied mortgage credit than were identicallyqualified non-Hispanic whites (Munnell et al. 1996). The Boston Federal Reservestudy generated a torrent of criticism that focused on issues of omitted-variable bias,data-coding errors, econometric specification, and model fit (Horne 1993; Liebowitzand Day 1992; Schill and Wachter 1993; Zandi 1993). Nevertheless, several reanalysesof the same data yielded similar results (Carr and Megbolugbe 1993), and the BostonFederal Reserve researchers convincingly defended their conclusions in a point-by-point response to their critics (Browne and Tootell 1995). In sum, debate persists onthe question of discrimination, and there is little prospect for widespread consensus.While there is universal agreement on disparities in mortgage market outcomes, thequestion of discrimination remains.
Our study does not address mortgage market discrimination. We start with the ob-served differences in homeownership attainment and, holding aside its causality, lookto ways, through mortgage innovation, to increase the homeownership rate, especiallyamong minority, LMI, immigrant, and similar households. The U.S. National Home-ownership Strategy, while attempting to help all Americans, underscores a “specialresponsibility and an important opportunity to target underserved populations and
4 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
communities,” including minorities (HUD 1996, 1). Because financing is key to realiz-ing homeownership, there are growing efforts to improve traditionally underservedpopulations’ access to credit. Given the multiple barriers faced by minorities and LMIpopulations in securing credit, the response by the public and private sectors has beenmultifaceted. Approaches have included attempts to address differential treatmentdiscrimination and differential impact discrimination as well as efforts to relax mort-gage borrowing constraints.
Attempts to eliminate discrimination involve strengthened enforcement of existinglaws (Federal Reserve Bank of Boston 1993; Federal Reserve System 1996). The FairHousing Act makes it unlawful for any person who engages in the business of mak-ing or purchasing residential real estate loans, or in the selling, brokering, or apprais-ing of residential real estate property, to discriminate based on race, color, nationalorigin, or religion. The Equal Credit Opportunity Act prohibits lending discriminationbased on the sex, marital or family status, disability, age, race, color, national origin,or public assistance status of the borrower (Federal Reserve Bank of Boston 1993,26–27). Although most analysts believe that blatant discrimination is no longer wide-spread in mortgage financing, there is credible evidence that subtle forms of discrimi-nation do persist (Holloway 1998; Hunter and Walker 1996; Yinger 1995, 1998). More-over, the mortgage transaction is only one part of a complex set of institutional andsocial processes involved in housing markets, and various types of discrimination inother housing transactions and in other markets can affect minority access to home-ownership (Yinger 1995). Segmentation of minorities into unstable or lower-wage jobs,for example, historically has placed minorities at a disadvantage when underwritingguidelines have required applicants to demonstrate long-term employment or a singlesource of income. Furthermore, minority home seekers sometimes experience discrim-ination in their encounters with home sales agents (Turner 1992; HUD 1991). As aconsequence, efforts to eliminate discrimination in mortgage financing must be coor-dinated with a broad-based assault on discrimination in other markets and in thehousing search process (Listokin and Wyly 1998; Vartanian et al. 1995; Yinger 1995).
There have also been efforts to expand the availability of more affordable and flexiblemortgages. The Community Reinvestment Act (CRA) provides a major incentive. En-acted in 1977 and since amended, CRA requires financial institutions to meet theircommunity’s need for credit in low- and moderate-income neighborhoods, consistentwith safe and sound operation of the institution. CRA applies to depository lendinginstitutions with offices in a metropolitan area and with assets above a specific thresh-old. These lenders are required to prepare CRA statements that define their marketcommunity and list the types of credit offered (Fishbein 1992).
Fannie Mae and Freddie Mac, the government-sponsored enterprises (GSEs) that arethe preeminent forces in the nation’s secondary mortgage market, have also beencalled upon to broaden access to mortgage credit and homeownership. The 1992 Fed-eral Housing Enterprises Financial Safety and Soundness Act (FHEFSSA) mandatedthat the GSEs increase their acquisition of primary-market loans made to lower-income borrowers and to areas unserved by private mortgage credit institutions(Carr et al. 1994a, 1994b). Spurred in part by the FHEFSSA mandate, Fannie Mae
Background: Homeownership in the United States 5
The Potential and Limitations of Mortgage Innovation
announced a trillion-dollar commitment in 1994 to help 10 million families—particu-larly those most in need—buy homes of their own (Fannie Mae 1997, 2). Freddie Mac,the other dominant secondary-market institution, has pursued similar initiatives.
The result has been a wider variety of innovative mortgage products. The GSEs haveintroduced a new generation of affordable, flexible, and targeted mortgages, therebyfundamentally altering the terms upon which mortgage credit was offered in theUnited States from the 1960s through the 1980s (Gyourko, Linneman, and Wachter1997; Linneman and Wachter 1989; Listokin and Wyly 1998). Moreover, these second-ary-market innovations have proceeded in tandem with shifts in the primary markets:depository institutions, spurred by the threat of CRA challenges and the lure of sig-nificant profit potential in underserved markets, have pioneered flexible mortgageproducts (Listokin and Wyly 1998; Listokin et al. 2000; Schwartz 1998). For years, de-positories held these products in portfolios when their underwriting guidelines ex-ceeded benchmarks set by the GSEs. Current shifts in government policy, GSE ac-quisition criteria, and the primary market have fostered greater integration of capitaland lending markets.
These changes in lending herald what we refer to as mortgage innovation. To betterappreciate the paradigm shift in financing introduced by mortgage innovation, it isinstructive to organize the evolving mortgage products into a typology:
1. Historical Mortgages—standard mortgages as they existed through the two decadesbefore 1990.
2. GSE Standard Mortgages—loans conforming to the current (1990 and subsequent)standard mortgage guidelines of the GSEs.
3. GSE Affordable Mortgages—more affordable and flexible current loans developedby the GSEs (e.g., the Fannie Mae Community Home Buyer’s Program™ [CHBP]and Freddie Mac’s Affordable Gold™ [AG]) that share characteristics such as highloan-to-value ratios (LTVs; value of property) and credit flexibility. Emerging GSEAffordable Mortgages represent further innovations in GSE products that haverecently become available.
4. Portfolio Affordable Mortgages—more affordable and flexible current mortgageinstruments that are held in portfolios by individual lenders. These loans exceedthe parameters of the GSE Affordable Mortgages.
5. Governmental Affordable Mortgages3—publicly backed loans, such as mortgagesinsured by the Federal Housing Administration (FHA; e.g., Section 203), that areoften used by lower-income and minority borrowers.
6 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
3 This study includes the Governmental Affordable Mortgages as contemporary products (i.e., nonhistorical)even though Federal Housing Administration loans have been offered for many decades.
This study examines the potential and limitations of mortgage innovation in fosteringhomeownership by current renters, especially those who are members of traditionallyunderserved populations. While we recognize that homeownership is not for every-one (e.g., lifestyle renters) and that it would be counterproductive to bring the mar-ginally qualified to homeownership if they would shortly thereafter fail, it is instruc-tive to establish the baseline potential of renters to realize homeownership. To thisend, we estimate the ability of specific mortgage instruments, grouped according tothe mortgage typology described above, to expand homeownership. For example, weexamine how many renters can become homeowners through the application of theCHBP, AG, and other GSE Affordable Mortgages. The study estimates the number ofrenter families who would be served by the various loan products—that is, the num-ber of renters who would qualify for a home purchase loan given the confluence ofthe families’ financial resources and the terms of the respective mortgages. Rentersnot realizing homeownership from the alternative loan products are referred to asthe unserved. We apply a mortgage simulation—a modeled mortgage underwriting—to determine who is served and who is unserved.
As noted, the analysis focuses on renter families, defined here as related individualsresiding together or persons living alone. Our information on families comes from the1993 Survey of Income and Program Participation (SIPP). We use the seventh-wavepanel of the 1993 survey, which reports information for 1995. Based on our definitionof family and wave seven of the 1993 SIPP, our data set contains 25,762,939 renterfamilies in the United States. Our estimate varies from the American Housing Sur-vey (AHS)4 estimate for the total number of renters (34,150,000) for several reasons.First, we counted renter families while AHS counted renter households. Second, theAHS estimate of 34,150,000 renters is the total number of renter households in 1995.Our estimate is for families who were renters in 1993 and continued to be renters in1995. Thus, anyone who owned a home in 1993 but rented one in 1995, or who enteredthe SIPP after 1993, or who was under 18 in wave one, would not be counted, for ourpurposes, as a renter in 1993.
The analysis of the served compared with the unserved is made for all renters as agroup and for all renters by race/ethnicity. Our race/ethnicity categories are non-Hispanic white, non-Hispanic black, non-Hispanic other, and Hispanic. These race/ethnicity groups include both native-born Americans and immigrants. Immigrantsinclude all people not born in the United States, regardless of when they arrived. Inaddition, we present numbers for recent immigrants, defined as renter families whose
Background: Homeownership in the United States 7
The Potential and Limitations of Mortgage Innovation
4 The AHS figure is taken from the 1995 American Housing Survey, table 2–1, Introductory Characteristics—Occupied Units (HUD 1997, 42).
householder entered the United States after 1984.5 This last category is a subset ofall renters and is not differentiated by race/ethnicity.6
The analysis is applied at the national level because there is insufficient sample sizeto differentiate mortgage simulations at a more microgeographic level (e.g., metro-politan statistical areas [MSAs]). For example, we cannot with statistical confidencesay that x percent of the unserved are in the Washington, D.C., MSA compared withy percent in the Atlanta MSA. Yet while we do not report separate results at the MSAlevel, we do incorporate important geographic distinctions in our analysis. In gauginghomeownership, we incorporate differences in housing prices and renter incomesacross the four census regions and nine census divisions.7 We also incorporate arealdistinctions in property tax rates, closing costs, and other home-buying expenses foreight major and geographically dispersed MSAs: Atlanta; Chicago; Houston; Los Ange-les; Miami; New York City; Philadelphia; and Washington, D.C. These area variations,however, are internal to the calculations and, as noted in our findings, are reportedon only an aggregate national basis.
We also compare the served and unserved renters by income category, and, in this re-gard, divide renters into four relative income groups: low, moderate, middle, andupper. Low-income families earn between 0 percent and 49 percent of median familyincome, moderate-income families 50 percent to 79 percent; middle-income families80 percent to 119 percent, and upper-income families at least 120 percent. The respec-tive percentages of median income that define each income group are for a family offour, and adjustments are applied for smaller and larger family sizes.8
Our approach both builds from and extends prior literature on mortgage simulationand other relevant studies (Bogdon and Can 1997; Gyourko and Tracy 1999; Herbert1995; Linneman and Megbolugbe 1992).
8 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
5 The SIPP uses the following categories to define periods of immigration: (1) 1911–59, (2) 1960–64, (3) 1965–69,(4) 1970–74, (5) 1975–79, (6) 1980–81, (7) 1982–84, and (8) 1985–93. Our recent immigrant groups are thoseindicating response (8) above.
6 Non-Hispanic white, non-Hispanic black, Hispanic, and non-Hispanic other renter families are discrete. Com-bined, they sum to all renter families. Recent immigrant families, those entering the United States after 1984,overlap with the previously listed race/ethnic groups (e.g., non-Hispanic whites and non-Hispanic blacks).
7 The census areas are the Northeast Region (New England and Middle Atlantic Divisions), Midwest Region(East North Central and West North Central Divisions), South Region (South Atlantic, East South Central, andWest South Central Divisions), and West Region (Mountain and Pacific Divisions).
6 For example, for a family of two, the four member–based median family income is reduced by 20 percent; for afamily of seven, the median is increased by 24 percent.
LITERATURE REVIEW
In a recent paper, Calhoun and Stark spoke of “synthetic loan underwriting simula-tions as studies determining the number of households that would qualify to pur-chase a home under various mortgage assumptions” (1997, 3–4). The most relevantsynthetic underwriting research has been produced by the National Association ofHome Builders (NAHB 1997), the National Association of Realtors (NAR 1998), Sav-age (1997, 1999), Savage and Fronczek (1993), and Calhoun and Stark (1997). Anotherrelevant study by Galster et al. (1996) examined the potential size of the homeown-ership market. Each of these studies is discussed below and summarized in table 1.(See appendix A for a more detailed description.)
Prior Studies
Industry Estimates of Housing Affordability
The intent of the NAHB and NAR work is to develop indices—the NAHB’s HousingOpportunity Index (HOI) and the NAR’s Housing Affordability Index (HAI)—of rela-tive housing affordability over time and by place. The higher the HOI and HAI values,the greater the housing affordability. The HOI and HAI tap a variety of data sourcesand include numerous underwriting considerations. However, each measure has limi-tations. The HAI considers the affordability of only median-priced housing (the HOIconsiders the distribution of home prices), and both the HAI and HOI limit their af-fordability analysis to median-income families.9 Our research design seeks a morediverse range of potential housing selection (i.e., not just the median-priced home) anda wider pool of would-be home buyers (i.e., not just median-income families). Further,the HOI and HAI approaches were never intended to be comprehensive financial un-derwriting models. The HAI considers only principal and interest (PI) payments aspart of the housing expense–to-income (front-end) ratio. It omits property taxes (T),property insurance (I), and mortgage insurance (MI), all of which are included intypical front-end ratios used by underwriters.10 Typical mortgage qualification pro-cesses also consider the combination of housing and other debt as a share of income—commonly known as the back-end ratio. Neither the HAI nor the HOI factors in theback-end ratio.
These indices have other limitations. Research has shown that limited assets constrainrenters’ ability to purchase a home because they lack sufficient resources for the downpayment, closing costs, and other related outlays (Duca and Rosenthal 1994; Linne-man and Wachter 1989). Yet neither the HAI nor the HOI considers renters’ assets:
Literature Review 9
The Potential and Limitations of Mortgage Innovation
9 One HAI variation, for first-time home buyers, targets units priced at 85 percent of the median and uses themedian income of families with heads of household aged 25 to 44.
1o The HOI compensates, in part, for this by applying a front-end maximum ratio, 25 percent, that is lower thanthe typical maximum allowed (28 percent).
10 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 1.
Syn
thet
ic U
nd
erw
riti
ng
Stu
die
s an
d H
om
eow
ner
ship
Aff
ord
abili
ty In
dic
es
NA
HB
Hou
sing
NA
RH
ousi
ng A
fford
abili
ty I
ndex
(H
AI)
Sav
age
(199
7, 1
999)
Opp
ortu
nity
FR
M,a
AR
M,b
and
Firs
t-tim
eS
avag
e an
dC
alho
un a
ndC
hara
cter
istic
Inde
x (H
OI)
Com
posi
te H
AIs
Hom
e-B
uyer
HA
IFr
oncz
ek (
1993
)S
tark
(19
97)
Targ
et h
ome
buye
r(s)
Med
ian-
inco
me
fam
ilyM
edia
n-in
com
e fa
mily
25-
to 4
4-ye
ar-o
ld
All
owne
rs a
nd r
ente
rs
All
rent
ers
diffe
rent
iate
dre
nter
sdi
ffere
ntia
ted
by
by r
ace
and
othe
r in
com
e, r
ace,
and
ch
arac
teris
tics
othe
r ch
arac
teris
tics
Targ
et h
ousi
ng u
nit(
s)N
ew a
nd e
xist
ing
hom
e M
edia
n-pr
iced
exi
stin
g 85
% o
f med
ian-
Crit
erio
n ho
mes
—
Max
imum
-affo
rdab
lesa
les
hom
e so
ldpr
iced
exi
stin
g in
clud
ing
low
-pric
ed
hom
es:5
0th
hom
e so
ld(1
0th
perc
entil
e),
perc
entil
e co
mpa
red
mod
estly
pric
ed (
25th
w
ith ta
rget
-pric
ed
perc
entil
e), a
nd
hom
es s
ugge
sted
m
edia
n-pr
iced
hom
es
by c
onsu
mpt
ion
of
(50t
h pe
rcen
tile—
and
sim
ilarly
situ
ated
m
axim
um-a
fford
able
ow
ners
ho
mes
Geo
grap
hic
area
(s)
200
larg
est
Uni
ted
Sta
tes/
Uni
ted
Sta
tes/
Uni
ted
Sta
tes/
Uni
ted
Sta
tes/
cens
us
met
ropo
litan
are
asce
nsus
reg
ions
cens
us r
egio
nsce
nsus
reg
ions
/div
isio
nsre
gion
s/di
visi
ons
Fin
anci
al Q
ualif
icat
ion
Con
side
ratio
ns1.
Mor
tgag
e ty
peF
RM
-AR
M C
ompo
site
Var
ies:
FR
M, A
RM
, F
RM
-AR
M
FR
M
FR
Man
d F
RM
-VR
Mc
Com
posi
te
Com
posi
te2.
Mor
tgag
e LT
V90
%80
%90
%95
% (
FH
A—
95%
–97%
)80
%–1
00%
(ra
nge)
3.D
own
paym
ent
10%
20%
10%
5% (
FH
A—
3%–5
%)
20%
–0%
(ra
nge)
4.H
ome-
buye
r as
sets
No
No
No
Yes
Yes
5.H
ousi
ng e
xpen
ses
a.P
rinci
pal a
nd
Yes
Yes
Yes
Yes
Yes
inte
rest
b.P
rope
rty
taxe
sYe
sN
oN
oYe
sIn
dire
ctly
c.H
ome
insu
ranc
eYe
sN
oN
oYe
sIn
dire
ctly
d.M
ortg
age
No
Not
app
licab
led
Yes
Yes
Indi
rect
lyin
sura
nce
6.M
axim
um
25%
25%
25%
28%
(F
HA
—29
%)
28%
–38%
(ra
nge)
perc
enta
ge o
f in
com
e al
low
ed
for
hous
ing
expe
nses
7.E
xist
ing
debt
N
oN
oN
oYe
sYe
s
Literature Review 11
The Potential and Limitations of Mortgage Innovation
Tabl
e 1.
Syn
thet
ic U
nd
erw
riti
ng
Stu
die
s an
d H
om
eow
ner
ship
Aff
ord
abili
ty In
dic
es (
cont
inue
d)
NA
HB
Hou
sing
NA
RH
ousi
ng A
fford
abili
ty I
ndex
(H
AI)
Sav
age
(199
7, 1
999)
Opp
ortu
nity
FR
M,a
AR
M,b
and
Firs
t-tim
eS
avag
e an
dC
alho
un a
ndC
hara
cter
istic
Inde
x (H
OI)
Com
posi
te H
AIs
Hom
e-B
uyer
HA
IFr
oncz
ek (
1993
)S
tark
(19
97)
8.M
axim
um
Not
con
side
red
Not
con
side
red
Not
con
side
red
36%
(F
HA
—41
%)
33%
–43%
(ra
nge)
pe
rcen
tage
of
inco
me
allo
wed
fo
r ho
usin
g ex
pens
es a
nd
debt
9.C
redi
t rec
ord/
othe
r N
oN
oN
oN
oN
oun
derw
ritin
g va
riabl
es
10.C
losi
ng c
osts
N
oN
oN
oYe
sN
oco
nsid
ered
11.M
ajor
dat
a E
xper
ian,
Cen
sus
Cen
sus
Cen
sus
Cen
sus,
SIP
P,
NS
FH
,eot
hers
sour
ces
man
y ot
hers
aF
ixed
-rat
e m
ortg
age.
bA
djus
tabl
e-ra
te m
ortg
age.
cV
aria
ble-
rate
mor
tgag
e.d
Mor
tgag
e in
sura
nce
is n
ot r
equi
red
for
loan
s w
ith a
t le
ast
a 20
per
cent
dow
n pa
ymen
t.e
Nat
iona
l Sur
vey
of F
amili
es a
nd H
ouse
hold
s.
Each merely assumes that renters will be able to come up with a 20 percent downpayment and closing and other costs (HAI) or a 10 percent down payment and closingand other costs (HOI, as well as the first-time home buyer under HAI). Althoughdifficulty in establishing a credit record and a blemished credit experience also mayconstrain a renter’s ability to purchase a home, especially among minority and LMIhouseholds (Listokin and Wyly 1998), this underwriting factor—admittedly difficultto obtain—is not considered in the HAI and HOI.
In short, the HOI and HAI serve the role they were designed for: they are roughgauges of relative affordability over time and by place. In the current investigation,we seek a more refined synthetic underwriting approach that will inform us, for ex-ample, about the share of African-American, white, and Hispanic families that canafford a home using a GSE Affordable Mortgage compared with the share of thesefamilies that can afford a home using a Portfolio Affordable Mortgage.
Savage and Fronczek’s Estimates
A more expansive analysis is offered by Savage (1997, 1999) and Savage and Fronczek(1993) in their Who Can Afford to Buy a House periodic series. The authors analyzeaffordability for both owners and renters. They examine these groups using a richdatabase, the SIPP. Housing affordability is gauged for a range of six alternativelypriced units, termed criterion homes by Savage and Fronczek. The criterion homesinclude units priced at various ranges: the median, 25th percentile (modestly pricedhouse), 10th percentile (low-priced house), new housing, and so on. Savage and Fron-czek specify a criterion home and then determine its affordability on the basis of apotential buyer’s financial profile (e.g., income, debt, assets). They also calculate themaximum home affordable based on the applicant’s financial characteristics.
Savage and Fronczek’s financial model is comprehensive. Two alternative mortgageinstruments—conventional and FHA—are factored with an assumed fixed interestrate and a 30-year term. All of the components of the front-end ratio—PI, T, I, MI—are calibrated. Existing debt is considered as well, thereby permitting calculation ofa would-be buyer’s total debt obligations and back-end ratio. These front-end andback-end costs are compared with the maximum front-end and back-end ratios allow-ed in conventional mortgages (28 percent/36 percent) and FHA mortgages (29 per-cent/41 percent). Similarly, Savage and Fronczek match the down payment require-ments for conventional and FHA mortgages against the actual assets available tovarious individuals and households. Closing costs are also compared with the would-bebuyer’s available assets. In addition, Savage and Fronczek assemble data for keyvariables (e.g., income, assets, debt) from the SIPP.
Much valuable information is contained in the Savage and Fronczek analysis, andtheir contribution is critically important to the field. However, their research haslimitations. Credit record—an important underwriting consideration, especially forthose at the economic margin—is not considered. In practice, however, “lenders placegreat emphasis on credit scores as the first step in determining whether the house-
12 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
hold can qualify for an A [credit] mortgage” (Caplin et al. 1997, 2.1). Additionally,although dual mortgage instruments—conventional and FHA—are incorporated bySavage and Fronczek, there is a much richer mosaic of loan products with respect tovarying LTVs, front-end and back-end ratios, credit acceptability, and the like.
The use of criterion-priced homes is also limiting. Although the use of six criterionhomes is preferable to, say, the HAI’s single median-priced property, the appropriatelinkage between these levels of housing consumption and various household typesand incomes (Follain 1999) remains unclear. For example, what is the “appropriatehome” for a middle-aged household earning twice the median income? Should theassociation here be to the median-priced criterion home, a new house, or an upscaleunit? A single-person household with an annual income of $20,000 is not likely toseek the same type of house as a household that has the same income but consists ofa married couple with two children. Should the “appropriate home” be the same foryoung white, African-American, and Hispanic renters? Should it be the same in allmetropolitan areas?
Calhoun and Stark’s Approach
This discussion points to the many influences that bear on specifying the referencehouse in any housing affordability analysis—that is, the appropriate pairing of agiven household to a given housing unit. In their paper titled “Credit Quality andHousing Affordability of Renter Households,” Calhoun and Stark (1997) describe an-other approach to specifying the reference house price. These authors also use a data-base that heretofore has not been tapped in determining housing affordability—theNational Survey of Families and Households (NSFH). Calhoun and Stark apply multi-variate regression to the NSFH data to specify reference house prices that can bereasonably paired with different renters based on the observed housing consumptionbehavior of similarly situated owners. Thus, a better-educated, higher-income renterliving in a more affluent area would be linked with a more expensive reference housethan would a less educated, lower-earning renter residing in a less affluent location.In both instances, African-American renters would be paired with less expensive unitsthan would their white counterparts—reflecting African Americans’ historically moremodest housing consumption (Galster, Aron, and Reeder 1999).
Next, Calhoun and Stark calculate the maximum-priced home affordable to eachrenter household. This is accomplished by linking a household’s financial resourcesto mortgage terms. The authors assume a 30-year fixed-rate mortgage with a rangeof terms; they do not use the specific requirements of actual mortgage products. Cal-houn and Stark incorporate affordability results with mortgage LTVs ranging from80 percent to 100 percent (20 percent to 0 percent down payments), front-end ratiosfrom 28 percent to 38 percent, back-end ratios from 33 percent to 43 percent, and in-terest rates from 7 percent to 13 percent. The confluence of these items and the renterhousehold’s financial resources produces the maximum-priced housing unit affordableto each renter. Calhoun and Stark then compare the target-priced home for each rent-er household—that is, the one based on historically observed consumption by compa-
Literature Review 13
The Potential and Limitations of Mortgage Innovation
rable owners—with the maximum-priced unit affordable to the renter household. Ifthe maximum-priced unit is more costly than the target-priced unit, the household isable to afford homeownership.
Calhoun and Stark have advanced the discussion on the subject, and our researchmodel is grounded, in part, on their conceptual framework. Still, there are somelimitations to their analysis. They did not factor closing costs in their financial calcu-lations, even though closing costs are not inconsequential outlays: depending on loca-tion, closing costs can amount to 3 percent to 4 percent of the cost of the house (Caplinet al. 1997). There is also a far-from-transparent identification by Calhoun and Starkof the housing costs included in the front-end ratio. As noted earlier, these housingcosts encompass principal and interest, taxes, property insurance, and mortgage in-surance. Ideally, these would be individually expensed—a procedure included bySavage and Fronczek. Calhoun and Stark did not take this approach; instead, theycharge directly only the principal and interest expenses, and the other outlays areindirectly subsumed in the range of the front-end ratios that they factor.
Galster’s Model
Galster and colleagues suggest an alternative approach for specifying the reference-price house in their 1996 study entitled Estimating the Number, Characteristics, andRisk Profile of Potential Homeowners. This important analysis developed a modelthat predicted the likelihood of renter households moving into homeownership duringan 18-month period. In formulating that model, the basis was the observed behaviorof white suburban renters—those thought to be the least constrained in buying ahome. The renter-to-homeowner transition of the white suburban subgroup was thenapplied to all households to gauge the potential overall homeowner market. Galsterand colleagues project that just over 600,000 renter households (2 percent) wouldbecome homeowners over an 18-month period if the underwriting practices found inwhite suburban areas were employed uniformly across the nation.
This is significantly different from the model applied by Calhoun and Stark. Theseresearchers paired a household with a reference-price house based on the observedconsumption patterns of owners in the same racial/ethnic group. Thus, if minoritiesin the past bought lower-priced homes, then minorities are paired with the same ref-erence-price house in the affordability analysis. In contrast, Galster and colleaguesmodel potential behavior on the basis of the consumption patterns of the least-con-strained population—white suburbanites who entered the homeownership marketin the favorable economic and housing market conditions of the mid-1990s. Galsterand colleagues thus pair minority renters with the higher-priced housing sought bywhite renters in the suburbs and reveal the size of the unserved market in a scenariofree of all discrimination.11
14 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
11 Other researchers have grappled with the issue of pairing certain types of housing in terms of price, amenities,location, and so on with various categories of households (e.g., minority versus majority; higher- versus lower-
Context of the Current Investigation and Prior Literature
It is useful at this point to place our current investigation in the context of relevantpast literature and research:
1. Because the purpose of this study is to identify who is served and who is notserved by mortgage innovation, we need to synthetically model the mortgageprocess as closely as possible. For this reason, the HOI and HAI approaches do notsuffice (they were not intended to perform such a duty), but the work of others ismore relevant.
2. We model our synthetic underwriting financial calculations on the Who Can Affordto Buy a House approach, the most comprehensive synthetic macroscale under-writing research done to date (Savage 1997, 1999; Savage and Fronczek 1993).Savage and Fronczek itemize the housing expenditures of the front-end ratio (PI,T, I, MI); include debt, so that back-end ratio considerations can be incorporated;factor in closing costs; and in other ways approximate the realities of mortgagequalification. We also use the same database (the SIPP) as that used by Savageand Fronczek.
3. We build upon the Savage and Fronczek financial calculations by applying thesynthetic underwriting for a menu of mortgages organized by the typology of loans(GSE Standard, GSE Affordable, Portfolio Affordable, Governmental Affordable,and Historical Mortgages). By comparison, Savage and Fronczek incorporated onlytwo mortgage types: conventional and FHA. In this regard, our work functionallyresembles the wide range of mortgage terms incorporated by Calhoun and Stark(1997); however, we target specific mortgage products, whereas Calhoun and Starkdid not.
4. We further augment the Savage and Fronczek (and Calhoun and Stark) analysesby incorporating credit record information into the synthetic underwriting. Sinceour credit data and credit score assignment have severe limitations, our credit un-derwriting must be regarded as an exploratory effort to be refined in the future.(The results of the pilot credit underwriting are reported in appendix B.)
5. Savage and Fronczek measure housing affordability with respect to the house-hold’s ability to purchase a benchmark (i.e., criterion) home and through dollarvalues (e.g., a mean- or median-priced affordable home under different mortgageproducts). We follow Savage and Fronczek in this regard and use both a bench-mark approach and a dollar measure of home-buying capacity.
Literature Review 15
The Potential and Limitations of Mortgage Innovation
income). For instance, Linneman and Wachter (1989, 391) related the housing services a family desires to pur-chase (with that capitalized value designated as V*) as a function of the family’s income (labeled as I) and a vec-tor of preference variables labeled X, with that overall relationship expressed as V* = V(I, X; b), where b is avector of parameters.
6. Although our financial analysis is an enhanced version of the basic Savage andFronczek approach, we depart somewhat from these authors with respect to thebenchmark housing used in analyzing affordability. Savage and Fronczek examinethe affordability of criterion homes. As there is an inherent attraction to using thecriterion homes as a benchmark, we present estimates for the criterion homes atthe metropolitan median, as well as estimates for modestly priced homes (25thpercentile) and low-priced homes (10th percentile).12 Yet, as discussed earlier, suchan arbitrary price-point approach does not consider the appropriate-shelter frameof reference for any given family. We therefore also link a renter family to a hous-ing unit based on observed consumption, as introduced by Calhoun and Stark andothers (e.g., Linneman and Wachter 1989). Our basic consumption model identifiesthe price of the housing unit each renter family would seek if its behavior conform-ed to the behavior of comparable renters who previously moved to homeowner-ship. The unit with a price calibrated in this way is termed the target house.
As noted, there can be various target houses. Galster and colleagues (1996) use asa reference the housing consumption of the least constrained group—white renterswho recently attained homeownership. Calhoun and Stark present affordabilityanalyses based on the consumption of all recent buyers. Galster and colleagues’standard benchmarks are appropriate in a world unconfounded by historicalracial and other barriers, whereas Calhoun and Stark orient their benchmarks tothe world as it exists.
In the current investigation, we benchmark the target house according to the ap-proach used by Calhoun and Stark—that is, the analysis is based on the consump-tion of all renters who recently moved to homeownership. We note, however, thattarget prices would be higher were the Galster and colleagues approach used.
16 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
12 While Savage and Fronczek examine six different criterion homes, our investigation will use three that span themore modestly priced housing spectrum.
SIMULATION FRAMEWORK AND MODELS
Our overall simulation framework and the data used to calibrate its component mod-els are summarized in figure 1. The framework comprises three major models. TheHousing Consumption Model estimates the price of the housing unit that each renterfamily would seek if its behavior conformed to that of all first-time home buyers enter-ing the market in the mid-1990s. The Housing Consumption Model yields a target orreference unit of housing consumption for each renter family. To provide comparabilitywith Savage and Fronczek, criterion-priced homes are considered here as well. Thetarget or criterion home thus provides the reference or benchmarked unit of housingconsumption. This is followed by the Mortgage Model, in which alternative mortgageinstruments are used to estimate the maximum purchase price for which each renterfamily could qualify based on its financial characteristics. The Mortgage Model alsoprovides other measures of home-buying capacity for each mortgage product, such asthe aggregate value of housing that could be purchased using the product. Given thatcredit is a key underwriting consideration, a pilot credit submodel, admittedly withmany limitations, informs the Mortgage Model. Finally, the Affordability Analysismodel compares a family’s reference house prices with the maximum-priced house forwhich it could qualify and provides other measures of affordability.
The overall simulation framework provides a measure of reference home-buying ca-pacity because it establishes a benchmark (e.g., individually calibrated target orprice-point criterion home) and then determines whether that referenced housing unitcan be afforded, given the confluence of the renters’ financial profile and the mortgageterms. There is much to be gained from such an approach; by establishing a bench-mark of housing consumption, we gain a point of reference by which we can gaugeachievement. If the renter can afford the reference home with a given mortgage prod-uct, the renter is served; if not, the renter is unserved. Thus, the loan product may bejudged in terms of its capacity to allow the renter to achieve the housing benchmark.
While there is much logic to the establishment of a reference point for measuringhousing affordability, such a reference point does not fully convey the ability of agiven mortgage to expand purchasing power, because purchasing capacity below thebenchmark is noted but not quantified. Assume, for instance, that a reference pur-chase price is $100,000 and that mortgage product A offers a given renter $50,000in purchasing power while mortgage product B conveys $75,000. If one simply notesthat both products fail to achieve the $100,000 benchmark, product B’s greater capac-ity to expand purchasing power is unrecognized.13 If, however, we incorporate a dollarhome-buying capacity, then the second product’s greater potential can be measured.
Consequently, in addition to the reference home-buying capacity, our AffordabilityAnalysis includes absolute home-buying capacity. The latter provides monetized mea-sures of the purchasing power afforded by the respective mortgage products. These
Simulation Framework and Models 17
The Potential and Limitations of Mortgage Innovation
13 Calhoun and Stark (1997) measure the relative level of achievement through the use of ratios. For more detail,see appendix A.
18 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Fig
ure
1.H
ou
sin
g A
ffo
rdab
ility
un
der
Alt
ern
ativ
e M
ort
gag
es:
Ove
rall
An
alyt
ic F
ram
ewo
rk
* Fa
ir, I
saac
& C
ompa
ny,
Inc.
Ho
usi
ng
Co
nsu
mp
tio
n M
od
elC
orre
late
the
hous
e pu
rcha
se p
rice
with
the
soci
odem
ogra
phic
char
acte
ristic
s of
fam
ilies
who
mov
ed fr
om r
entin
g to
ow
ning
betw
een
1993
and
199
5 }M
ort
gag
e M
od
elC
alcu
late
the
max
imum
hou
se p
rice
for
whi
ch r
ente
rs c
an q
ualif
yba
sed
on fi
nanc
ial c
hara
cter
istic
s }H
OU
SIN
GA
FF
OR
DA
BIL
ITY
AN
ALY
SIS
:1.
Ref
eren
ce h
ome-
buyi
ng c
apac
ity:a
com
paris
on o
f val
ues
from
the
Hou
sing
Con
sum
ptio
n an
d M
ortg
age
Mod
els.
2.A
bsol
ute
hom
e-bu
ying
cap
acity
:a ta
bula
tion
of to
tals
from
the
Mor
tgag
e M
odel
.
Sav
age’
s C
riter
ion
Hom
e A
ppro
ach
Est
imat
ed M
axim
um
Ho
use
Pu
rch
ase
Pri
ceC
alcu
late
d fo
r ea
ch f
amily
for
eac
h m
ortg
age
inst
rum
ent;
code
d “N
A”
if ap
plic
ant
cann
ot b
e qu
alifi
ed
Co
re D
atab
ase
1993
SIP
P
Cre
dit
Sco
reM
od
el
Fro
m m
ortg
age
com
pany
FH
Asa
mpl
e:•
Pre
dict
edF
ICO
*sc
ores
Pre
dic
ted
Ho
use
Pu
rch
ase
Pri
ces
Ref
eren
ce g
roup
:•
Hom
e bu
yers
199
3–19
95
Com
paris
on r
ente
rs:
•N
ativ
e-bo
rn w
hite
•N
ativ
e-bo
rn b
lack
•N
ativ
e-bo
rn H
ispa
nic
Con
sum
ptio
n M
odel
Impl
ied
Pur
chas
e P
rice
10th
Per
cent
ileof
MS
A25
th P
erce
ntile
of M
SA
Med
ian
ofM
SA
His
toric
al
Pre
dic
ted
Pu
rch
ase
Pri
ce V
aria
ble
s
Alt
ern
ativ
e M
ort
gag
e In
stru
men
ts
GS
ES
tand
ard
GS
EA
fford
able
Por
tfolio
Affo
rdab
leG
over
nmen
tA
fford
able
Hou
se P
urch
ase
Pric
e =
f(h
ousi
ng-n
eed
varia
bles
):
•In
com
e•
Age
•H
ouse
hold
siz
e•
Mar
ital s
tatu
s•
Chi
ldre
n•
Edu
catio
n•
Occ
upat
ion
•S
elf-
empl
oym
ent
•P
ublic
ass
ista
nce
•R
esid
entia
l loc
atio
n
coef
ficie
ntes
timat
es
▼
measures include the total home-purchasing power of current renters under the dif-ferent mortgage instruments as well as the mean/median value of house prices afford-able to renters using the respective products.
In measuring the absolute home-buying capacity, we stipulate a threshold of consump-tion related to the “lumpy” and expensive consumption of housing; that is, we establisha minimum level of buying power required for inclusion in the absolute home-buyingcapacity measure. In this study, we stipulate the low-priced house as the minimumthreshold. If the low-priced house is, for example, priced at $40,000, then buying powerof $40,000 and up with no ceiling (unless otherwise indicated) would count toward theabsolute home-buying capacity.
The absolute home-buying capacity builds off the Mortgage Model. As noted, thatmodel estimates the maximum purchase price for which each renter could qualify,based on his or her financial characteristics. A maximum purchase price estimate thatexceeds the stipulated threshold equals the renter’s total absolute home-buying capac-ity. The aggregate of each of these renter calculations constitutes the total absolutehome-buying capacity. Mean/median values can be similarly derived.
In sum, the overall research framework comprises the Housing Consumption Model,the Mortgage Model, and the Affordability Analysis, with the latter incorporating boththe reference and the absolute home-buying capacities. The paired results of theHousing Consumption Model and the Mortgage Model provide the reference home-buying capacity. The Mortgage Model alone is the basis for the absolute home-buyingcapacity.
Ideally, an affordability analysis would account for the full array of circumstancesconfronted by different kinds of families attempting to locate and purchase homes.Unfortunately, few existing data sources provide sufficient information or sample sizeto permit such a comprehensive analysis, and researchers must therefore evaluatetradeoffs between spatial, temporal, and thematic coverage.14 In our judgment, theSIPP, released by the U.S. Census Bureau (U.S. Bureau of the Census 1996), providesthe most balanced compromise among these many tradeoffs. In appendix C, we pre-sent detailed evaluations of the SIPP and several alternative information sources.For the bulk of the analysis, the SIPP constitutes our core database, and we minimizethe use of imputed information from other sources. Almost all of the Housing Con-sumption Model (i.e., derivation of the target-priced homes) is derived from the SIPP,
Simulation Framework and Models 19
The Potential and Limitations of Mortgage Innovation
14 For example, the microdata samples from the decennial census of population and housing provide the largestpossible sample of individuals and families, and are the only source of information that includes rigorous spatialsampling protocols to illuminate local variations in population characteristics within metropolitan areas (U.S.Bureau of the Census 1993). Yet the Public Use Microdata Sample (PUMS) files are now a decade old; they aredesigned only for cross-sectional analyses, and the census long form includes no detailed questions regardingassets, employment history, or other factors that are crucial in the underwriting decision. In our initial researchdesign we considered the possibility of assigning to PUMS more recent estimates of population characteristicsfrom the Current Population Survey and other sources, and adding variables from other databases. That approach,however, would require the adoption of many unrealistic assumptions, and there is currently no reliable meansof cross-validating the accuracy of imputed data from an independent source.
and most of the Mortgage Model (with the exception of the credit submodel) is basedon renter financial data from the SIPP. In the following sections, we calibrate andapply the Housing Consumption Model and the Mortgage Model and describe ingreater detail their inclusive data sets.
20 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
MODEL CALIBRATION AND ANALYSIS
Housing Consumption Model
Model Framework and Data
The Housing Consumption Model builds on the work of Galster et al. (1996), Calhounand Stark (1997), Linneman and Wachter (1989), Gyourko, Linneman, and Wachter(1997), Jones (1995), and other literature. The development of details on this model isdescribed in a separate paper (Wyly et al. 1999). In brief, after deciding to enter thehomeownership market, a family (as earlier defined) chooses a desired level of hous-ing consumption that is influenced by income (I), wealth (W), demographic and lifecourse factors (X), and regional housing market conditions (R). These factors relateas follows:
H = f (I, W, X, R; β), (1)
where f is commonly modeled in a linear regression framework. In our model, the de-pendent variable is defined as the natural log of house value, and we refer to H asthe target house price that captures the various needs, financial resources, and hous-ing market circumstances of different families.
We predict the expected housing consumption of current renters, H, on the basis ofobserved choices of renter families that purchased homes during the recovery of themid-1990s. The analysis involves three main procedures. First, we use the first andseventh waves of the 1993 SIPP files to identify those families that moved from rent-ing to homeownership between 1993 and mid-1995. (The 1993 SIPP contained a finalsample of 4,565 renter families, of which 456 moved from renting to owning duringthe first and seventh waves of the survey.) Second, we calibrate a regression modelthat relates reported house prices to a vector of social, demographic, and local hous-ing market characteristics; the sample used consists of all families that moved fromrenting to owning a home (the Calhoun and Stark [1997] approach). Finally, the co-efficients from this model are applied to the remaining renters in the core SIPP dataset, providing an estimate of expected housing consumption for all renters in the sam-ple. Together with criterion homes (the Savage [1997, 1999] approach), these target-priced homes are then used as the reference homes for each renter family.
The Housing Consumption Model, which develops the target-priced homes, certainlyhas limitations. On theoretical grounds, one may question the assumption that ob-served choices made by recent buyers can be used to describe the behavior of currentrenters if they were to move into homeownership. On methodological grounds, thetechnique relies on a fairly small sample of recent home buyers. Nevertheless, thetarget house price methodology provides a valuable alternative to arbitrary affordabil-ity thresholds. As with the simulations developed by Galster et al. (1996), our methodis an explicit attempt to use the behavior of recent home buyers as a critical bench-mark by which to measure the effects of alternative mortgage market innovations.
Model Calibration and Analysis 21
The Potential and Limitations of Mortgage Innovation
Housing Consumption Model Results
Table 2 defines the variables for the Housing Consumption Model and Table 3 presentsresults of ordinary least-squares estimates of equation (1). It estimates the level ofhousing consumption that would be expected of current renters if they made the samechoices as similar families who moved into homeownership during the mid-1990s.Demographic, regional, and financial variables are comparable to those used in sim-ilar studies of housing consumption and tenure choice (Calhoun and Stark 1997;Galster et al. 1996; Linneman and Wachter 1989; Quercia, McCarthy, and Wachter1998). Diagnostics suggest a fairly robust model fit, in line with similar studies.15
Variations in housing consumption reflect differences in individual preferences andneeds, family financial resources, and regional and local housing market conditions.White-collar workers typically seek out more expensive homes than do blue-collarworkers, although this difference is not statistically significant. House values riseamong families with higher incomes and assets, and for people with higher levels ofeducation. Model results imply an income elasticity of approximately 20 percent withrespect to house price: in other words, a 10 percent increase in family income amongrecent home buyers leads to a house purchase price increase of approximately 2 per-cent. By contrast, the elasticity measure for family assets is remarkably low: a 10percent increase in assets translates to only 0.1 percent capitalized into higher homeprices. Variations in family assets may be more important in the tenure decision thanin the desired amount of entry-level housing consumption once the decision is madeto purchase a home. Higher assets may simply be used to reduce mortgage debt or tomaintain liquidity.
Regional housing market variations correlate with wide contrasts in housing consump-tion. Not surprisingly, house prices rise in more populated metropolitan areas, withcosts at the peak of the urban hierarchy running 49 percent above prices in nonmetro-politan areas. Regional variations inscribe a clear bicoastal pattern, with prices inNew England and on the West Coast standing 50 percent above those in the UpperMidwest. Even after controlling for all other factors in the model, however, a premiumis still apparent in the distinctive housing markets of Washington, D.C. (29 percent),and New York (36 percent).
Racial and ethnic differences in house prices are not significant, although the relevantparameter estimates are in the hypothesized direction. This result may be related tothe relatively small number of minority home buyers (70) in the sample. But it is alsoconsistent with Gyourko, Linneman, and Wachter’s (1996) finding that racial differ-ences in homeownership are not significant among families with sufficient resourcesto meet underwriting and down payment requirements. Gyourko, Linneman, andWachter (1996) also find, however, that racial differences in homeownership are sig-nificant among families constrained by low assets.
22 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
15 The adjusted multiple coefficients of determination (0.59 and 0.60) are comparable to Calhoun and Stark’s(1997) value of 0.45 and Quercia, McCarthy, and Wachter’s (1998) value of 0.48.
Model Calibration and Analysis 23
The Potential and Limitations of Mortgage Innovation
Table 2. Variable Definitions for Housing Consumption Model
All Renters MovingFull Sample into Homeownership
Standard StandardVariable Name Mean Deviation Mean Deviation
Dependent VariableLHVALUE Natural log of reported house value 11.5 0.475
Financial/human capital characteristicsLFINC Natural log of family income 9.58 1.44 10.3 0.696LASSETS Natural log of family assets 3.10 4.14 5.19 4.41ED Years of education 12.2 2.92 13.3 2.05SELF Self-employed (dummy) 0.0465 0.0526WHTC1 White-collar occupation, worked 0.337 0.550
entire month (dummy)WHTC2 White-collar occupation, worked 0.00480 0.00438
on and off this month (dummy)BLUC1 Blue-collar occupation, worked 0.272 0.270
entire month (dummy)BLUC2 Blue-collar occupation, worked 0.0138 0.00877
on and off this month (dummy)
Demographic characteristicsA15_24a Age 15–24 (dummy) 0.124 0.127A25_34a Age 25–34 (dummy) 0.329 0.515A35_44a Age 35–44 (dummy) 0.219 0.193A45_54a Age 45–54 (dummy) 0.121 0.107A55_64a Age 55–64 (dummy) 0.706 0.0395MARST Married (dummy) 0.314 0.491CHILD Number of children under age 0.743 1.21 0.691 1.06
18 in familyCHILD*A35_44 Interaction term between 0.119 0.0965
CHILD and A35_44CHILD*A45_54 Interaction term between 0.0380 0.0417
CHILD and A45_54
Metropolitan/regional housing market dummiesSMALLb MSA population < 500,000 0.0398 0.0395MEDIUMb MSA population 500,001–1,500,000 0.154 0.158LARGEb MSA population 1,500,001–4,000,000 0.191 0.228V_LARGEb MSA population > 4,000,001 0.284 0.173SO_ESCc East South Central 0.0422 0.0614SO_WSCc West South Central 0.102 0.103SO_SOAc South Atlantic 0.151 0.210WEST_MTNc West (Mountain) 0.0410 0.0548WEST_CSTc West (Coast) 0.203 0.158NE_NEWc New England 0.0570 0.0395NE_MATc Mid Atlantic 0.173 0.101MW_ENCc East North Central 0.161 0.169DC Washington, D.C., MSA 0.0153 0.0263NY New York MSA 0.0979 0.0373
Race/ethnicity (reference category is non-Hispanic whites)BLACK Black reference person 0.139 0.0614HISPAN Hispanic reference person 0.129 0.0614OTHER Reference person of other race 0.0419 0.0263
Our results suggest that minorities’ level of housing consumption is not appreciablydifferent from that of whites with similar financial and demographic characteristics.Nevertheless, this finding flies in the face of widespread empirical evidence and the-oretical expectation that predict reduced consumption among minorities. For someanalysts, these reductions signify racial differences in tastes or preferences, yet a morecompelling case can be made for the effects of housing market processes that limitaccess to high-value suburban submarkets (Adams 1987; Badcock 1994; Gyourko,Linneman, and Wachter 1996; Ladd 1998). Either way, for the purpose of assessingmortgage-related criteria, it is reasonable to assume that the relative effects of con-sumer preference and housing market discrimination will continue to exist in the shortterm. Accordingly, in our subsequent analyses we use a model specification (tables 2and 3) that includes negative race/ethnicity coefficients. This approach reduces pre-dicted housing consumption among minorities, thereby narrowing the gap betweenthese families’ financial resources and the cost of homes deemed suitable for theirneeds.
The model allows us to predict desired housing consumption, which we term the tar-get house. We assign or impute a target house price for each of the 4,10916 remainingrenter families in the 1993 SIPP. To give a sense of that calculation, we report overallresults. Based on the observed experience of first-time buyers who purchased a homebetween 1993 and 1995, the Housing Consumption Model suggests a median target-priced house (in 1995 dollars) of $85,210 and a mean price of $92,781 (table 4). Thismedian estimate stands at 75 percent of the median price of all existing single-familyhomes sold nationwide in 1995 ($112,900) (U.S. Bureau of the Census 1996, 717).This gap reflects the higher purchase prices among trade-up home buyers, whoseaccumulated equity permits larger down payments and corresponding leveraging ofincome to purchase more expensive homes.
24 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
16 The 4,109 remaining renter families is the starting figure of 4,565 renter families in the 1993 SIPP less the 456who moved from renting to owning between the first and seventh waves of the survey.
Table 2. Variable Definitions for Housing Consumption Model (continued)
All Renters MovingFull Sample into Homeownership
Standard StandardVariable Name Mean Deviation Mean Deviation
Sample size 4,509 .456
Source: Authors’ analysis of the first and seventh waves of the 1993 SIPP.a Reference category is age 65 and over.b Reference category is nonmetropolitan areas.c Reference category is Midwest, West North Central.
Criterion-Priced Homes
In addition to the target homes based on the Housing Consumption Model, we con-sider as a benchmark criterion-priced housing. This follows the Savage and Fronczekapproach, and we use three criterion-priced homes: median-priced, modestly priced
Model Calibration and Analysis 25
The Potential and Limitations of Mortgage Innovation
Table 3. Housing Consumption Model for Identifying a Target-Priced House:All Renters Moving into Homeownership
Variable Name Parameter Estimate Standard Error T Value
Intercept 9.07 0.306 29.6**
LFINC 0.197 0.0281 6.99**LASSETS 0.0124 0.00378 3.28**ED 0.0149 0.00754 1.98*SELF 0.249 0.0675 3.69***WHTC1 0.0838 0.0482 1.74WHTC2 0.308 0.223 1.38BLUC1 –0.0841 0.048 –1.75BLUC2 –0.0166 0.160 –0.310
A15_24 –0.394 0.127 –2.32*A25_34 –0.202 0.120 –1.68A35_44 –0.189 0.125 –1.51A45_54 –0.369 0.128 –2.87**A55_64 –0.317 0.135 –2.34*MARST –0.0135 0.0351 –0.386CHILD –0.0697 0.0177 –3.94***CHILD*A35_44 0.138 0.0724 1.91CHILD*A45_54 0.382 0.0945 4.04***
SMALL 0.177 0.0783 2.26*MEDIUM 0.193 0.0454 4.24***LARGE 0.165 0.0419 3.94***V_LARGE 0.401 0.0544 7.37***SO_ESC 0.0896 0.0763 1.17SO_WSC 0.0230 0.0672 0.343SO_SOA 0.211 0.0597 3.53***WEST_MTN 0.219 0.0787 2.79**WEST_CST 0.407 0.0655 6.21***NE_NEW 0.405 0.0881 4.60***NE_MAT 0.176 0.0760 2.32*MW_ENC 0.130 0.0609 2.13*DC 0.262 0.102 2.58*NY 0.307 0.102 3.07**
BLACK –0.0536 0.0625 –0.858HISPAN –0.0881 0.0675 –1.30OTHER –0.0530 0.0911 –0.585
Source: Authors’ analysis of the first and seventh waves of the 1993 SIPP.Note: The dependent variable is the natural log of reported house value. Model F ratio: 20.34 (p ≤ 0.0001). AdjustedR2: 0.590. *p ≤ 0.05. **p ≤ 0.01. ***p ≤ 0.001.
(25th percentile), and low-priced (10th percentile). National criterion home pricesappear in table 5.
The national modestly priced and low-priced criterion homes are, as expected, con-siderably less expensive than the national Housing Consumption Model–ascribedtarget price of current renters ($85,210 median), assuming these renters made thesame choices as similar families who moved into homeownership. After all, rentersbuying homes typically want more than modest- or low-priced units. The nationalmedian-priced criterion home ($84,000) is just slightly less expensive than the con-sumption-ascribed target figure ($85,210 median). Prices of the criterion homes varyconsiderably by census region and division, as shown in table 4. The array of target
26 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 4. Target-Priced and Reference-Priced Houses Used in the Mortgage Simulation Analysis
Reference-Priced Housec (in $)Target-Priced Median Modest LowHousea (in $) (50th (25th (10th
Region/Area Meanb Medianb Percentile) Percentile) Percentile)
Nation 92,781 85,210 84,000 58,500 42,500
Census DivisionNortheast
New England Metro 103,065 99,880 137,500 103,500 75,000 New England Nonmetro 85,435 82,095 109,000 84,000 62,000 Middle Atlantic Metro 119,264 111,188 106,000 65,500 42,500 Middle Atlantic Nonmetro 69,081 73,351 79,000 56,000 40,000
MidwestEast North Central Metro 84,000 82,548 78,500 55,500 38,000 East North Central Nonmetro 62,136 62,221 62,000 43,000 29,000 West North Central 64,818 62,593 73,500 53,000 34,500 West North Central Nonmetro 60,280 58,046 52,000 35,000 22,000
SouthSouth Atlantic Metro 84,835 81,908 84,000 58,500 42,500 South Atlantic Nonmetro 68,064 68,887 68,000 46,000 30,000 East South Central Metro 69,605 71,440 65,500 49,000 32,000 East South Central Nonmetro 58,520 56,748 56,000 40,000 28,000 West South Central Metro 67,299 64,821 67,500 46,500 32,000 West South Central Nonmetro 54,005 53,765 53,000 33,000 17,000
WestMountain Metro 81,973 81,048 85,000 66,000 47,500 Mountain Nonmetro 72,439 76,485 71,000 52,000 35,000 Pacific Metro 117,768 111,605 164,500 108,500 68,500 Pacific Nonmetro 82,140 80,146 101,000 71,000 52,000
a This is the 1995 house value predicted using the Housing Consumption Model.b These statistics are computed for the sample of renter families that did not make the transition to homeownership
during 1993–95.c Reference-priced house estimates by census region are based on Savage’s study (1997). For each reference-
priced house, the national statistics represent the weighted median (based on SIPP family weights) of the regionalestimates; to facilitate comparison with the national median target-priced house, the statistics are computed for thesample of renter families that did not make the transition to homeownership during 1993–95.
and criterion-priced homes indicated in table 4 are the benchmarks to which weapply the Mortgage Model.
Mortgage Model
Mortgage Model Framework
Because a central theme of this study is the potential of mortgage innovation to ex-pand homeownership affordability, it is essential that details of the range and charac-teristics of alternative mortgage products be explained. Mortgages are often dividedinto two categories: governmental (those granted by the government or those havinggovernment backing, such as FHA insurance) and conventional (those that do nothave a public connection). The current investigation briefly examines GovernmentalAffordable Mortgages, most notably FHA loans insured under the Section 203 pro-gram. It also considers several conventional loan instruments. To better understandconventional mortgages, it is instructive to differentiate between three broad cate-gories of mortgages available since approximately 1990 (table 6).
The first is the Standard Mortgage, which is the current basic instrument of homefinance in the United States. The characteristics of this mortgage are heavily influ-enced by the GSEs that dominate the secondary market—Fannie Mae and FreddieMac. As such, we refer to this loan as the GSE Standard Mortgage.
Although it is the envy of other nations, the Standard Mortgage is beyond the finan-cial reach of many disadvantaged Americans. Over the last decade, a more affordableproduct has evolved. Its basic parameters have been set by Fannie Mae and FreddieMac, since primary-market lenders’ willingness to offer new products often dependson the GSEs’ willingness to buy the affordable loans on the secondary market. TheGSEs use different programmatic terms for these loans, for example, the FannieMae CHBP, CHBP with 3/2 Option, and Fannie 97, or Freddie Mac’s AG, AG with3/2 Option, and AG 97. To reflect their function—to reach the financially less advan-taged—and parentage, we refer to these loans as GSE Affordable Mortgages. At theleading edge of this affordable group are evolving products that we term EmergingGSE Affordable Mortgages. Examples are Fannie Mae’s Flex 97 and Freddie Mac’sCommunity Gold. (Because this subset was in a state of flux when we conducted ourresearch in 1998–99 and has evolved further since then, we will focus the followingdiscussion on the more established GSE Affordable Mortgages.)
Model Calibration and Analysis 27
The Potential and Limitations of Mortgage Innovation
Table 5. National Criterion Home Prices
National WeightedCriterion-Priced Home Median Price
Median-priced (50th percentile) $84,000Modestly priced (25th percentile) $58,500Low priced (10th percentile) $42,500
28 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 6.
Lo
an F
inan
cial
Ch
arac
teri
stic
s by
Mo
rtg
age
Typ
e
GS
E
GS
EP
ortfo
lio
Mor
tgag
e C
hara
cter
istic
His
toric
al M
ortg
age
Sta
ndar
d M
ortg
age
Affo
rdab
le M
ortg
age
Affo
rdab
le M
ortg
age
Pu
rpo
seS
erve
all
borr
ower
s/ar
eas
Ser
ve a
ll bo
rrow
ers/
area
sS
erve
low
- to
mod
erat
e-
Ser
ve lo
w-
to m
oder
ate
inco
me
and
unde
rser
ved
inco
me
and
unde
rser
ved
area
sar
eas
Ap
plic
atio
nP
urch
ase
or r
efin
ance
Pur
chas
e or
ref
inan
ceP
urch
ase
or r
efin
ance
—P
rimar
ily p
urch
ase
with
exc
eptio
nsa
Targ
eted
inco
me
No
inco
me
limits
No
inco
me
limits
Max
imum
100
% o
f are
a M
axim
um 8
0%–1
00%
of
med
ian
inco
me—
with
ar
ea m
edia
n in
com
e—w
ith
exce
ptio
nsb
exce
ptio
nsc
Inve
sto
r G
SE
GS
EG
SE
Por
tfolio
Deb
t rat
ios:
fron
t/bac
k-en
d25
%–2
8%/3
3%–3
6%28
%/3
6%33
%/4
0%e
Max
imum
35%
/42%
;(g
uid
elin
e m
axim
um
d)
maj
ority
at 3
3%/3
8%
Min
imu
m d
own
pay
men
t10
% w
ith m
ortg
age
5%fw
ith m
ortg
age
3%–5
%g
3% o
r le
ssin
sura
nce
insu
ranc
e20
% w
ithou
t20
% w
ithou
t m
ortg
age
insu
ranc
em
ortg
age
insu
ranc
e
Min
imu
m b
orr
ower
10%
5%3%
–5%
hM
inim
al—
1%–2
% o
r le
ss
con
trib
uti
on
(“sw
eat e
quity
”m
ay b
e al
low
ed)
Max
imu
m L
TV
90
% w
ith 1
0% d
own
95%
with
5%
dow
n 95
%–9
7%97
% o
r hi
gher
paym
ent
paym
ent
80%
with
20%
dow
n 80
% w
ith 2
0% d
own
paym
ent
paym
ent
Pri
vate
mor
tgag
e in
sura
nce
Req
uire
d fo
r LT
Vs
over
90%
Req
uire
d fo
r LT
Vs
over
80%
Req
uire
dM
ay o
r m
ay n
ot b
e re
quire
d
Inte
rest
rat
eM
arke
tM
arke
tM
arke
tO
ccas
iona
lly 0
.5 p
oint
sbe
low
mar
ket
Res
erve
s2
mon
ths
requ
ired
2 m
onth
s re
quire
dTy
pica
lly 1
mon
th
Freq
uent
ly n
ot r
equi
red
requ
ired—
with
ex
cept
ions
i
Co
un
selin
gN
ot r
equi
red
Not
req
uire
dR
equi
red—
with
R
equi
red
exce
ptio
ns
aT
he G
SE
Affo
rdab
le M
ortg
ages
are
ofte
n fo
rmal
ly r
efer
red
to a
s “c
omm
unity
lend
ing
prod
ucts
.”A
few
of t
he G
SE
Affo
rdab
le M
ortg
ages
(e.
g., F
anni
e 97
,A
fford
able
Gol
d 97
) ar
e lim
ited
to p
urch
ases
(w
ith e
xcep
tions
);m
ost
of t
he o
ther
GS
E A
fford
able
Mor
tgag
es c
an b
e us
ed fo
r ei
ther
pur
chas
es o
r re
fi-na
ncin
g.S
ee t
able
s D
.1 a
nd D
.2.
bIn
com
e m
ay e
xcee
d th
e m
edia
n in
hig
h-co
st a
reas
and
in t
arge
ted
loca
tions
(e.
g.,
high
-min
ority
cen
sus
trac
ts,
cent
ral-c
ity lo
catio
ns).
Model Calibration and Analysis 29
The Potential and Limitations of Mortgage Innovation
Tabl
e 6.
Lo
an F
inan
cial
Ch
arac
teri
stic
s by
Mo
rtg
age
Typ
e(c
ontin
ued)
cV
arie
s by
pro
gram
, w
ith m
axim
um in
com
es r
angi
ng f
rom
rou
ghly
50
perc
ent
to 1
50 p
erce
nt o
f ar
ea m
edia
n in
com
e.d
The
se a
re t
he “
norm
al”
guid
elin
e m
axim
ums
that
can
be
exce
eded
with
com
pens
atin
g fa
ctor
s.e
The
max
imum
gui
delin
e fr
ont-
end
ratio
in t
he F
anni
e M
ae a
fford
able
s is
33
perc
ent;
Fred
die
Mac
has
no
max
imum
fro
nt-e
nd r
atio
.The
max
imum
bac
k-en
d ra
tio is
38
perc
ent
for
Fann
ie M
ae’s
CH
BP
and
36
perc
ent
for
Fann
ie 9
7;Fr
eddi
e M
ac a
llow
s a
40 p
erce
nt g
uide
line
max
imum
.See
tab
les
D.1
and
D.2
.f
The
min
imum
bor
row
er c
ontr
ibut
ion
is 5
per
cent
(w
ith o
r w
ithou
t m
ortg
age
insu
ranc
e).A
min
imum
20
perc
ent
dow
n pa
ymen
t is
req
uire
d w
hen
mor
t-ga
ge in
sura
nce
is n
ot p
lace
d on
a lo
an;h
owev
er,
the
borr
ower
nee
d on
ly m
ake
a 5
perc
ent
cont
ribut
ion
from
his
or
her
own
fund
s.T
he b
alan
ce o
ffu
nds
may
be
in t
he fo
rm o
f a
gift.
gT
he m
inim
um d
own
paym
ent
is 3
per
cent
for
Fann
ie 9
7 an
d A
fford
able
Gol
d 97
;the
min
imum
dow
n pa
ymen
t is
5 p
erce
nt fo
r C
HB
P, C
HB
P w
ith 3
/2O
ptio
n, A
fford
able
Gol
d 3/
2, a
nd A
fford
able
Gol
d 5.
hT
he m
inim
um b
orro
wer
con
trib
utio
n is
3 p
erce
nt fo
r Fa
nnie
97,
CH
BP
with
3/2
Opt
ion,
Affo
rdab
le G
old
97,
and
Affo
rdab
le G
old
3/2.
It is
5 p
erce
nt fo
rC
HB
P a
nd A
fford
able
Gol
d 5.
Whe
re t
he m
inim
um b
orro
wer
con
trib
utio
n is
3 p
erce
nt,
the
addi
tiona
l 2 p
erce
nt c
an c
ome
from
gift
s, g
rant
s, a
nd s
ubsi
-di
zed
seco
nd m
ortg
ages
(co
mm
unity
sec
onds
).i
One
mon
th r
eser
ve is
req
uire
d by
Fan
nie
97,
Affo
rdab
le G
old
97,
and
Affo
rdab
le G
old
3/2.
One
mon
th is
rec
omm
ende
d bu
t no
t re
quire
d fo
r A
fford
able
Gol
d 5.
Res
erve
s ar
e no
t re
quire
d fo
r C
HB
P o
r C
HB
P w
ith 3
/2 O
ptio
n.
Portfolio Affordable Mortgages are loans kept in-house by a lender for its investmentpurposes, as opposed to loans sold to outside investors.17 There is nothing inherent ina portfolio loan that differentiates it from a standard GSE issue. In fact, banks maykeep loans in their portfolio for a while and then sell them on the secondary market.In practice, however, when dealing with less advantaged borrowers, the loans kept inportfolio tend to have somewhat more flexible financial and underwriting terms thanother loans do. These in-house loans are typically kept in portfolio because they ex-ceed the financial and underwriting parameters set by the GSEs.
To provide a perspective on recent advances in mortgage underwriting, we also referto the Historical Mortgage—the standard conventional mortgage that dominated themarket during the two decades before 1990.
The various mortgage typologies and the subsumed loan products differ in their fi-nancial characteristics and in their borrower and property underwriting criteria. Fi-nancial characteristics include such fundamental considerations as the loan’s interestrate, the LTV, and the front-end and back-end ratios. Borrower and property under-writing criteria consider the credit, employment, and income record of the would-bemortgagor; the condition and attractiveness of the property to be mortgaged (as wellas that of the surrounding neighborhood); and the assets acceptable to close the trans-action (table 7). Although all of the above items are commonly referred to as “under-writing” collectively, for the purposes of this discussion (and in tables 6 and 7), thefinancial characteristics and borrower and property criteria will be considered sepa-rately.
The Historical Mortgage required a minimum 10 percent down payment, thus allow-ing a maximum 90 percent LTV. The front-end ratio had a 25 percent to 28 percentrange, and the back-end ratio had a 33 percent to 36 percent range. Additional finan-cial characteristics are summarized in table 6.
The current version of this loan—the GSE Standard Mortgage—is more liberal. Itallows a maximum 95 percent LTV. The maximum guideline ratios18 are 28 percentfor the front-end ratio and 36 percent for the back-end ratio. For those who are ableto afford it, the GSE Standard Mortgage is a reliable vehicle for purchasing or refi-nancing a home.
The mortgage parameters described above exclude many financially disadvantagedhome seekers. The GSE Affordable Mortgages were created to address that gap andto serve LMI borrowers (including minorities and others not well served by the Stan-
30 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
17 The mortgage parameter of what is salable to the secondary market is a moving target. Thus, the GSEs arebeginning to buy portions of affordable loans that we have labeled as portfolio loans. Freddie Mac, for example,has informed the research team that it is purchasing some Bank of America products (e.g., Zero Down, CreditFlex) that we have labeled portfolio loans.
18 The GSE front- and back-end ratios are “guidelines” that a lender uses to qualify a borrower. If the ratios exceedthe “guidelines” of the program, a lender, with compensating factors documented, can make the loan with higherratios.
Model Calibration and Analysis 31
The Potential and Limitations of Mortgage Innovation
Tabl
e 7.
Bo
rro
wer
an
d P
rop
erty
Un
der
wri
tin
g C
rite
ria
by M
ort
gag
e Ty
pe
Und
erw
ritin
gG
SE
Sta
ndar
d M
ortg
age
and
Crit
eria
His
toric
al M
ortg
age
GS
EA
fford
able
Mor
tgag
eP
ortfo
lio A
fford
able
Mor
tgag
e
I.C
red
ith
isto
ry
II.Em
ploy
men
t/in
com
eh
isto
ry
a.S
tron
g fo
rmal
cre
dit
reco
rd is
requ
ired.
b.S
tron
g cr
edit
bure
au r
epor
t is
requ
ired.
c.A
pplic
ant
mus
t ha
ve h
ad n
o le
gal
actio
n (i.
e.,
judg
men
t, ba
nkru
ptcy
, or
fore
clos
ure)
with
in p
ast
7 ye
ars.
a.A
pplic
ant
mus
t do
cum
ent
sam
eoc
cupa
tion
and
empl
oyee
his
tory
for
at le
ast
2 ye
ars.
b.O
vera
ge (
e.g.
, bo
nuse
s) a
nd s
ea-
sona
l inc
ome
incl
uded
onl
y if
con-
sist
ent
over
a 2
-yea
r pe
riod
and
expe
cted
to
cont
inue
at
the
sam
ele
vel o
r in
crea
se in
the
fut
ure.
c.A
ll th
ird-p
arty
–pro
vide
d in
com
ein
clud
ed if
the
re is
a h
isto
ry o
f su
chpa
ymen
ts a
nd if
inco
me
is e
xpec
ted
to c
ontin
ue in
the
fut
ure
(no
men
tion
of s
peci
al t
reat
men
t, e.
g.,
gros
s up
of n
onta
xabl
e in
com
e).
d.Ta
x be
nefit
s of
hom
eow
ners
hip
are
not
incl
uded
as
inco
me.
e.“B
oard
er”
inco
me
is n
ot in
clud
ed(r
enta
l inc
ome
is in
clud
ed).
f.In
add
ition
to
the
prim
ary
appl
ican
t’sin
com
e, o
nly
the
inco
me
of c
omor
t-ga
gors
is in
clud
ed.
a.C
redi
t ca
n be
est
ablis
hed
thro
ugh
alte
rnat
e m
eans
.B
lem
ishe
d cr
edit
is n
ot s
uffic
ient
grou
nds
for
deni
al:
•O
ccas
iona
l lap
ses
acce
pted
•E
xten
uatin
g ci
rcum
stan
ces
acce
pted
b.T
here
is f
lexi
ble
trea
tmen
t of
cre
dit
scor
ing.
c.Le
gal a
ctio
ns a
t lea
st 2
yea
rs o
ld c
anbe
dis
coun
ted
with
ree
stab
lishe
d pa
y-m
ent h
isto
ry a
nd n
o la
te p
aym
ents
.
a.A
pplic
ant
mus
t do
cum
ent
inco
me
(not
nec
essa
rily
job/
empl
oyer
) st
a-bi
lity
for
2 ye
ars.
b.S
imila
r to
His
toric
al M
ortg
age:
over
-tim
e an
d bo
nuse
s ar
e av
erag
edov
er a
2-y
ear
perio
d.
c.A
ll th
ird-p
arty
–pro
vide
d in
com
e is
coun
ted;
nont
axab
le in
com
e is
gros
sed
up.(
Som
etim
es S
ocia
lS
ecur
ity in
com
e al
so is
gro
ssed
up.
)
d.Ta
x be
nefit
s of
hom
eow
ners
hip
are
not
incl
uded
as
inco
me.
e.B
oard
er in
com
e is
incl
uded
as
a co
m-
pens
atin
g fa
ctor
for
high
er q
ualif
ying
ratio
s fo
r G
SE
Affo
rdab
le M
ortg
age
but
not
GS
E S
tand
ard
Mor
tgag
e.f.
In a
dditi
on t
o th
e pr
imar
y ap
plic
ant’s
inco
me,
onl
y th
e in
com
e of
com
ort-
gago
rs is
incl
uded
.
a.T
his
mor
tgag
e ha
s es
sent
ially
the
sam
e pa
ram
eter
s as
GS
E A
fford
able
,w
ith s
omew
hat
grea
ter
flexi
bilit
y.E
xam
ples
:•
Med
ical
act
ions
dis
coun
ted
•G
reat
er r
ecep
tivity
to
exte
nuat
ing
circ
umst
ance
sb.
Cre
dit
scor
ing
may
be
som
ewha
tde
emph
asiz
ed.a
c.Le
gal a
ctio
ns m
ay b
e di
scou
nted
afte
r 1
to 2
yea
rs if
act
ion
isex
plai
ned
(e.g
., ex
tenu
atin
g ci
rcum
-st
ance
s) a
nd c
lean
rep
aym
ent
reco
rd h
as b
een
mai
ntai
ned.
a.A
pplic
ant
mus
t do
cum
ent
inco
me
(not
nec
essa
rily
job/
empl
oyer
) st
a-bi
lity
for
1 to
2 y
ears
.b.
Sim
ilar
to H
isto
rical
Mor
tgag
e:ov
er-
time
and
bonu
ses
are
aver
aged
over
a 2
-yea
r pe
riod.
c.A
ll th
ird-p
arty
–pro
vide
d in
com
e is
coun
ted;
nont
axab
le in
com
e is
gros
sed
up.(
Som
etim
es S
ocia
lS
ecur
ity in
com
e al
so is
gro
ssed
up.
)
d.Ta
x be
nefit
s of
hom
eow
ners
hip
may
be in
clud
ed a
s in
com
e.e.
Boa
rder
inco
me
is in
clud
ed a
s a
com
-pe
nsat
ing
fact
or fo
r hi
gher
qua
lifyi
ngra
tios
for
GS
E A
fford
able
Mor
tgag
ebu
t no
t G
SE
Sta
ndar
d M
ortg
age.
f.T
he in
com
e of
all
adul
t ho
useh
old
mem
bers
(ev
en n
on-c
omor
tgag
ors)
may
be
coun
ted.
32 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 7.
Bo
rro
wer
an
d P
rop
erty
Un
der
wri
tin
g C
rite
ria
by M
ort
gag
e Ty
pe
(con
tinue
d)
Und
erw
ritin
gG
SE
Sta
ndar
d M
ortg
age
and
Crit
eria
His
toric
al M
ortg
age
GS
EA
fford
able
Mor
tgag
eP
ortfo
lio A
fford
able
Mor
tgag
e
II.Em
ploy
men
t/in
com
eh
isto
ry(c
on
tinu
ed)
III.A
sset
veri
fica
tio
n
IV.P
rop
erty
—n
eig
hb
or-
ho
od
stan
dar
ds/
app
rais
al
g.S
tric
t ve
rific
atio
n of
all
inco
me
com
-po
nent
s is
req
uire
d.
a.A
sset
his
tory
mus
t be
ver
ifiab
le(c
ash
on-h
and
not
acce
pted
).
b.B
orro
wer
mus
t pr
ovid
e al
l fun
ds fo
rdo
wn
paym
ent
and
clos
ing
cost
s.
c.O
nly
gifts
fro
m fa
mily
mem
bers
are
allo
wed
;mon
ies
iden
tifie
d as
gift
sm
ust
be v
erifi
ed a
s su
ch (
e.g.
, le
tter
from
fam
ily m
embe
r sp
ecify
ing
that
the
tran
sfer
is a
bsol
ute
and
repa
y-m
ent
is n
ot r
equi
red)
.d.
Col
late
raliz
ed lo
an (
on p
erso
nal
prop
erty
, fo
r in
stan
ce)
is a
ccep
tabl
eas
an
asse
t w
ith s
tric
t ve
rific
atio
n.
a.N
eigh
borh
ood
char
acte
ristic
s co
unt
in a
ppra
isal
(po
sitiv
e an
d ne
gativ
e).
b.P
rope
rty
econ
omic
and
fun
ctio
nal
obso
lesc
ence
neg
ativ
ely
affe
cts
valu
atio
n.
c.“C
omps
”(c
ompa
rabl
e sa
les)
are
cons
ider
ed a
nd a
re r
estr
icte
d by
dist
ance
(fr
om s
ubje
ct p
rope
rty)
and
time
(sal
es w
ithin
the
last
6 m
onth
s).
g.Ve
rific
atio
n of
all
inco
me
com
pone
nts
is r
equi
red;
verb
al c
onfir
mat
ion
may
suffi
ce.b
a.A
sset
his
tory
mus
t be
verif
iabl
e;fo
rG
SE
Affo
rdab
le M
ortg
age,
cas
h on
-ha
nd is
acc
epta
ble
unde
r sp
ecifi
csi
tuat
ions
.b.
Bor
row
er is
not
req
uire
d to
pro
vide
all f
unds
for
dow
n pa
ymen
t and
clo
s-in
g co
sts;
a po
rtio
n m
ay b
e pr
ovid
edby
out
side
sou
rces
.
c.G
ifts
from
fam
ily m
embe
rs a
re a
llow
-ed
(af
ter
the
borr
ower
con
trib
utio
n is
satis
fied)
;pub
lic g
rant
s ar
e ac
cept
-ab
le b
ut tr
ansf
ers
from
the
selle
r an
dth
e re
alto
r ar
e re
stric
ted.
d.C
olla
tera
lized
loan
s ar
e m
ore
read
ilyac
cept
able
.
a.P
ositi
ve a
spec
ts o
f a n
eigh
borh
ood
(e.g
., re
cent
reh
abili
tatio
n or
new
cons
truc
tion)
are
str
esse
d.b.
Obs
oles
cenc
e di
scou
nt is
dee
mph
a-si
zed
by r
ecog
nitio
n of
a w
ider
ran
geof
acc
epta
ble
stan
dard
s.
c.G
reat
er r
ange
of c
omps
is a
ccep
ted.
g.T
he r
equi
rem
ents
are
ess
entia
lly t
hesa
me
as t
hose
for
GS
E A
fford
able
Mor
tgag
e, w
ith m
inor
exc
eptio
ns.
a.A
sset
s no
t al
way
s ve
rifie
d an
d ca
shon
-han
d is
fre
quen
tly a
ccep
ted.
b.R
equi
rem
ents
are
gen
eral
ly t
hesa
me
as t
hose
for
GS
E A
fford
able
,bu
t P
ortfo
lio A
fford
able
may
allo
wm
ost
or a
ll of
the
dow
n pa
ymen
t to
com
e fr
om o
utsi
de s
ourc
es.
c.P
ortfo
lio A
fford
able
allo
ws
mor
e gi
ftso
urce
s th
an G
SE
Affo
rdab
le a
llow
s(e
.g.,
tran
sfer
s fr
om t
he s
elle
r or
the
real
tor
are
perm
issi
ble)
.
d.C
olla
tera
lized
loan
s ar
e m
ore
read
ilyac
cept
able
.
a.P
ositi
ve a
spec
ts o
f a n
eigh
borh
ood
(e.g
., re
cent
reh
abili
tatio
n or
new
cons
truc
tion)
are
str
esse
d.b.
Sam
e as
GS
E A
fford
able
:ofte
n th
ego
od c
ondi
tion
of a
pro
pert
y is
stre
ssed
and
a h
ome
insp
ectio
n is
requ
ired
to a
void
a s
ituat
ion
whe
re a
prop
erty
nee
ds m
ajor
rep
airs
ear
lyon
.c.
Gre
ater
ran
ge o
f com
ps is
acc
epte
d.
Model Calibration and Analysis 33
The Potential and Limitations of Mortgage Innovation
Tabl
e 7.
Bo
rrow
er a
nd
Pro
per
ty U
nd
erw
riti
ng
Cri
teri
a by
Mo
rtg
age
Typ
e (c
ontin
ued)
Und
erw
ritin
gG
SE
Sta
ndar
d M
ortg
age
and
Crit
eria
His
toric
al M
ortg
age
GS
EA
fford
able
Mor
tgag
eP
ortfo
lio A
fford
able
Mor
tgag
e
aT
hat
is n
ot t
he c
ase
in t
he t
wo
Ban
k of
Am
eric
a m
ortg
ages
con
side
red
here
.b
Alte
rnat
ive
docu
men
tatio
n of
inco
me
is a
ccep
tabl
e;th
e bo
rrow
er m
ay p
rovi
de o
rigin
al p
ayst
ub(s
) an
d In
tern
al R
even
ue S
ervi
ce o
r W
-2 fo
rms
for
the
prev
ious
tw
o ye
ars,
and
the
lend
er m
ay v
erify
cur
rent
em
ploy
men
t by
tel
epho
ne.
V.O
ther
a.C
ontin
gent
liab
ility
deb
t (e
.g.,
cosi
gned
loan
) is
incl
uded
in t
otal
debt
(ba
ck-e
nd)
ratio
.
b.P
oole
d as
set f
unds
(i.e
., th
ose
draw
nfr
om fo
rmal
/info
rmal
alte
rnat
ive
sav-
ings
acc
ount
s, s
uch
as s
ous-
sous
)m
ay n
ot b
e us
ed a
s as
sets
, ev
en if
the
fund
s dr
awn
are
not
repa
yabl
e.
a.G
SE
gui
delin
es r
equi
re th
e le
nder
tove
rify
a hi
stor
y of
doc
umen
ted
pay-
men
ts o
n a
debt
by
the
prim
ary
oblig
or a
nd a
scer
tain
that
a h
isto
ryof
del
inqu
ent p
aym
ents
doe
s no
tex
ist f
or th
at d
ebt.
Whe
n th
is c
anno
tbe
don
e, th
e le
nder
incl
udes
the
cont
inge
nt li
abili
ty a
s lo
ng-t
erm
deb
tw
hen
calc
ulat
ing
the
borr
ower
’squ
alify
ing
ratio
s.b.
Poo
led
asse
t fun
ds m
ay b
e us
ed fo
rbo
th G
SE
Sta
ndar
d an
d A
fford
able
Mor
tgag
es a
nd a
re s
ubje
ct to
gui
de-
lines
est
ablis
hed
for
cash
on-
hand
.
a.S
ame
as G
SE
Affo
rdab
le:c
ontin
gent
liabi
lity
debt
may
be
excl
uded
fro
mth
e ba
ck-e
nd r
atio
if t
he p
rimar
y bo
rrow
er h
as a
con
sist
ent
paym
ent
hist
ory.
b.P
oole
d as
set
fund
s ar
e m
ore
read
ilyac
cept
able
.
34 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
dard Mortgage). Several variations of this type are covered in tables D.1 and D.2 inappendix D (e.g., Fannie Mae’s Fannie 97 and Freddie Mac’s Affordable Gold 5™);their general characteristics are summarized in table 6.
The focus of these affordable loans is to help less advantaged people achieve home-ownership; accordingly, the GSE Affordable Mortgages are sometimes limited to homepurchases,19 and the loans are typically targeted to people earning no more than thearea median income. The GSE Standard Mortgage has no such income limitation—one just needs to be able to afford it. The GSE Affordable Mortgage is not subsidizedwith a below-market interest rate; a market rate is charged. However, the standardfinancial requirements are significantly modified. Because LMI mortgagors have verymodest assets, only a minimum down payment is required—3 percent to 5 percent ofthe purchase price. Not all of the down payment has to come from the borrower; 2percent of the 5 percent is sometimes allowed to come from gifts or grants. The lowdown payments are mirrored by very high LTVs (as high as 97 percent). Becausethese ratios are so high, private mortgage insurance (PMI) is always required. More-over, since closing costs can also be a hurdle to homeownership, GSE Affordable Mort-gages allow payment of these costs to come from sources other than the borrower (forexample, a grant or a subsidized loan from a nonprofit group).20 Because the mort-gagor’s income is constrained, higher front-end and back-end ratios of 33 percent and40 percent, respectively, are allowed.21
Some lenders offer terms that are even more liberal than those offered by the GSEs.These terms are listed under the Portfolio Affordable Mortgage column in table 6. Toillustrate, both of the more liberal GSE Affordable Mortgages—Fannie Mae’s Fannie97 and Freddie Mac’s AG 97—require a 3 percent minimum borrower contribution.In contrast, one portfolio mortgage, Bank of America’s Neighborhood Advantage ZeroDown™, requires no down payment. Portfolio Affordable Mortgage allowances wereintroduced in recognition of the minimal assets available to many traditionally un-derserved families for the down payment and closing costs. The resources of thesefamilies are so insubstantial that even the greatly reduced GSE Affordable Mortgagerequirements (e.g., a 3 percent down payment) can prove too great an obstacle. To en-hance affordability, some portfolio mortgages have rescinded PMI requirements, evenon high LTV loans. For example, this was the policy of some lenders participating inthe Delaware Valley Mortgage Plan (DVMP). (We refer to no-PMI portfolio productsreflective of the DVMP experience as a Portfolio Composite.)
A further level of affordability is offered by underwriting flexibility. Historically, manyunderwriting standards concerning credit history, employment history, asset verifica-tion, and so on were quite strict. These standards are detailed in table 7 under His-
19 This is the case for Fannie Mae’s Fannie 97 and Freddie Mac’s AG 97. Fannie 97 is also available for no-cashout rate/term refinances of existing Fannie 97 mortgages.
20 This depends on the specific GSE Affordable Mortgage, as detailed in tables 6 and 7.
21 Freddie Mac’s GSE Affordable Mortgages have no maximum front-end ratio.
Model Calibration and Analysis 35
The Potential and Limitations of Mortgage Innovation
torical Mortgage. Since 1990, however, underwriting requirements have been mademore flexible for both the GSE Standard Mortgages and the GSE Affordable Mort-gages. In fact, they currently have nearly indistinguishable underwriting standards.22
Still, the portfolio loans of some institutions venture even further, incorporating ad-ditional provisions for mortgage-qualification flexibility.
The changes in underwriting are summarized in table 7 under five sections: (1) cred-it history; (2) employment/income history; (3) asset verification; (4) property—neigh-borhood standards/appraisal; and (5) other. For example, the credit bar for the His-torical Mortgage was high. Institutions typically demanded a strong credit bureaureport. Credit misdeeds were treated harshly. It was difficult to obtain a mortgage ifa bankruptcy, foreclosure, or other legal action had taken place within the past sevenyears.
The current GSE Standard and GSE Affordable Mortgages have changed the creditworld, however. Recognizing that some people may not have obtained formal credit inthe past, the GSE Standard and GSE Affordable Mortgages allow credit to be estab-lished through alternative means. Also realizing that credit misdeeds may occur, es-pecially among lower-income families and those new to using credit, the GSE Stan-dard and GSE Affordable Mortgages are more tolerant of occasional credit lapses, aslong as a pattern is not indicated and the lapses are due to extenuating circumstances(e.g., illness, unemployment linked to downsizing, a spouse’s death).
Credit underwriting for portfolio mortgages is a “nuanced change” from GSE policy.It may mean greater tolerance for payment lapses common among the less advan-taged, such as medical collection or missed winter utility bills. Portfolio lending mayalso pay less heed to credit scoring. The GSEs encourage lenders to use credit scoringas one of the numerous components of risk assessment, and they mention Fair, Isaac& Company (FICO) specifically (Fannie Mae 1995, 2).23 The GSEs have referred to abroad parameter or guideline FICO score of approximately 620 for their standardproducts. At the same time, the GSEs encourage lenders to factor the FICO scorealong with compensating factors (e.g., a lower mortgage LTV) and extenuating cir-cumstances (Fannie Mae 1995, 1997; Freddie Mac 1995, 1996). The portfolio lenderssometimes accept applicants with scores lower than the comfort range of the GSEs(e.g., the Portfolio Composite accepted FICO scores in the mid-500s). Each portfolioprogram, however, has its own requirements. For example, the Bank of AmericaNeighborhood Advantage Zero Down mortgage grants a 100 percent LTV mortgagebut has a guideline FICO score of 660, whereas a sister Bank of America Neighbor-hood Advantage Credit Flex portfolio product has somewhat lower FICO guidelinesbut also a lower LTV.
22 There are very few technical differentiations. For instance, boarder income is included as a compensating factorfor higher qualifying ratios for GSE Affordable Mortgages but not for GSE Standard Mortgages.
23 After the research for this report was completed, Fannie Mae announced that it was replacing credit scores withspecific credit characteristics in the most recent release of its automated underwriting system (Fannie Mae 2000).
36 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Current underwriting with respect to employment, income, and assets has been sig-nificantly liberalized compared with the Historical Mortgage. This is true for the cur-rent GSE Standard Mortgages, the GSE Affordable Mortgages, and the Portfolio Af-fordable Mortgages (see table 7 for a comparison of the underwriting criteria for eachmortgage type).
It is important to note that elements of each mortgage product are subject to change.For example, the Emerging GSE Affordable category shares numerous characteristicswith the Portfolio Affordable category (e.g., very high front-end and back-end ratios).The Government Affordable Mortgage also has evolved. Loans guaranteed by theU.S. Department of Veterans Affairs, important in the earlier post–World War II era,are no longer significant today. The primary unsubsidized Government AffordableMortgage loans insured under the FHA 203(b) program have also been changing. Be-fore 1997, FHA 203(b) mortgages, termed FHA 203(b)–prior, had a maximum LTV ofabout 96 percent24 on purchase amounts up to $125,000. Since 1997, the FHA 203(b)mortgage, termed FHA 203(b)–current, has increased the LTV to about 98 percent25
on purchase amounts up to $125,000. Although the FHA 203(b)–prior allowed non-recurring closing costs to be financed, that option ended with the FHA 203(b)–cur-rent. Both forms of the FHA 203(b) mortgage have been relatively lenient on creditunderwriting.
To summarize, our mortgage framework comprises the types and products listed intable 8.
Mortgage Model Data
The next step in the Mortgage Model is to obtain appropriate data so that the finan-cial and underwriting facets of the alternative mortgages can be applied to the renterfamilies. While data are not available for every component cited in the preceding sec-tion, much information can be obtained from various sources, and we can calibratethe more critical financial and underwriting elements. This process is described indetail in appendices B and D; we summarize the major points below.
Some of the mortgage data elements are not family specific. These include the mort-gage interest rate, property taxes, property insurance, mortgage insurance, and clos-ing costs, all of which are determined in as specific a fashion as possible, as noted inappendix D.
24 The exact maximum LTV was 97 percent of the first $25,000; 95 percent of the amount between $25,001 and$125,000; and 90 percent of the amount over $125,000.
25 The exact maximum LTV is 98.75 percent on the first $50,000; 97.65 percent of the amount between $50,001and $125,000; and 97.15 percent of the amount over $125,000.
For instance, it is important to carefully specify closing costs, as these are quite signif-icant, often exceeding the amount of the down payment in this era of very high LTVloans. Closing cost information for eight major metropolitan areas in different regionsof the country were obtained from Countrywide Home Loans (1998), one of the nation’slargest lenders. From these actual transactions, we derived closing costs as a percent-age of the housing unit price. These costs varied from 3 percent to 4 percent of thehousing unit price in the Atlanta and Los Angeles metropolitan areas to 5 percent to6 percent in the New York and Philadelphia metropolitan areas. Closing costs weredifferent for several reasons. Certain areas—for example, New York and Philadel-phia—had very high mortgage taxes, while other locations had no mortgage taxes ormuch lower mortgage levies. Prepaid property taxes also varied; these payments werelow in the Atlanta and Los Angeles metropolitan areas and high in the Northeastlocations. Interestingly, there was variation in the price of items that one would havethought would be uniformly priced commodity items. For instance, title insuranceamounted to 0.6 percent of the home’s value in Miami, compared with 0.2 percent inAtlanta. There were also specific area idiosyncrasies. For example, the cost of hazardinsurance was three times higher in Miami than in the other metropolitan areas—aresult of the tremendous insurance company losses suffered in the wake of HurricaneAndrew.
Model Calibration and Analysis 37
The Potential and Limitations of Mortgage Innovation
Table 8. Mortgage Types and Products in Report Tables
Mortgage Types/Products Detailed in Tables
Historical Mortgage 6, 7, D.6
GSE Standard Mortgage D.1, D.2, D.6(Fannie Mae and Freddie Mac)
GSE Affordable MortgagesFannie Mae
CHBP D.1, D.6CHBP, 3/2 Option D.1, D.6Fannie 97 D.1, D.6
Freddie MacAG D.2, D.6AG, 3/2 Option D.2, D.6AG 97 D.2, D.6
Emerging GSE Affordable MortgagesFannie Mae
Flex 97 D.3, D.6Freddie Mac
Community Gold D.3, D.6
Portfolio Affordable MortgagesBank of America Zero Down D.4, D.6Bank of America Credit Flex D.4, D.6Portfolio Composite D.6
Government Affordable MortgagesFHA 203(b)–prior D.5, D.6FHA 203(b)–current D.5, D.6
The Mortgage Model also calibrates family characteristics that influence mortgagequalification: (1) income, (2) debt, (3) assets, (4) credit record, and (5) other items (e.g.,employment stability). There are various sources of data for these items, as explainedin appendix C. Our choice of which source to use is governed by the demands of thecurrent investigation. Since our objective is to determine the degree of housing af-fordability ensuing from the alternative mortgages, the data sought must correlatewith each of the mortgage’s financial and underwriting characteristics. For instance,the GSE Standard, GSE Affordable, and Portfolio Affordable Mortgages “gross up”nontaxable sources of income and count boarder income; the Historical Mortgagedoes not.
These specific mortgage-driven data needs are then compared with the range of po-tential available data sources on families and households, and an appropriate sourceis selected (table 9). For example, income information found in the SIPP is the richestof all the sources. The SIPP disaggregates income into 38 categories, including “room-er and boarder” income as well as nontaxable sources (e.g., child support, disability),which allows for a more accurate mortgage qualification. The SIPP is also the pre-ferred source for asset, debt, and other characteristics, as explained in appendix C.Furthermore, the SIPP has the advantage of consistency because the ConsumptionModel is based on the SIPP.
Appendices C and D describe in detail how each SIPP variable is treated to developthe data for the Mortgage Model. For instance, debt is reported in the SIPP on a totalbasis, and this has to be translated into a periodic debt-repayment amount for under-writing the back-end ratio. Similarly, as the alternative mortgage products treat cat-egories of income differently, we adjust the various income components in the SIPPaccordingly. Thus, Social Security is not grossed up under the Historical Mortgage,but it is under today’s more innovative mortgage products. The Historical Mortgagewould discount public support, but today’s mortgages fully count that source. Further,the Historical Mortgage would penalize the borrower with occupational instability,income instability, or both, in effect discounting income in these situations. Currentmortgages are much more lenient in this regard. We therefore adjust income to re-flect these mortgage traits, as detailed in appendix D.
The results of these adjustments provide the asset, debt, and income statistics reportedin table 10 for all renters and for renter subgroups (non-Hispanic white; non-Hispanicblack; Hispanic; non-Hispanic other; and recent immigrant).
A number of characteristics are significant. Renter income is modest, especially forthe traditionally underserved, as noted in table 11 based on 1993 SIPP data.
The modest renter incomes above would not always be fully credited by underwriters.Thus, since the Historical Mortgage would discount or exclude such income compo-nents (reported in the SIPP) as Aid to Families with Dependent Children, child sup-port, and alimony, which are drawn upon by some renters, the median mortgage-available family income (or adjusted family income) for all renters and for white, black,and Hispanic renters under the Historical Mortgage would be $15,957, $17,879,
38 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Model Calibration and Analysis 39
The Potential and Limitations of Mortgage Innovation
Tabl
e 9.
Su
mm
ary
An
alys
is o
f D
ata
So
urc
es C
on
sid
ered
fo
r th
e H
ou
sin
g A
ffo
rdab
ility
an
d M
ort
gag
e Q
ual
ific
atio
n R
esea
rch
Mos
tM
ortg
age
Und
erw
ritin
g C
riter
ia
Sou
rce
Rec
ent Y
ears
Inco
me
Ass
ets
Liab
ilitie
sC
redi
tE
mpl
oym
ent
Sou
rce:
See
app
endi
x C
(ta
bles
C.1
, C
.2,
and
C.3
).N
ote:
Sha
ded
boxe
s ar
e so
urce
s ch
osen
in t
he a
naly
sis.
Co
nsu
mer
Exp
end
itu
reS
urv
ey
Su
rvey
of
Inco
me
and
Pro
gra
m
Par
tici
pat
ion
Pan
el S
tud
y o
fIn
com
eD
ynam
ics
(PS
ID)
Su
rvey
of
Co
nsu
mer
Fin
ance
Nat
ion
al
Su
rvey
of
Fam
ilies
an
dH
ou
seh
old
s
Pu
blic
Use
Mic
rod
ata
Sam
ple
Am
eric
anH
ou
sin
g
Su
rvey
1990
–95
1991
–93,
199
6
1990
–96
1991
, 19
93,
and
1995
1992
–94
(wav
e2
data
)
1990
Nat
iona
l sam
ple
is 1
999
(oth
ers
vary
by
MS
A)
Non
e
Non
e
•Q
uest
ions
on
fisca
l str
ess
byye
ar•
Ava
ilabl
e in
onl
y19
96 s
urve
yN
one
•O
rdin
al s
cale
•
Que
stio
n on
over
due
pay-
men
ts:t
otal
inpa
ymen
ts t
hat
wer
e 2
mon
ths
over
due
Non
e
Non
e
•In
terv
al s
cale
•
11 s
ourc
es•
Bef
ore
and
afte
rta
xes
•In
terv
al s
cale
•38
sou
rces
•B
y in
divi
dual
s in
hou
seho
ld
•In
terv
al s
cale
•
8 so
urce
s•
By
indi
vidu
als
inho
useh
old
•O
rdin
al s
cale
•5
cate
gorie
s•
Ove
rsam
ples
wea
lthy
hous
e-ho
lds
•O
rdin
al s
cale
•7
sour
ces
•In
terv
al s
cale
•8
sour
ces
•O
rdin
al s
cale
•9
sour
ces
•In
terv
al s
cale
•7
sour
ces
•A
ggre
gate
est
i-m
ates
not
rob
ust
•In
terv
al s
cale
•
15 f
inan
cial
ass
etca
tego
ries
•5
nonf
inan
cial
asse
t ca
tego
ries
•A
ggre
gate
ass
ets
avai
labl
e •
PS
ID c
onta
ins
asu
pple
men
t w
itha
few
inte
rval
estim
ates
of
asse
ts•
Ord
inal
sca
le•
10 c
ateg
orie
s fo
rfin
anci
al a
sset
s;5
for
nonf
inan
cial
asse
ts•
Ord
inal
sca
le
•5
sour
ces
Non
e
Non
e
•In
terv
al s
cale
•M
onth
ly p
ay-
men
ts o
n ou
t-st
andi
ng d
ebt
•In
terv
al s
cale
•
Ans
wer
s on
3ca
tego
ries
in19
93 S
urve
y(w
ave
7)
•P
SID
con
tain
s a
supp
lem
ent
with
a fe
w in
terv
ales
timat
es o
f lia
bilit
ies
•O
rdin
al s
cale
•6
cate
gorie
s
•O
rdin
al s
cale
•M
onth
ly p
ay-
men
ts a
nd t
otal
outs
tand
ing
•5
sour
ces
Non
e
Non
e
•12
mon
ths
•A
vera
ge h
ours
wor
ked
•48
+ m
onth
s•
Ave
rage
hou
rsw
orke
d•
Occ
upat
ion
and
indu
stry
•S
elf-
empl
oym
ent
•Le
ngth
of
time
with
cur
rent
empl
oyer
•C
urre
nt e
mpl
oy-
men
t st
atus
•E
mpl
oym
ent
hist
ory
1987
–94
•O
ccup
atio
n
•12
mon
ths
•A
vera
ge h
ours
wor
ked
Non
e
40 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 10
.Fin
anci
al a
nd
Dem
og
rap
hic
Pro
file
s o
f R
ente
r G
rou
ps
(in
Do
llars
,exc
ept
N R
ow
s an
d H
ou
seh
old
Siz
e,Fa
mily
Siz
e,an
d F
amily
Wei
gh
t C
olu
mn
s)
Nom
inal
Adj
uste
dA
djus
ted
Nom
inal
Adj
uste
dA
djus
ted
Targ
etH
ouse
hold
Hou
seho
ldH
ouse
hold
Fam
ilyFa
mily
Fam
ilyH
ouse
hold
Fam
ilyFa
mily
Ren
ter
Gro
upa
Ass
ets
Deb
tH
ouse
Inco
me
Inco
meb
Inco
mec
Inco
me
Inco
meb
Inco
mec
Siz
eS
ize
Wei
ght
All M
ean
7,90
4 3,
609
92,7
81
27,3
98
23,1
78
26,2
00
24,1
17
19,9
33
22,9
28
2.50
2.24
6,95
9 M
edia
n30
0 45
85
,210
21
,600
18
,580
20
,857
18
,786
15
,957
18
,000
2.
002.
005,
933
Std
.Dev
iatio
nd38
,900
8,
250
41,2
88
22,6
65
24,9
02
22,7
65
20,6
31
23,1
00
20,7
57
1.61
1.59
2,08
0 M
inim
um
0 0
7,52
9 0
–123
,764
–2
1,11
2 0
–123
,764
–2
1,11
2 1.
001.
0077
2 M
axim
um71
2,87
7 21
1,50
0 32
7,59
6 20
8,09
5 19
2,11
1 19
2,11
1 20
8,09
5 19
2,11
1 19
2,11
1 13
.00
13.0
027
,728
N
(w
eigh
ted)
25,8
27,5
2625
,827
,526
25,8
27,5
2625
,827
,526
25
,827
,526
25
,827
,526
25
,827
,526
25,8
27,5
2625
,827
,526
25,8
27,5
2625
,827
,526
25,8
27,5
26N
(un
wei
ghte
d)4,
120
4,12
0 4,
109
4,12
0 4,
120
4,12
0 4,
120
4,12
0 4,
120
4,12
0 4,
120
4,12
0
Whi
te (
non-
His
pani
c)M
ean
11,3
68
4,24
7 96
,649
30
,196
25
,912
29
,107
26
,243
21
,987
25
,153
2.
181.
936,
358
Med
ian
649
500
87,7
99
24,2
40
21,1
98
23,3
62
20,5
56
17,8
79
19,9
92
2.00
1.00
5,77
5 S
td.D
evia
tiond
45,5
65
9,06
5 42
,906
23
,955
26
,102
23
,872
21
,489
23
,878
21
,442
1.
411.
391,
703
Min
imum
0
0 8,
286
0 –1
16,6
47
–20,
449
0 –1
16,6
47
–20,
449
1.00
1.00
772
Max
imum
712,
877
211,
500
327,
596
208,
095
192,
111
192,
111
208,
095
192,
111
192,
111
13.0
013
.00
27,7
28
N (
wei
ghte
d)16
,235
,457
16,2
35,4
5716
,183
,879
16,2
35,4
5716
,235
,457
16
,235
,457
16
,235
,457
16,2
35,4
5716
,235
,457
16,2
35,4
5716
,235
,457
16,2
35,4
57N
(un
wei
ghte
d)2,
769
2,76
9 2,
760
2,76
9 2,
769
2,76
9 2,
769
2,76
9 2,
769
2,76
9 2,
769
2,76
9
Bla
ck (
non-
His
pani
c)M
ean
1,60
1 2,
443
83,4
83
20,9
17
17,0
48
19,5
62
19,4
71
15,6
66
18,1
54
2.69
2.50
8,44
7 M
edia
n0
0 73
,205
15
,402
12
,342
14
,319
13
,770
11
,220
12
,993
2.
002.
006,
824
Std
.Dev
iatio
nd9,
165
5,95
5 36
,600
18
,533
20
,356
18
,742
18
,149
20
,083
18
,354
1.
691.
662,
850
Min
imum
0
0 8,
704
0 –7
5,08
0 –1
5,01
1 0
–75,
080
–15,
011
1.00
1.00
2,43
1 M
axim
um11
4,99
9 76
,000
24
5,29
8 15
7,80
6 15
7,80
6 15
7,80
6 15
7,80
6 15
7,80
6 15
7,80
6 12
.00
12.0
022
,863
N
(w
eigh
ted)
4,47
1,07
0 4,
471,
070
4,46
4,52
5 4,
471,
070
4,47
1,07
0 4,
471,
070
4,47
1,07
0 4,
471,
070
4,47
1,07
0 4,
471,
070
4,47
1,07
0 4,
471,
070
N (
unw
eigh
ted)
609
609
608
609
609
609
609
609
609
609
609
609
His
pani
cM
ean
2,00
0 2,
308
86,8
00
23,0
26
18,7
25
21,6
77
20,5
46
16,2
65
19,2
01
3.38
3.06
7,75
8 M
edia
n0
0 82
,812
18
,266
15
,010
17
,283
15
,314
12
,844
14
,178
3.
003.
006,
745
Std
.Dev
iatio
nd23
,148
5,
613
33,6
17
17,4
24
21,4
99
18,3
13
16,8
19
21,0
81
17,7
04
1.89
1.88
2,17
3 M
inim
um
0 0
8,48
5 0
–123
,764
–2
1,11
2 0
–123
,764
–2
1,11
2 1.
001.
001,
995
Max
imum
524,
999
59,5
00
221,
505
119,
196
119,
196
119,
196
119,
196
119,
196
119,
196
11.0
011
.00
22,9
42
N (
wei
ghte
d)3,
986,
596
3,98
6,59
6 3,
980,
131
3,98
6,59
6 3,
986,
596
3,98
6,59
6 3,
986,
596
3,98
6,59
6 3,
986,
596
3,98
6,59
6 3,
986,
596
3,98
6,59
6 N
(un
wei
ghte
d)56
2 56
2 56
1 56
2 56
2 56
2 56
2 56
2 56
2 56
2 56
2 56
2
Oth
er r
ace
(non
-His
pani
c)M
ean
3,92
7 3,
647
95,1
68
28,2
54
23,8
49
26,6
51
24,5
38
20,2
43
22,9
89
3.25
2.88
6,88
2 M
edia
n28
2 0
89,5
48
22,6
55
19,0
94
22,2
66
17,5
89
14,1
65
16,4
25
3.00
3.00
5,94
2 S
td.D
evia
tiond
17,4
93
7,60
2 44
,494
21
,650
23
,124
22
,093
21
,087
22
,531
21
,445
1.
751.
781,
917
Min
imum
0
0 7,
529
180
–35,
845
–4,8
72
0 –3
5,84
5 –4
,872
1.
001.
002,
941
Max
imum
214,
000
59,0
00
237,
383
100,
296
100,
152
100,
152
100,
296
100,
152
100,
152
10.0
010
.00
23,5
97
N (
wei
ghte
d)1,
134,
403
1,13
4,40
3 1,
134,
403
1,13
4,40
3 1,
134,
403
1,13
4,40
3 1,
134,
403
1,13
4,40
3 1,
134,
403
1,13
4,40
3 1,
134,
403
1,13
4,40
3 N
(un
wei
ghte
d)18
0 18
0 18
0 18
0 18
0 18
0 18
0 18
0 18
0 18
0 18
0 18
0
Model Calibration and Analysis 41
The Potential and Limitations of Mortgage Innovation
Tabl
e 10
.Fin
anci
al a
nd
Dem
og
rap
hic
Pro
file
s o
f R
ente
r G
rou
ps
(in
Do
llars
,exc
ept
N R
ows
and
Ho
use
ho
ld S
ize,
Fam
ily S
ize,
and
Fam
ily W
eig
ht
Co
lum
ns)
(con
tinue
d)
Nom
inal
Adj
uste
dA
djus
ted
Nom
inal
Adj
uste
dA
djus
ted
Targ
etH
ouse
hold
Hou
seho
ldH
ouse
hold
Fam
ilyFa
mily
Fam
ilyH
ouse
hold
Fam
ilyFa
mily
Ren
ter
Gro
upa
Ass
ets
Deb
tH
ouse
Inco
me
Inco
meb
Inco
mec
Inco
me
Inco
meb
Inco
mec
Siz
eS
ize
Wei
ght
Rec
ent i
mm
igra
ntM
ean
9,07
6 2,
696
86,0
15
26,7
57
22,7
21
25,3
38
21,8
36
17,8
00
20,4
18
2.48
2.
18
6,95
9 M
edia
n25
0 31
4 77
,938
21
,843
20
,739
22
,794
16
,710
14
,690
16
,200
2.
00
1.00
6,
075
Std
.Dev
iatio
nd52
,306
4,
142
37,2
75
20,9
03
20,8
65
20,0
28
17,6
60
18,4
43
16,6
55
1.54
1.
62
1,98
6 M
inim
um
0 0
8,79
4 0
–21,
716
–4,3
86
0 –3
8,60
2 –4
,386
1.
00
1.00
2,
296
Max
imum
514,
100
17,0
00
212,
301
116,
826
102,
297
113,
872
115,
845
93,7
56
93,7
56
7.00
7.
00
14,1
93
N (
wei
ghte
d)70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 70
4,03
5 N
(unw
eigh
ted)
111
111
111
111
111
111
111
111
111
111
111
111
Sou
rce:
All
stat
istic
s ar
e co
mpu
ted
usin
g th
e w
ave
7 da
ta o
f S
IPP
199
3.a
Non
-His
pani
c w
hite
, no
n-H
ispa
nic
blac
k, H
ispa
nic,
and
oth
er n
on-H
ispa
nic
rent
er fa
mili
es a
re d
iscr
ete.
Com
bine
d, t
hey
sum
to
all r
ente
r fa
mili
es.R
ecen
tim
mig
rant
fam
ilies
, th
ose
ente
ring
the
Uni
ted
Sta
tes
afte
r 19
84,
over
lap
with
the
pre
viou
sly
liste
d ra
cial
/eth
nic
grou
ps (
e.g.
, no
n-H
ispa
nic
whi
tes
and
blac
ks).
Fin
ally
, if
not
othe
rwis
e no
ted,
whi
tes,
bla
cks,
and
oth
er g
roup
s ar
e no
n-H
ispa
nic.
bA
djus
ted
for
stab
ility
and
oth
er c
hara
cter
istic
s ac
cord
ing
to t
he H
isto
rical
Mor
tgag
e st
anda
rds
(see
app
endi
x D
).c
Adj
uste
d fo
r st
abili
ty a
ccor
ding
to
curr
ent
mor
tgag
e in
stru
men
ts’s
tand
ards
(se
e ap
pend
ix D
).d
Sta
ndar
d de
viat
ion—
unw
eigh
ted
valu
es.
$11,220, and $12,844, respectively. These mortgage-available incomes are approxi-mately 20 percent less than the gross (unadjusted, or what we term nominal) renterincomes (table 10).
Renters are not heavily indebted. The median debt of all renters and white, black, andHispanic renters, as derived from the SIPP, is $45, $500, $0, and $0, respectively.Assets, however, are of a trace magnitude. The median assets of all renters and white,black, and Hispanic renters are $300, $649, $0, and $0, respectively.
Also of note in table 10 is the financial profile of recent immigrants. With medianassets, debt, and nominal family income of $250, $314, and $16,710, respectively (meanvalues of $9,076, $2,696, and $21,836, respectively), recent immigrant renters areconsiderably more advantaged than their black and Hispanic counterparts. Recentimmigrant assets, debt, and income, however, approach but fall short of the respectiveresources available to white renters.
In short, the SIPP allows calibration of assets, debt, and income—critical input intothe Mortgage Model. That leaves credit as a final key mortgage underwriting element.Unfortunately, the publicly available sources that contain data on renters have eitherno information on credit (that is true of the SIPP), or they have such rough data (e.g.,the Panel Study of Income Dynamics reports whether a household has experienced“financial distress”) that the information is not very useful (see appendix C). More-over, we specifically seek credit data that are actually used by underwriters; increas-ingly, that includes FICO scores. Appendix B explains how FICO scores were imputedto the renters in the SIPP.
After assembling or estimating renter financial characteristics and factoring thesedata according to the standards used by the alternative mortgage products, the Mort-gage Model calculates the maximum-priced affordable home. This is done for each ofthe renter families in the SIPP. Because of the number of families we are attemptingto qualify and the 15 individual mortgage products considered, the Mortgage Modeluses a software program to facilitate the calculations. The mortgage simulation soft-ware is described in appendix D.
To summarize, the Affordability Analysis encompasses two measures of housingaffordability:
42 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 11. Median Family Income (1995)
All FamiliesRenter (Renters and
Families ($) Homeowners) ($)
All 18,786 28,710Non-Hispanic whites 20,556 30,600Non-Hispanic blacks 13,770 19,515Hispanics 15,314 19,791
Note: The figures shown are the unadjusted family incomes. See table 10 forfurther details.
1. Reference home-buying capacity is the ability of current renters to afford eitherthe target-priced home (derived from the all-renter consumption model) or thecriterion home (median priced, modestly priced [25th percentile], and low priced[10th percentile]).
2. Absolute home-buying capacity is the dollar purchasing power of renters (mea-sured in the aggregate and by considering mean/median purchasing ability).
We anticipate significant variations in the home-buying capacities for our typologyof mortgage products.
• Historical Mortgage. With comparatively rigid down payment minimums and debtratios and other stringent financial and underwriting requirements, this type ofmortgage is expected to yield the lowest home-buying capacity.
• GSE Standard Mortgage. We expect that the GSE Standard Mortgage’s higherLTVs and front-end and back-end debt ratios (compared with those of the Histor-ical Mortgage) will result in a corresponding increase in home-buying capacity.Given the similar financial characteristics of Fannie Mae and Freddie Mac GSEStandard products, we expect no significant differences in the performance ofthese products.
• GSE Affordable Mortgages and Emerging GSE Affordable Mortgages. The relaxedfinancial characteristics of this group (e.g., high debt ratios, LTVs approaching 100percent, or both) should increase affordability in comparison to the GSE StandardMortgages. It is also reasonable to expect greater variation in affordability amongthese products because they incorporate varying debt, LTV, and other criticalparameters.
• Portfolio Affordable Mortgages. We have selected portfolio products that are espe-cially aggressive and innovative; therefore, we expect these instruments to achievea particularly high level of home-buying capacity.
• Governmental Affordable Mortgages. Because the FHA’s 203(b) loan has tradition-ally been used to expand homeownership opportunities, we expect it to achieve areasonably high level of home-buying capacity. The prior and current FHA 203(b)products have some distinctive loan terms; we therefore expect some differencesin affordability between these versions.
Given the very limited financial resources of renters, it will be a challenge for eventhe most aggressive mortgage products to bring large numbers of renters to home-ownership. Thus, we anticipate a large share of renters to remain unserved. We antic-ipate that more whites and recent immigrant renters, given these groups’ greaterrelative financial resources, than black and Hispanic renters will be served.
Model Calibration and Analysis 43
The Potential and Limitations of Mortgage Innovation
When renters are categorized according to income group (low, moderate, middle, andupper), we anticipate that the number of middle- and upper-income renters servedwill be greater than the number of low- and moderate-income renters served.
In the body of this report, we present our affordability results without consideringFICO scores imputed by the credit submodel. We apply FICO information in appen-dix B. Because credit is an additional underwriting hurdle, we expect the number ofpeople served by the respective mortgage instruments to decrease when the FICOresults are applied.
44 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
RESULTS: HOMEOWNERSHIP AFFORDABILITY
The results of the affordability analysis are summarized in table 12 and detailedbelow.
Absolute Home-Buying Capacity
We begin with the absolute home-buying capacity measures, which provide a “first-cut” estimate of the potential mortgage market. Consider first the case in which allcurrent renter families apply for a mortgage and are evaluated on the basis of His-torical Mortgage underwriting standards. Our models suggest that the HistoricalMortgage provides approximately $314 billion in aggregate national home-purchasingpower (table 12). This figure is simply the sum total of the prices of homes above thestipulated threshold (i.e., a low-priced house) for which renters can qualify with thisloan product. Thus, it includes some renters who are able to afford very expensivehomes as well as a few renters unable to qualify for anything more than a low-pricedhouse. Moreover, this figure does not consider variations in credit history or differ-ences in preference for homeownership among different demographic groups. Never-theless, the estimate of $314 billion may be interpreted as a baseline against whichother mortgage instruments may be measured.
The past decade has brought dramatic integration in America’s housing and capitalmarkets. The Historical Mortgage has been replaced with standardized products con-forming to the GSEs’ purchasing guidelines. Our models suggest that the advent anddiffusion of the GSE Standard Mortgage boosts total purchasing capacity by approx-imately $74 billion, to $388 billion. Two factors have contributed to the increase inpurchasing power: first, relaxed debt ratios and higher LTVs allow renters to qualifyfor larger loans; second, many families that would be disqualified by historical crite-ria from purchasing even a low-priced house can now afford such a unit with the GSEStandard Mortgage guidelines.
The GSE Affordable Mortgages (when the 3/2 Option is not effected) increase aggre-gate home-buying capacity to a range between $445 billion and $518 billion, depend-ing on the individual product. GSE Affordable Mortgages from Fannie Mae provideaggregate home-buying capacity in the $445 billion to $477 billion range, and FreddieMac GSE Affordable Mortgages yield higher aggregate home-buying capacity—inthe $516 billion to $518 billion range. The Freddie Mac GSE Affordable products pro-vide greater home-buying capacity because they have no front-end ratio; Fannie MaeGSE Affordable loans have a 33 percent front-end ratio. Additionally, the Freddie Macback-end ratios are higher than those allowed by Fannie Mae in the GSE Affordablecategory (39 percent compared with 36 percent to 38 percent).
For both the Fannie Mae and the Freddie Mac GSE Affordable Mortgages, home-buying capacity is the same for the “base” affordable product (CHBP, AG) as for the3/2 variation because we do not assume here that the 2 percent down payment is
Results: Homeownership Affordability 45
The Potential and Limitations of Mortgage Innovation
46 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 12
.Im
pac
t o
f A
lter
nat
ive
Mo
rtg
age
Inst
rum
ents
:A
bso
lute
an
d R
efer
ence
Ho
me-
Bu
yin
g C
apac
ity
for
All
Ren
ters
,Nat
ion
wid
e
Ref
eren
ce H
ome-
Buy
ing
Cap
acity
(Per
cent
age
of R
ente
rs W
ho C
ould
Abs
olut
e H
ome-
Buy
ing
Cap
acity
Affo
rd th
e In
dica
ted
Hou
se)
Tota
l Nat
iona
lA
vera
ge/M
edia
n50
th25
th10
thH
ome-
Buy
ing
Cap
acity
aH
ome-
Buy
ing
Cap
acity
bTa
rget
Per
cent
ileP
erce
ntile
Per
cent
ileLo
an N
ame
(in $
Bill
ions
)M
ean(
$)M
edia
n($)
Hou
seH
ouse
Hou
seH
ouse
His
toric
al M
ortg
age
314.
312
,199
0 3.
85.
17.
510
.0
GS
E S
tand
ard
Mor
tgag
e (F
anni
e M
ae
387.
715
,049
0 5.
06.
39.
212
.1an
d Fr
eddi
e M
ac)
GS
E A
fford
able
Mor
tgag
esc
Fann
ie M
ae
CH
BP
;CH
BP,
3/2
Opt
ion
(NA
)47
7.3
18,5
260
6.4
7.8
11.0
14.3
CH
BP,
3/2
Opt
ion
(A)
538.
420
,898
0 7.
28.
812
.316
.2Fa
nnie
97
445.
017
,273
0 5.
97.
110
.614
.1Fr
eddi
e M
acA
G;A
G, 3
/2 O
ptio
n (N
A)
516.
120
,032
0 7.
08.
411
.414
.9A
G, 3
/2 O
ptio
n (A
)58
3.7
22,6
580
7.9
9.4
12.9
16.9
AG
97
518.
020
,108
0 7.
08.
511
.514
.8
Em
ergi
ng G
SE
Affo
rdab
le M
ortg
ages
Fann
ie M
aeFl
ex 9
748
7.5
18,9
240
6.7
8.1
11.3
14.3
Fred
die
Mac
Com
mun
ity G
old
539.
520
,940
0 7.
58.
711
.814
.9
Por
tfolio
Affo
rdab
le M
ortg
ages
d
Ban
k of
Am
eric
a Ze
ro D
own
511.
419
,850
0 7.
18.
511
.714
.9B
ank
of A
mer
ica
Cre
dit F
lex
546.
421
,208
0 7.
78.
811
.914
.9P
ortfo
lio C
ompo
site
e58
4.1
22,6
740
8.0
9.4
13.0
17.0
Gov
ernm
ent A
fford
able
Mor
tgag
esFH
A 2
03(b
)–pr
iorf
601.
323
,341
0 7.
99.
414
.220
.1FH
A 2
03(b
)–cu
rren
tg56
2.4
21,8
320
7.8
9.2
13.0
17.3
Sou
rce:
Aut
hors
’ana
lysi
s of
199
3 S
IPP
data
.a
The
agg
rega
te v
alue
of
hous
es t
hat
coul
d be
pur
chas
ed b
y cu
rren
t re
nter
s ab
ove
a lo
w-p
riced
hou
se t
hres
hold
.b
For
all
curr
ent
rent
ers,
not
just
ren
ters
who
can
affo
rd a
t le
ast
a lo
w-p
riced
hou
se.
cA
= 3
/2 O
ptio
n is
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
g is
fort
hcom
ing)
;NA
= 3
/2 O
ptio
n is
not
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
g is
not
fort
h-co
min
g).
dT
hese
are
sel
ecte
d as
exa
mpl
es o
f th
e m
any
port
folio
mor
tgag
es c
urre
ntly
offe
red.
eIn
corp
orat
es t
he m
ortg
age
char
acte
ristic
s of
a c
ompo
site
of
port
folio
pro
duct
s.f
Inco
rpor
ated
FH
A(b
) te
rms
befo
re 1
997.
gIn
corp
orat
ed F
HA
(b)
term
s fr
om 1
997
onw
ard.
provided by a source other than the buyer.26 For both the Fannie Mae and FreddieMac Affordables, the “97” product (Fannie 97 and AG 97) has a somewhat lower (orequivalent) home-buying capacity than the base product; although the 97 product hasa higher LTV (97 percent compared with 95 percent), this is somewhat offset by itsother requirements (e.g., more costly mortgage insurance [because the LTV is higher],more demanding reserves).
The Emerging GSE Affordable Mortgages incrementally increase home-buying capac-ity from the GSE Affordable baseline. Fannie Mae’s Flex 97 realizes approximately$488 billion in home-buying capacity (compared with $445 billion to $477 billion withthe Fannie Mae Affordable Mortgages). Freddie Mac’s Community Gold reaches ap-proximately $540 billion in home-buying capacity (compared with $516 billion to $518billion with the Freddie Mac Affordable Mortgages). For both GSEs, the gain isachieved because the Emerging GSE Affordable Mortgages have more aggressive fea-tures (e.g., higher debt ratios, higher LTVs), albeit with some offsets (e.g., higher re-quired reserves, higher mortgage insurance costs).
The portfolio products studied here achieve significant aggregate home-buying capac-ity in the $511 billion to $584 billion range—a function of their high LTVs and fairlyaggressive front-end and back-end ratios. Of the two Bank of America portfolio prod-ucts examined here, Zero Down and Credit Flex, the latter realizes greater home-buying capacity: Although Credit Flex has a lower LTV than does Zero Down (97 per-cent compared with 100 percent), it has lower mortgage insurance requirements (a“payback” from its lower LTV), no front-end ratio (recalling the Freddie Mac approach),and a higher back-end ratio (43 percent compared with 41 percent).
The Portfolio Composite product has the highest home-buying capacity ($584 billion)considered thus far because its parameters extend so far beyond the “normal” market(e.g., the product does not require mortgage insurance on a very high LTV and allowsvery high front-end and back-end ratios) that it is essentially a subsidized loan. (TheDVMP-inspired Portfolio Composite Mortgage was so costly to its originators that itultimately was rescinded. That recalls the 45 percent back-end ratio allowed by lendersin the Atlanta Mortgage consortium, which also was ultimately dropped.)
The Governmental Affordable Mortgages realize very high home-buying capacity inthe $562 billion to $601 billion range—a reflection of their high LTVs and high front-end and back-end ratios. Of the two FHA 203(b) products, the “prior” product hassomewhat greater home-buying prowess than does the “current” product—perhapsbecause the former permits a portion of the closing costs to be financed.
Other measures in parallel denote varying housing affordability by mortgage type(table 12). Under the Historical Mortgage, the mean home affordable, including
Results: Homeownership Affordability 47
The Potential and Limitations of Mortgage Innovation
26 In the Policy Considerations section, we examine the impacts assuming that the 2 percent down payment isprovided by a source other than the buyer. Results with the 3/2 Option activated are also shown in all of the ref-erenced tables.
those with no home-buying capacity, is $12,199.27 The mean affordable increases to$15,049 with the GSE Standard Mortgage; to a range between $17,273 and $20,940with the GSE Affordable Mortgages (without the 3/2 Option activated), including theEmerging GSE Affordable Mortgages group; to a range between $19,850 and $22,674for the Portfolio Affordable Mortgages; and to a range between $21,832 and $23,341for the Governmental Affordable Mortgages.
Table 13 disaggregates the absolute home-buying capacity by subgroups of renter fam-ilies: whites, blacks, Hispanics, recent immigrants, and others. (Here and elsewherein this study, the referenced white, black, and other groups are non-Hispanic.) Forwhite renter families, absolute home-buying capacity increased from approximately$285 billion under the Historical Mortgage to approximately $347 billion under theGSE Standard Mortgage, and to approximately $505 billion for the Portfolio Com-posite. For black renter families, total purchasing capacity under these three mort-gage products increased from approximately $11 billion to $17 billion and then to$36 billion. This progression means that mortgage innovation increases the totalpurchase power of whites by about 77 percent; blacks, starting from a smaller base,realize a 227 percent gain in purchasing power. Hispanic families, like blacks, simi-larly experience a quantum increase of total purchasing power (a 139 percent gain)from mortgage innovation. The pattern for recent immigrant families resemblesthat described for whites.
Mortgage innovation is thus dramatically expanding the total purchasing capacity oftraditionally underserved groups, most notably blacks and Hispanics. Yet because thelatter groups have fewer financial resources (e.g., lower income and assets), thesegroups capture a disproportionately small fraction of total buying capacity relative totheir share of the total population. We can express that relationship in the form of aratio (table 14). For example, black renter families make up 17.3 percent of all renterfamilies (4,464,525 of 25,762,939). Under the Historical Mortgage, black renter fam-ilies capture 3.6 percent of the total home-buying capacity ($11.356 billion of $314.291billion)—representing a 0.21 ratio (3.6 percent divided by 17.3 percent) of their dollarhome-buying capacity relative to their population representation. Greater home-buyingaccess for blacks in the form of rising ratios is afforded by today’s more aggressivemortgage products; however, even with these, the ratios never exceed 0.40. Hispanicfamilies fare similarly. Their ratio, starting at 0.20 for the Historical Mortgage, riseswith today’s more aggressive loan products, yet it never exceeds 0.30. (For both blacksand Hispanics, the highest ratios are realized by the FHA 203(b)–prior product.)
In contrast, white renter families capture a disproportionately large share of the totalpurchasing capacity relative to their population representation—their ratios are sig-nificantly above 1.00 (table 14). The ratio for whites is 1.45 for the Historical Mort-gage and drops slightly to no lower than approximately 1.34 under today’s more ag-gressive products. Recent immigrant families, realizing relative total purchasing capac-ity that is nearly identical to their proportional share of the renter population, have
48 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
27 That is, renters who have a home-buying capacity of zero are included in the calculation of the mean.
ratios near 1.00. For example, under the GSE Standard Mortgage, recent immigrantfamilies realize $10.724 billion of the total $387.695 billion purchasing capacity—or2.8 percent—a capture rate that is nearly identical to their 2.7 percent share of therenter population.
This does not presume that there should be full equality in home-buying prowess (i.e.,that the ratios should always be 1.0). Under our political and economic system, thereis far from equal access to a broad array of goods and services. The point of the aboveexercise is merely to gauge the degree of relative home-buying capacity; the fact thatthe ratios are for the most part far from 1.0 points to considerable unevenness. Wealso learn that mortgage innovation is reducing some of the disparity, albeit by aminor measure.
In sum, analysis of absolute home-buying capacity reveals much about the ability ofmortgage innovation to expand homeownership. Yet because it is an aggregate mea-sure that is unlinked to a benchmark, it obscures certain aspects of affordability. Con-sider, for example, how the aggregate capacity measure treats two hypothetical fami-lies. Renter A is able to qualify for a home purchase of $200,000 under the HistoricalMortgage but might be able to secure a $300,000 unit under the more liberal GSEStandard guidelines. By contrast, renter B might be unable to purchase anything
Results: Homeownership Affordability 49
The Potential and Limitations of Mortgage Innovation
Table 13. Absolute Home-Buying Capacity for Renter Groupsa
by Mortgage Instrument, Nationwide (in $ Millions)
White Black OtherFamilies Families Families Recent
All (Non- (Non- Hispanic (Non- ImmigrantLoan Name Families Hispanic) Hispanic) Families Hispanic) Families
1.Historical 314,291 285,347 11,356 9,631 7,957 9,0592.GSE Standard 387,695 347,340 16,702 13,422 10,231 10,7243.Fannie Mae CHBP; CHBP 3/2 (NA)b 477,289 421,454 24,602 16,604 14,629 13,2434.Fannie Mae CHBP 3/2 (A)c 538,396 466,417 32,271 20,639 19,069 14,6645.Fannie Mae 97 445,006 391,387 24,016 15,822 13,781 12,5786.Freddie Mac AG 5; AG 3/2 (NA)b 516,087 455,364 26,427 18,680 15,616 14,4977.Freddie Mac AG 3/2 (A)c 583,749 505,947 34,754 22,764 20,284 16,4248.Freddie Mac AG 97 518,046 457,728 26,054 18,783 15,481 14,7479.Fannie Mae Flex 97 487,533 430,821 24,941 17,048 14,723 13,607
10.Freddie Mac Community Gold 539,486 475,988 27,567 19,784 16,147 15,22911.Bank of America Zero Down 511,393 446,771 27,186 19,492 17,943 14,54912.Bank of America Credit Flex 546,372 482,359 27,681 19,963 16,369 15,41813.Portfolio Composite 584,144 505,042 35,649 23,053 20,400 16,44614.FHA 203(b)–prior 601,325 507,648 41,687 27,887 24,103 15,79215.FHA 203(b)–current 562,449 480,712 36,183 23,335 22,219 14,920
Weighted sample size 25,762,939 16,183,879 4,464,525 3,980,131 1,134,403 704,035
Source: Authors’ analysis of 1993 SIPP data.a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,
they sum to all renter families. Recent immigrant families—those entering the United States after 1984—overlapwith the previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted,whites, blacks, and other groups are non-Hispanic.
b NA = 3/2 Option is not activated (i.e., 2 percent outside funding is not forthcoming).c A = 3/2 Option is activated (i.e., 2 percent outside funding is forthcoming).
under the Historical Mortgage but might secure a home of $100,000 under the GSEStandard. The aggregate home-buying measure treats both of these families equal-ly—each inducing an increase of $100,000 in total purchasing power. Clearly, however,affordability means very different things to each of these families, and it is arguablymuch more important to expand homeownership opportunities than to boost the
50 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 14. Relative Home-Buying Capacity for Renter Groupsa
by Mortgage Instrument, NationwideRecent
All White Black Hispanic Other ImmigrantLoan Name Families Families Families Families Families Families
Percentage of Families 100.0 62.9 17.3 15.4 4.4 2.7
Percentage of home-buying capacity by mortgage instrument1.Historical 100.0 90.8 3.6 3.1 2.5 2.92.GSE Standard 100.0 89.6 4.3 3.5 2.6 2.83.Fannie Mae CHBP; CHBP 3/2 (NA)b 100.0 88.2 5.2 3.5 3.1 2.84.Fannie Mae CHBP 3/2 (A)c 100.0 86.7 6.0 3.8 3.5 2.75.Fannie Mae 97 100.0 88.0 5.4 3.6 3.1 2.86.Freddie Mac AG 5; AG 3/2 (NA)b 100.0 88.3 5.1 3.6 3.0 2.87.Freddie Mac AG 3/2 (A)c 100.0 86.7 6.0 3.9 3.5 2.88.Freddie Mac AG 97 100.0 88.4 5.0 3.6 3.0 2.89.Fannie Mae Flex 97 100.0 88.4 5.1 3.5 3.0 2.8
10.Freddie Mac Community Gold 100.0 88.2 5.1 3.7 3.0 2.811.Bank of America Zero Down 100.0 87.4 5.3 3.8 3.5 2.812.Bank of America Credit Flex 100.0 88.2 5.1 3.7 3.0 2.813.Portfolio Composite 100.0 86.5 6.1 3.9 3.5 2.814.FHA 203(b)–prior 100.0 84.5 6.9 4.6 4.0 2.615.FHA 203(b)–current 100.0 85.5 6.4 4.1 4.0 2.7
Relative ratiod by mortgage instrument1.Historical 1.00 1.45 0.21 0.20 0.57 1.052.GSE Standard 1.00 1.43 0.25 0.22 0.60 1.013.Fannie Mae CHBP; CHBP 3/2 (NA)b 1.00 1.41 0.30 0.23 0.70 1.024.Fannie Mae CHBP 3/2 (A)c 1.00 1.38 0.35 0.25 0.80 1.005.Fannie Mae 97 1.00 1.40 0.31 0.23 0.70 1.036.Freddie Mac AG 5; AG 3/2 (NA)b 1.00 1.40 0.30 0.23 0.69 1.037.Freddie Mac AG 3/2 (A)c 1.00 1.38 0.34 0.25 0.79 1.038.Freddie Mac AG 97 1.00 1.41 0.29 0.23 0.68 1.049.Fannie Mae Flex 97 1.00 1.41 0.30 0.23 0.69 1.02
10.Freddie Mac Community Gold 1.00 1.40 0.29 0.24 0.68 1.0311.Bank of America Zero Down 1.00 1.39 0.31 0.25 0.80 1.0412.Bank of America Credit Flex 1.00 1.41 0.29 0.24 0.68 1.0313.Portfolio Composite 1.00 1.38 0.35 0.26 0.79 1.0314.FHA 203(b)–prior 1.00 1.34 0.40 0.30 0.91 0.9615.FHA 203(b)–current 1.00 1.36 0.37 0.27 0.90 0.97
Source: Authors’ analysis of 1993 SIPP data.a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,
they sum to all renter families. Recent immigrant families, those entering the United States after 1984, overlap withthe previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted, whites,blacks, and other groups are non-Hispanic.
b NA = 3/2 Option is not activated (i.e., 2 percent outside funding is not forthcoming).c A = 3/2 Option is activated (i.e., 2 percent outside funding is forthcoming).d Equals group’s share of home-buying capacity divided by group’s share of all renter families.
housing consumption of middle- or higher-income buyers. Similarly, any gain abovea stipulated threshold (i.e., the low-cost houses) is counted—even a $1 advantage inthe ability to secure a housing asset. Thus, the absolute aggregate home-buying capac-ity must be viewed as a largely theoretical measure of the home-buying and business-expanding potential of mortgage product innovation. Despite these caveats, the aggre-gate home-buying measure powerfully communicates the dollar value of the housingasset reach of different mortgage instruments, and we find that there are multi-bil-lion-dollar differences in this aggregate reach across the different mortgages. Furtherinsight in this regard is afforded in considering the reference home-buying capacity.
Reference Home-Buying Capacity
Both the effectiveness and the shortfalls of the mortgage instruments in expandinghomeownership are apparent when we examine the percentage of renters able tosecure the various reference-priced homes. These figures are summarized in tables15 through 18 for the four reference-priced homes: the target house and the threecriterion units (homes priced at the 50th [median-priced], 25th [modestly priced], and10th [low-priced] percentiles). Only 3.8 percent of all renter families (approximately0.986 million) can afford the target house under the Historical Mortgage. With today’sinnovative mortgage products, the percentage of renters able to afford the targethouse increases to the following levels: 5.0 percent (approximately 1.288 million renterfamilies) under the GSE Standard Mortgage; 5.9 percent to 7.0 percent (approximately1.518 million to 1.804 million renters) under the GSE Affordable Mortgages;28 6.7percent to 7.5 percent (approximately 1.718 million to 1.932 million renters) underthe Emerging GSE Affordable Mortgages; 7.1 percent to 8.0 percent with the Portfo-lio Affordable Mortgages (approximately 1.830 million to 2.049 million renters); and7.9 percent (approximately 2.028 million renters) for the highest-achieving Govern-mental Affordable Mortgage (FHA 203(b)–prior). There is a similar continuum bymortgage product—but at a higher share affording—when the reference home is thelow-priced unit. Ten percent of all renters (approximately 2.580 million) can affordthe low-priced home under the Historical Mortgage. This percentage rises to the fol-lowing levels under the current more innovative mortgage products: 12.1 percent (ap-proximately 3.106 million renters) under the GSE Standard Mortgage; 14.1 percentto 14.9 percent (approximately 3.635 to 3.842 million renters) under the GSE Afford-able Mortgages,29 including the Emerging GSE Affordable Group; 14.9 percent to 17.0percent (approximately 3.828 million to 4.373 million renters) under the Portfolio Af-fordable Mortgages; and 20.1 percent (5.168 million renters) for the best-performingGovernmental Affordable Mortgage (FHA 203(b)–prior). (As observed in the preced-ing section, the FHA 203(b)–prior again slightly outperforms the FHA 203(b)–cur-rent with respect to the reference home-buying capacity.)
Results: Homeownership Affordability 51
The Potential and Limitations of Mortgage Innovation
28 Without the 3/2 Option activated.
29 Without the 3/2 Option activated.
52 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 15
.Ref
eren
ce H
om
e-B
uyi
ng
Cap
acit
y by
Mo
rtg
age
Inst
rum
ent
and
Ren
ter
Gro
up
,aN
atio
nw
ide:
Targ
et H
ou
se (
Nu
mb
er a
nd
Per
cen
tag
e o
f R
ente
rs W
ho
Co
uld
Aff
ord
th
e In
dic
ated
Ho
use
)
Rec
ent
His
pani
cIm
mig
rant
All
Fam
ilies
Whi
te F
amili
esB
lack
Fam
ilies
Fam
ilies
Oth
er F
amili
esFa
mili
es
Loan
Nam
eN
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%
1.H
isto
rical
986,
175
3.8
887,
739
5.5
43,1
13
1.0
16,9
28
0.4
38,3
97
3.4
32,4
81
4.6
2.G
SE
Sta
ndar
d1,
288,
186
5.0
1,13
8,26
6 7.
066
,321
1.
539
,384
1.
044
,215
3.
932
,481
4.
63.
Fann
ie M
ae C
HB
P;C
HB
P 3
/2 (
NA
)b1,
655,
784
6.4
1,44
5,20
48.
995
,041
2.
160
,199
1.
555
,343
4.
945
,747
6.
54.
Fann
ie M
ae C
HB
P 3
/2 (
A)c
1,86
0,78
0 7.
21,
621,
625
10.0
115,
810
2.6
68,0
60
1.7
55,3
48
4.9
45,7
47
6.5
5.Fa
nnie
Mae
97
1,51
7,66
9 5.
91,
317,
934
8.1
89,6
66
2.0
60,1
99
1.5
49,8
82
4.4
32,4
81
4.6
6.Fr
eddi
e M
ac A
G 5
;AG
3/2
(N
A)b
1,80
3,89
5 7.
01,
570,
079
9.7
95,0
41
2.1
72,5
82
1.8
66,1
95
5.8
45,7
47
6.5
7.Fr
eddi
e M
ac A
G 3
/2 (
A)c
2,04
5,19
1 7.
91,
782,
816
11.0
115,
810
2.6
80,4
38
2.0
66,1
92
5.8
45,7
47
6.5
8.Fr
eddi
e M
ac A
G 9
71,
793,
796
7.0
1,57
2,26
49.
795
,041
2.
165
,975
1.
760
,527
5.
345
,747
6.
59.
Fann
ie M
ae F
lex
971,
717,
512
6.7
1,49
9,27
59.
395
,041
2.
160
,199
1.
563
,005
5.
645
,747
6.
510
.Fre
ddie
Mac
Com
unity
Gol
d1,
932,
272
7.5
1,68
5,22
710
.410
4,96
1 2.
465
,991
1.
776
,164
6.
753
,724
7.
611
.Ban
k of
Am
eric
a Ze
ro D
own
1,83
0,45
7 7.
11,
604,
680
9.9
109,
943
2.5
60,1
99
1.5
55,6
29
4.9
45,7
47
6.5
12.B
ank
of A
mer
ica
Cre
dit F
lex
1,96
9,90
7 7.
61,
722,
792
10.6
104,
968
2.4
65,9
82
1.7
76,1
65
6.7
53,7
24
7.6
13.P
ortfo
lio C
ompo
site
2,04
8,78
1 8.
01,
785,
274
11.0
115,
814
2.6
73,8
37
1.9
73,8
56
6.5
45,7
47
6.5
14.F
HA
203
(b)–
prio
r2,
028,
033
7.9
1,73
6,20
710
.713
3,22
1 3.
074
,986
1.
983
,651
7.
445
,747
6.
515
.FH
A 2
03(b
)–cu
rren
t1,
997,
478
7.8
1,72
3,58
310
.712
5,05
1 2.
868
,060
1.
780
,872
7.
145
,747
6.
5
Wei
ghte
d sa
mpl
e si
ze25
,762
,939
100
.016
,183
,879
100.
04,
464,
525
100.
03,
980,
131
100.
01,
134,
403
100.
070
4,03
5 10
0.0
Sou
rce:
Aut
hors
’ana
lysi
s of
199
3 S
IPP
data
.a
Non
-His
pani
c w
hite
, non
-His
pani
c bl
ack,
His
pani
c, a
nd o
ther
non
-His
pani
c re
nter
fam
ilies
are
dis
cret
e.C
ombi
ned,
they
sum
to a
ll re
nter
fam
ilies
.Rec
ent
imm
igra
nt fa
mili
es,
thos
e en
terin
g th
e U
nite
d S
tate
s af
ter
1984
, ov
erla
p w
ith t
he p
revi
ousl
y lis
ted
raci
al/e
thni
c gr
oups
(e.
g.,
non-
His
pani
c w
hite
s an
dbl
acks
).F
inal
ly,
if no
t ot
herw
ise
note
d, w
hite
s, b
lack
s, a
nd o
ther
gro
ups
are
non-
His
pani
c.b
NA
= 3
/2 O
ptio
n is
not
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
g is
not
fort
hcom
ing)
.c
A =
3/2
Opt
ion
is a
ctiv
ated
(i.e
., 2
perc
ent
outs
ide
fund
ing
is fo
rthc
omin
g).
Results: Homeownership Affordability 53
The Potential and Limitations of Mortgage Innovation
Tabl
e 16
.Ref
eren
ce H
om
e-B
uyi
ng
Cap
acit
y by
Mo
rtg
age
Inst
rum
ent
and
Ren
ter
Gro
up
,aN
atio
nw
ide:
50th
Per
cen
tile
(M
edia
n-P
rice
d)
Ho
use
(N
um
ber
an
d P
erce
nta
ge
of
Ren
ters
Wh
o C
ou
ld A
ffo
rd t
he
Ind
icat
ed H
ou
se)
Rec
ent
His
pani
cIm
mig
rant
All
Fam
ilies
Whi
te F
amili
esB
lack
Fam
ilies
Fam
ilies
Oth
er F
amili
esFa
mili
es
Loan
Nam
eN
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%
1.H
isto
rical
1,31
4,43
6 5.
11,
204,
814
7.4
59,6
62
1.3
30,5
44
0.8
19,4
17
1.7
38,0
82
5.4
2.G
SE
Sta
ndar
d1,
631,
007
6.3
1,49
5,35
1 9.
271
,036
1.
639
,384
1.
025
,235
2.
244
,904
6.
43.
Fann
ie M
ae C
HB
P;C
HB
P 3
/2 (
NA
)b2,
021,
824
7.8
1,80
8,06
3 11
.210
5,27
3 2.
460
,458
1.
548
,076
4.
251
,348
7.
34.
Fann
ie M
ae C
HB
P 3
/2 (
A)c
2,27
4,32
6 8.
82,
010,
361
12.4
130,
364
2.9
73,5
93
1.8
59,9
31
5.3
51,3
48
7.3
5.Fa
nnie
Mae
97
1,83
3,29
1 7.
11,
633,
763
10.1
114,
515
2.6
59,8
61
1.5
25,2
40
2.2
44,9
04
6.4
6.Fr
eddi
e M
ac A
G 5
;AG
3/2
(N
A)b
2,15
5,61
1 8.
41,
923,
940
11.9
115,
229
2.6
60,4
58
1.5
56,0
51
4.9
59,3
25
8.4
7.Fr
eddi
e M
ac A
G 3
/2 (
A)c
2,41
3,93
6 9.
42,
126,
400
13.1
140,
320
3.1
73,5
93
1.8
73,6
57
6.5
59,3
25
8.4
8.Fr
eddi
e M
ac A
G 9
72,
180,
704
8.5
1,93
8,34
3 12
.012
0,54
2 2.
765
,752
1.
756
,051
4.
959
,325
8.
49.
Fann
ie M
ae F
lex
972,
082,
882
8.1
1,84
7,87
5 11
.412
1,16
7 2.
765
,752
1.
748
,076
4.
251
,348
7.
310
.Fre
ddie
Mac
Com
unity
Gol
d2,
243,
952
8.7
1,98
9,64
6 12
.312
5,45
3 2.
867
,065
1.
761
,791
5.
459
,325
8.
411
.Ban
k of
Am
eric
a Ze
ro D
own
2,20
2,34
5 8.
51,
966,
180
12.1
114,
515
2.6
73,5
93
1.8
48,0
76
4.2
60,8
89
8.6
12.B
ank
of A
mer
ica
Cre
dit F
lex
2,27
0,01
9 8.
82,
021,
017
12.5
120,
130
2.7
67,0
79
1.7
61,7
93
5.4
59,3
25
8.4
13.P
ortfo
lio C
ompo
site
2,43
4,29
4 9.
42,
139,
630
13.2
145,
212
3.3
81,5
40
2.0
67,9
12
6.0
59,3
25
8.4
14.F
HA
203
(b)–
prio
r2,
411,
669
9.4
2,11
2,64
4 13
.115
3,66
9 3.
481
,553
2.
063
,776
5.
660
,889
8.
615
.FH
A 2
03(b
)–cu
rren
t2,
365,
476
9.2
2,07
2,18
4 12
.813
9,11
5 3.
181
,553
2.
072
,602
6.
460
,889
8.
6
Wei
ghte
d sa
mpl
e si
ze25
,762
,939
100.
016
,183
,879
100.
04,
464,
525
100.
03,
980,
131
100.
01,
134,
403
100.
070
4,03
510
0.0
Sou
rce:
Aut
hors
’ana
lysi
s of
199
3 S
IPP
data
.a
Non
-His
pani
c w
hite
, non
-His
pani
c bl
ack,
His
pani
c, a
nd o
ther
non
-His
pani
c re
nter
fam
ilies
are
dis
cret
e.C
ombi
ned,
they
sum
to a
ll re
nter
fam
ilies
.Rec
ent
imm
igra
nt fa
mili
es,
thos
e en
terin
g th
e U
nite
d S
tate
s af
ter
1984
, ov
erla
p w
ith t
he p
revi
ousl
y lis
ted
raci
al/e
thni
c gr
oups
(e.
g.,
non-
His
pani
c w
hite
s an
dbl
acks
).F
inal
ly,
if no
t ot
herw
ise
note
d, w
hite
s, b
lack
s, a
nd o
ther
gro
ups
are
non-
His
pani
c.b
NA
= 3
/2 O
ptio
n is
not
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
g is
not
fort
hcom
ing)
.c
A =
3/2
Opt
ion
is a
ctiv
ated
(i.e
., 2
perc
ent
outs
ide
fund
ing
is fo
rthc
omin
g).
54 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Tabl
e 17
.Ref
eren
ce H
om
e-B
uyi
ng
Cap
acit
y by
Mo
rtg
age
Inst
rum
ent
and
Ren
ter
Gro
up
,aN
atio
nw
ide:
25th
Per
cen
tile
(M
od
estl
y P
rice
d)
Ho
use
(N
um
ber
an
d P
erce
nta
ge
of
Ren
ters
Wh
o C
ou
ld A
ffo
rd t
he
Ind
icat
ed H
ou
se)
Rec
ent
His
pani
cIm
mig
rant
All
Fam
ilies
Whi
te F
amili
esB
lack
Fam
ilies
Fam
ilies
Oth
er F
amili
esFa
mili
es
Loan
Nam
eN
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%
1.H
isto
rical
1,92
6,93
9 7.
51,
742,
420
10.8
71,0
36
1.6
51,6
90
1.3
61,7
93
5.4
52,8
81
7.5
2.G
SE
Sta
ndar
d2,
381,
156
9.2
2,11
9,39
9 13
.111
9,60
0 2.
772
,703
1.
869
,455
6.
159
,325
8.
43.
Fann
ie M
ae C
HB
P;C
HB
P 3
/2 (
NA
)b2,
835,
727
11.0
2,51
6,10
8 15
.514
5,23
1 3.
392
,936
2.
381
,314
7.
264
,229
9.
14.
Fann
ie M
ae C
HB
P 3
/2 (
A)c
3,17
5,28
2 12
.32,
761,
617
17.1
193,
225
4.3
106,
389
2.7
113,
860
10.0
83,4
70
11.9
5.Fa
nnie
Mae
97
2,72
1,33
9 10
.62,
388,
741
14.8
145,
231
3.3
100,
021
2.5
87,4
40
7.7
68,8
69
9.8
6.Fr
eddi
e M
ac A
G 5
;AG
3/2
(N
A)b
2,94
5,99
2 11
.42,
620,
170
16.2
145,
231
3.3
99,3
44
2.5
81,3
14
7.2
64,2
29
9.1
7.Fr
eddi
e M
ac A
G 3
/2 (
A)c
3,32
3,16
1 12
.92,
891,
897
17.9
202,
154
4.5
115,
185
2.9
113,
860
10.0
83,4
70
11.9
8.Fr
eddi
e M
ac A
G 9
72,
962,
996
11.5
2,62
7,45
3 16
.214
1,65
9 3.
210
6,38
9 2.
787
,440
7.
764
,229
9.
19.
Fann
ie M
ae F
lex
972,
915,
592
11.3
2,57
1,45
7 15
.915
0,14
2 3.
410
6,38
9 2.
787
,440
7.
764
,229
9.
110
.Fre
ddie
Mac
Com
unity
Gol
d3,
037,
966
11.8
2,68
6,03
8 16
.615
5,32
1 3.
511
5,18
5 2.
981
,552
7.
264
,229
9.
111
.Ban
k of
Am
eric
a Ze
ro D
own
3,02
0,44
7 11
.72,
651,
081
16.4
146,
883
3.3
106,
946
2.7
115,
607
10.2
73,7
69
10.5
12.B
ank
of A
mer
ica
Cre
dit F
lex
3,05
8,50
8 11
.92,
706,
465
16.7
155,
321
3.5
115,
170
2.9
81,5
51
7.2
64,2
26
9.1
13.P
ortfo
lio C
ompo
site
3,34
6,09
9 13
.02,
906,
418
18.0
210,
651
4.7
115,
170
2.9
113,
860
10.0
83,4
73
11.9
14.F
HA
203
(b)–
prio
r3,
654,
988
14.2
3,07
3,48
1 19
.028
4,25
6 6.
415
0,36
9 3.
814
6,89
4 12
.987
,244
12
.415
.FH
A 2
03(b
)–cu
rren
t3,
351,
243
13.0
2,87
7,49
4 17
.823
0,54
8 5.
212
2,34
9 3.
112
0,89
3 10
.783
,632
11
.9
Wei
ghte
d sa
mpl
e si
ze25
,762
,939
100
.016
,183
,879
100
.04,
464,
525
100.
03,
980,
131
100.
01,
134,
403
100.
070
4,03
5 10
0.0
Sou
rce:
Aut
hors
’ana
lysi
s of
199
3 S
IPP
data
.a
Non
-His
pani
c w
hite
, non
-His
pani
c bl
ack,
His
pani
c, a
nd o
ther
non
-His
pani
c re
nter
fam
ilies
are
dis
cret
e.C
ombi
ned,
they
sum
to a
ll re
nter
fam
ilies
.Rec
ent
imm
igra
nt fa
mili
es,
thos
e en
terin
g th
e U
nite
d S
tate
s af
ter
1984
, ov
erla
p w
ith t
he p
revi
ousl
y lis
ted
raci
al/e
thni
c gr
oups
(e.
g.,
non-
His
pani
c w
hite
s an
dbl
acks
).F
inal
ly,
if no
t ot
herw
ise
note
d, w
hite
s, b
lack
s, a
nd o
ther
gro
ups
are
non-
His
pani
c.b
NA
= 3
/2 O
ptio
n is
not
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
g is
not
fort
hcom
ing)
.c
A =
3/2
Opt
ion
is a
ctiv
ated
(i.e
., 2
perc
ent
outs
ide
fund
ing
is fo
rthc
omin
g).
Results: Homeownership Affordability 55
The Potential and Limitations of Mortgage Innovation
Tabl
e 18
.Ref
eren
ce H
om
e-B
uyi
ng
Cap
acit
y by
Mo
rtg
age
Inst
rum
ent
and
Ren
ter
Gro
up
,aN
atio
nw
ide:
10th
Per
cen
tile
(L
ow
-Pri
ced
) H
ou
se (
Nu
mb
er a
nd
Per
cen
tag
e o
f R
ente
rs W
ho
Co
uld
Aff
ord
th
e In
dic
ated
Ho
use
)
Rec
ent
His
pani
cIm
mig
rant
All
Fam
ilies
Whi
te F
amili
esB
lack
Fam
ilies
Fam
ilies
Oth
er F
amili
esFa
mili
es
Loan
Nam
eN
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%N
umbe
r%
Num
ber
%
1.H
isto
rical
2,58
0,00
6 10
.02,
296,
689
14.2
128,
499
2.9
79,0
90
2.0
75,7
29
6.7
59,3
25
8.4
2.G
SE
Sta
ndar
d3,
105,
523
12.1
2,73
5,40
8 16
.915
9,59
4 3.
611
5,17
0 2.
995
,352
8.
468
,825
9.
83.
Fann
ie M
ae C
HB
P;C
HB
P 3
/2 (
NA
)b3,
692,
602
14.3
3,20
4,73
2 19
.822
8,36
0 5.
112
5,29
5 3.
113
4,24
5 11
.893
,616
13
.34.
Fann
ie M
ae C
HB
P 3
/2 (
A)c
4,18
0,03
7 16
.23,
545,
564
21.9
298,
186
6.7
161,
991
4.1
174,
426
15.4
107,
091
15.2
5.Fa
nnie
Mae
97
3,63
5,15
1 14
.13,
141,
453
19.4
236,
843
5.3
122,
110
3.1
134,
699
11.9
97,9
38
13.9
6.Fr
eddi
e M
ac A
G 5
;AG
3/2
(N
A)b
3,84
1,51
2 14
.93,
312,
840
20.5
246,
040
5.5
148,
339
3.7
134,
245
11.8
99,4
87
14.1
7.Fr
eddi
e M
ac A
G 3
/2 (
A)c
4,34
3,88
9 16
.93,
671,
313
22.7
310,
284
7.0
187,
982
4.7
174,
426
15.4
120,
981
17.2
8.Fr
eddi
e M
ac A
G 9
73,
804,
928
14.8
3,29
5,84
7 20
.422
9,52
1 5.
114
5,15
5 3.
613
4,24
5 11
.899
,487
14
.19.
Fann
ie M
ae F
lex
973,
694,
663
14.3
3,21
5,25
1 19
.922
3,04
8 5.
012
2,11
0 3.
113
4,24
5 11
.893
,616
13
.310
.Fre
ddie
Mac
Com
unity
Gol
d3,
828,
115
14.9
3,29
9,89
3 20
.424
2,20
0 5.
415
1,84
2 3.
813
4,24
5 11
.899
,487
14
.111
.Ban
k of
Am
eric
a Ze
ro D
own
3,82
8,63
0 14
.93,
290,
183
20.3
231,
753
5.2
145,
036
3.6
161,
698
14.3
111,
498
15.8
12.B
ank
of A
mer
ica
Cre
dit F
lex
3,83
8,07
6 14
.93,
309,
805
20.5
242,
185
5.4
151,
846
3.8
134,
240
11.8
99,4
87
14.1
13.P
ortfo
lio C
ompo
site
4,37
3,12
4 17
.03,
687,
791
22.8
322,
936
7.2
187,
974
4.7
174,
422
15.4
120,
983
17.2
14.F
HA
203
(b)–
prio
r5,
168,
303
20.1
4,25
6,03
7 26
.342
1,09
4 9.
426
9,69
4 6.
822
1,50
4 19
.513
9,92
0 19
.915
.FH
A 2
03(b
)–cu
rren
t4,
467,
036
17.3
3,72
1,48
3 23
.034
2,83
1 7.
720
5,09
6 5.
219
7,73
8 17
.412
0,28
4 17
.1
Wei
ghte
d sa
mpl
e si
ze25
,762
,939
100
.016
,183
,879
100
.04,
464,
525
100.
03,
980,
131
100.
01,
134,
403
100.
070
4,03
5 10
0.0
Sou
rce:
Aut
hors
’ana
lysi
s of
199
3 S
IPP
data
.a
Non
-His
pani
c w
hite
, non
-His
pani
c bl
ack,
His
pani
c, a
nd o
ther
non
-His
pani
c re
nter
fam
ilies
are
dis
cret
e.C
ombi
ned,
they
sum
to a
ll re
nter
fam
ilies
.Rec
ent
imm
igra
nt fa
mili
es,
thos
e en
terin
g th
e U
nite
d S
tate
s af
ter
1984
, ov
erla
p w
ith t
he p
revi
ousl
y lis
ted
raci
al/e
thni
c gr
oups
(e.
g.,
non-
His
pani
c w
hite
s an
dbl
acks
).F
inal
ly,
if no
t ot
herw
ise
note
d, w
hite
s, b
lack
s, a
nd o
ther
gro
ups
are
non-
His
pani
c.b
NA
= 3
/2 O
ptio
n is
not
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
g is
not
fort
hcom
ing)
.c
A =
3/2
Opt
ion
is a
ctiv
ated
(i.e
., 2
perc
ent
outs
ide
fund
ing
is fo
rthc
omin
g).
We further detail home-buying capacity by population subgroup. Tables 15 through18 identify the number, type, and percentage of various categories of renter familieswho can afford the range of reference-priced units under the respective mortgageproducts. For instance, under the Historical Mortgage, approximately 1.314 millionrenter families (5.1 percent of the total 25.8 million renter families), including approx-imately 0.060 million black renters (1.3 percent of the total black renter families), canafford the median-priced house (table 16). With the application of the GSE StandardMortgage, approximately 1.631 million renter families (6.3 percent of the total), in-cluding approximately 0.071 million black renters (1.6 percent of the total), are ableto afford the median-priced house. Similar small increments of gain are achieved withthe more liberal mortgage instruments—the GSE Affordable, Governmental Afford-able, and Portfolio products. Thus, the most aggressive of the products consideredhere brings just over 2.4 million renter families (9.4 percent of the total), includingapproximately 0.154 million black renter families (3.4 percent of the total), to home-ownership at the median-priced reference point (table 16). Greater numbers of fami-lies are served at the less expensive reference-priced levels. For example, under theHistorical Mortgage, approximately 0.128 million black renters (2.9 percent of thetotal) can secure the low-priced house (table 18). That low-priced homeownership at-tainment reaches approximately 0.421 million black renter families (9.4 percent ofthe total) with the FHA 203(b)–prior mortgage.
A similar result is observed for Hispanic renter families (tables 15 through 18). Theirability to purchase a reference-priced home increases from the Historical Mortgageas the more liberal loan instruments are applied. For any given mortgage, equal orgreater numbers of Hispanic families qualify for homeownership at the less expensivereference prices. In general, the percentage of Hispanic families able to secure a homefor any given mortgage and reference-priced home combination is in order of magni-tude similar to or slightly less than that of black renters. For example, 0.4 percent ofHispanic renter families can afford the target-priced house with the Historical Mort-gage (compared with 1.0 percent of black renters); that share rises very slightly toabout 2 percent of Hispanic renters affording the target house (compared with approx-imately 3 percent of black renters) with the GSE Affordable and other more aggres-sive products. A maximum of about 7 percent of Hispanic30 renters can purchase alow-priced house with the most aggressive Governmental Affordable Mortgages—similar to but somewhat lower than the roughly 10 percent of black renters who canafford the low-priced home with these same most aggressive loans. Recent immigrantrenter families fare better, however, and come closest to the homeownership capabil-ity of white renter families. This last finding is consistent with the observed vigoroushomeownership gains among immigrants in recent years (Joint Center for HousingStudies 1998).
56 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
30 Hispanic renters’ home-buying capacity is similar but somewhat lower than that of their black counterparts,despite Hispanic renters having on average higher incomes and assets (table 10). We hypothesize that the His-panic home-buying capacity may be less than that of blacks for various reasons. First, the Hispanics’ target-pricedhomes are more expensive (table 10). Second, Hispanics, relative to blacks, are disproportionately concentrated inregions (e.g., California) with higher-priced criterion homes (e.g., the mean modestly priced house for Hispanicsis $77,049 compared with $61,802 for blacks).
We further break down reference home-buying capacity in table 19. This table countsonly the purchasing capacity of renter families specifically oriented to the target price.For example, if family A can afford a $200,000 house, only the purchase capacity ofthe target-priced house (about $90,000) would be counted in table 19. If all currentrenters were to seek a level of housing consumption similar to that attained by demo-graphically similar families who actually did purchase houses between 1993 and 1995,the Historical Mortgage would yield a total purchasing power of approximately $122billion. All of the mortgage market innovations allow dramatic increases in this im-pact: the GSE Standard guidelines yield a total impact of approximately $144 billion;the GSE Affordable Mortgages31 (including the Emerging GSE Affordable Mortgagegroup) realize a maximum of $166 billion in buying power; the Portfolio AffordableMortgages realize a maximum of $183 billion; and the Governmental Affordable Mort-gages realize a maximum of $233 billion in buying power (table 19).
We can extend the type of calculation just noted to further express the shortfall inrenters’ ability to become owners. We can do this by calculating the difference be-tween the price of the four reference homes (target, median-priced, modestly priced,and low-priced) and the value that each renter can afford (capped at the four refer-ence points)32 under the various mortgage instruments. This difference is termed thereference gap. The reference gap figures are shown in table 20.
Consider first the gap between (1) the sum total of houses for which renters can quali-fy based on each of the 15 mortgage instruments, and (2) the sum total of the targethouse prices for each renter. This affordability gap represents the shortfall betweenmortgage market innovations and the kinds of homes that were purchased by first-time buyers in the mid-1990s. Clearly, this gap is most pronounced for the HistoricalMortgage. Were all of the nation’s renter families to make the same choices made bydemographically similar families who purchased homes between 1993 and 1995,total housing demand would be in the $2.2 trillion to $2.3 trillion range (about 25.8million renter families, each seeking a target-priced home of about $90,000). However,the Historical Mortgage allows only a very low percentage of all renters to buy thetarget house, and the most they can afford is a fraction of the target house price (re-call that the mean uncapped home-buying capacity with the Historical Mortgage is$12,199 and the median is $0). The gap between what renters can afford with theHistorical Mortgage and the price of that house, summed for all renters, or the targethouse reference gap, is $2.2 trillion (table 20). Today’s more aggressive mortgagesreduce this gap somewhat, with the most flexible products considered here (e.g., theFlex 97, Community Gold, and Portfolio Composite mortgages) narrowing the gap toabout $2.0 trillion. Nevertheless, these staggering figures illustrate the enormousbarriers to homeownership and housing affordability among the nation’s renters. Evenvery aggressive mortgage market innovations are unable to address the low incomeand negligible savings of many renters.
Results: Homeownership Affordability 57
The Potential and Limitations of Mortgage Innovation
31 Without the 3/2 Option activated.
32 The absolute home-buying capacity shown in tables 12 and 13 did not cap the affordability.
One way to reduce that gap is to ratchet down the reference home from the higher-amenity target-priced unit to a less expensive home—in our current example, themodestly priced and low-priced houses. Table 20 shows that lowering housing expec-tation reduces the housing affordability gap by hundreds of billions of dollars butstill leaves a yawning chasm in the trillion-dollar range, with the exact amountvarying by product. For example, the reference gap on the low-priced house for theFreddie Mac Community Gold Mortgage stands at about $1.02 trillion.
Thus, while any gains are welcome, a large gap remains. Even very liberal mortgageproducts leave the vast majority of renters unserved at any level. Every mortgageproduct leaves at least 21 million renter families—or about 80 percent of the total—unable to enter homeownership at even the low-priced threshold. They are unservedas defined in this study. Less advantaged subgroups (e.g., blacks, Hispanics) fare evenworse.
58 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 19. Impact of Alternative Mortgage Instruments:Reference Home-Buying Capacity for All Renters, Nationwide with Target-Price Cap
Total National Average/Median
Home-Buying Home-Buying
Capacitya Capacityb
Loan Name (in $ Billions) Mean ($) Median ($)
Historical Mortgage 122.0 4,736 0
GSE Standard Mortgage (Fannie Mae and Freddie Mac) 144.1 5,594 0
GSE Affordable MortgagesFannie Mae
CHBP; CHBP, 3/2 Option (NA)c 161.7 6,278 0 CHBP, 3/2 Option (A)d 182.3 7,076 0 Fannie 97 165.7 6,430 0
Freddie MacAG; AG, 3/2 Option (NA)c 161.5 6,270 0AG, 3/2 Option (A)d 178.1 6,914 0 AG 97 160.7 6,238 0
Emerging GSE Affordable MortgagesFannie Mae
Flex 97 158.5 6,153 0 Freddie Mac
Community Gold 151.7 5,888 0
Portfolio Affordable MortgagesBank of America Zero Down 161.9 6,285 0 Bank of America Credit Flex 150.0 5,824 0 Portfolio Composite 182.9 7,100 0
Governmental Affordable MortgagesFHA 203(b)–prior 233.3 9,054 0 FHA 203(b)–current 193.7 7,519 0
Source: Authors’ analysis of 1993 SIPP data.a The aggregate value of houses that could be purchased by current renters above a low-priced house threshold.b For all current renters, not just just renters who can afford at least a low-priced house.c NA = 3/2 Option is not activated (i.e., 2 percent outside funding is not forthcoming).d A = 3/2 Option is activated (i.e., 2 percent outside funding is forthcoming).
Therefore, the crucial differences are in how the mortgages incrementally, albeit oftenmarginally, increase affordability for the renter population. Under the Historical Mort-gage, only 1.0 million to 2.6 million renter families of approximately 25.8 million totalrenter families could afford the various reference-priced homes. Each of the alterna-tive mortgages improves these figures slightly.
Results: Homeownership Affordability 59
The Potential and Limitations of Mortgage Innovation
Table 20. Reference Gap for All Renters, Nationwide by Mortgage Instrument
Reference Gapa (in $ Billions)
50th 25th 10thTarget Percentile Percentile Percentile
Loan Name House House House House
Historical Mortgage 2,159.5 2,406.8 1,615.8 1,077.5
GSE Standard Mortgage (Fannie Mae 2,103.3 2,359.8 1,580.8 1,052.9 and Freddie Mac)
GSE Affordable MortgagesFannie Mae
CHBP; CHBP, 3/2 Option (NA)b 2,048.4 2,311.1 1,544.1 1,028.0 CHBP, 3/2 Option (A)c 2,006.1 2,272.2 1,513.4 1,005.8 Fannie 97 2,059.6 2,319.5 1,548.4 1,029.5
Freddie MacAG; AG, 3/2 Option (NA)b 2,031.7 2,297.3 1,533.7 1,020.3 AG, 3/2 Option (A)c 1,988.0 2,257.3 1,502.5 997.9 AG 97 2,033.2 2,298.7 1,535.4 1,021.2
Emerging GSE Affordable MortgagesFannie Mae
Flex 97 2,044.4 2,308.1 1,542.7 1,027.3 Freddie Mac
Community Gold 2,026.6 2,293.9 1,532.8 1,019.9
Portfolio Affordable MortgagesBank of America Zero Down 2,026.9 2,291.7 1,530.7 1,019.6 Bank of America Credit Flex 2,024.6 2,292.4 1,532.0 1,019.4 Portfolio Composite 1,984.9 2,254.5 1,500.5 996.2
Governmental Affordable MortgagesFHA 203(b)–prior 1,941.8 2,209.2 1,458.8 964.7 FHA 203(b)–current 1,981.7 2,248.8 1,494.1 991.3
Source: Authors’ analysis of 1993 SIPP data.Note: The reference gap is largest for the 50th percentile reference category, even though the national median targethouse price exceeds the weighted median of the regional 50th percentile house prices. This somewhat counterintu-itive finding reflects the distribution of individual families according to their respective target house prices and geographicregions. The national medians for the 50th percentile and target house prices cannot be used to infer the underlyingdistribution of values that yield the aggregate reference gaps.a The difference between the price of the four reference homes (target-, median-, modestly, and low-priced) and the
value that each renter can afford (capped at the four reference points) summed across all renter families is the refer-ence gap.
b NA = 3/2 Option is not activated (i.e., 2 percent outside funding is not forthcoming).c A = 3/2 Option is activated (i.e., 2 percent outside funding is forthcoming).
Absolute and Reference Home-Buying Capacity by Family Income Level
Similar results are obtained when the analysis is applied by income level. Renters aresorted into four income categories: low income (0 percent to 49 percent of the medi-an family income); moderate income (50 percent to 79 percent of the median familyincome); middle income (80 percent to 119 percent of the median family income);and upper income (at least 120 percent of the median family income). To maintainsufficient sample size, we do not cross-tabulate by both income and race/ethnicity.
The mortgage simulation results by income category are detailed in appendix E andare summarized below.
1. As expected, home-buying capacity is positively correlated with income. For exam-ple, of the total $388 billion in absolute home-buying capacity under the GSEStandard Mortgage, the low-, moderate-, middle-, and upper-income renter fami-lies realize very different purchasing power—$2.8 billion, $9.8 billion, $28.3, and$346.8 billion, respectively (tables E.1 through E.4).
2. Homeownership affordability also tracks income in the sense that an increasingshare of renters is served at higher income levels. For example, 0.4 percent oflow-income renters can afford a modestly priced house with the GSE StandardMortgage, compared with 17.2 percent of their upper-income counterparts.
3. For any given income group, mortgage innovation increases homeownership af-fordability, but the gains are slightest for the least advantaged, who are simply,for the most part, too poor to purchase irregardless of mortgage product or refer-ence home price. For example, only 0.4 percent of low-income renters can afford tobuy a low-priced house with the Historical Mortgage. That rises, but barely so, toabout 0.9 percent with the most aggressive loan products. For moderate-incomerenters, the gain in the low-priced home-buying capacity rises from 2.5 percentof these renters, using the Historical Mortgage, to about 7 percent with the mostaggressive mortgage products considered in this study.
Once a threshold of home-buying capacity is reached, the gains from mortgageinnovation become more marked. For example, 6.6 percent of middle-income fam-ilies can afford the low-priced house with the Historical Mortgage. This percent-age doubles to 13.4 percent with mortgage innovation.
4. Stratifying by income reaffirms the overall finding of the mortgage simulationthat there is a tremendous challenge in bringing renters to homeownership. Forall income categories, the overwhelming share of renters remain unserved. Only15.4 percent of upper-income renters can afford the target-priced house with themost aggressive mortgage product. Even when the housing benchmark is ratch-eted down to a low-priced unit, only 35.7 percent of the upper-income renters areserved with the most liberal mortgage.
60 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Comparison with Previous Research
Our findings are similar to those of prior researchers. As we have conceptually mod-eled much of our work on and used the same database (SIPP) as Savage (1997, 1999),it is appropriate to compare our results to Savage’s findings. Our results differ slightlyfrom those of Savage because of differences in the following:
1. SIPP panels. In Who Can Afford to Buy a House in 1993 (1997) Savage combinedresults from wave four of the 1992 SIPP and wave seven of the 1991 SIPP (inter-views conducted between February and May 1993). We were able to access a laterpanel of the SIPP; we include results from the 1993 panel. In Who Can Afford toBuy a House in 1995 (Savage 1999), which was released after we had completedthe major portion of our investigation, Savage uses results from wave seven of the1993 SIPP.
2. Criterion homes. In Savage’s study (1997), the criterion-priced homes were basedon 1993 house values, differentiated by location (e.g., census regions and divisions).We incorporated these 1993 values in our investigation. Savage’s more recent WhoCan Afford to Buy a House in 1995 (Savage 1999) considers 1995 house values.
3. Interest rates. The earlier Savage study (1997) used a 7.55 percent conventionalmortgage contract interest rate; the rate is 8.79 percent in his latest study (1999).Our study was conducted during a time period between the two Savage studies,and, therefore, the rate we applied was 8.05 percent.
4. Closing costs. Savage (1997, 1999) calculates closing costs at about 3 percent; weestimate closing expenses at approximately 4 percent.
5. Income used for underwriting. We adjust income based on underwriting consider-ations such as income stability (table 10). This has the effect of reducing the in-come counted by underwriters for mortgage-qualifying purposes. (Note that theadjusted incomes in table 10 are almost always lower than the nominal income.)Savage did not make such an adjustment.
6. Definition of “family.” The definition of “family” used in this study is broader thanthat used by Savage. He defines a family as a group of two or more related personswho reside together, and he treats a person living alone as an “unrelated individ-ual.” In contrast, we include related individuals as well as a person living alone asa “family.” That, along with the timing difference noted above, results in differencesin the number of families counted; we counted 25.8 million families, comparedwith 21.4 million counted by Savage in his earlier study (1997) and 20.7 millioncounted in his later study (1999).
Given the above differences, we do not expect our results to mirror those of Savage.Nonetheless, given the overall commonality of database and approach, we expect ourinvestigations to yield similar outcomes. This order of magnitude parallelism is shownwhen we examine homeownership affordability for all renters (table 21).
Results: Homeownership Affordability 61
The Potential and Limitations of Mortgage Innovation
The overall similarity of outcomes is also evident when homeownership affordabilityof a modestly priced house is compared by renter category (e.g., race and Hispanicorigin; table 22).
We also find overall equivalent results when homeownership affordability by pricelevel is compared (see table 23). For example, Savage (1999) found that 69.9 percentof all renters could afford a house priced at less than $20,000 with an FHA mort-gage. Our figure is 69.1 percent. In short, Savage’s findings—that relatively few cur-rent renters (about 1 in 10) can afford even a modestly priced house, and that fewerthan 1 in 30 black or Hispanic renter families can realize modestly priced homeown-ership—are essentially reaffirmed in our mortgage simulations.
62 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 21. Percentage of All Renter Families Who Can Afford Variously Priced HousesUsing a GSE Standard Mortgage
House Price Savage (1997) Savage (1999) Listokin et al.
Target priced NA NA 5.0Median priced 8.6 6.7 6.3Modestly priced 11.7 9.9 9.2Low priced 15.1 12.8 12.1
NA = not applicable.
Table 22. Percentage of Renter Families by Type Who Can Afford a Modestly Priced HouseUsing a GSE Standard Mortgage
Renter Type Savage (1997) Savage (1999) Listokin et al.
All 11.7 9.9 9.2White (non-Hispanic) 14.2 14.4 13.1Black (non-Hispanic) 3.3 3.2 2.7Hispanic 4.1 3.3 1.8
Results: Homeownership Affordability 63
The Potential and Limitations of Mortgage Innovation
Tabl
e 23
.Co
mp
aris
on
of
Lis
toki
n e
t al
.’s a
nd
Sav
age’
s (1
999)
Aff
ord
abili
ty R
esu
lts
(Nu
mb
er a
nd
Per
cen
tag
e o
f R
ente
rs W
ho
Co
uld
Aff
ord
th
e In
dic
ated
Ho
use
Pri
ce)
Con
vent
iona
l Mor
tgag
eF
HA
Mor
tgag
e
List
okin
et a
l.*S
avag
e (1
999)
List
okin
et a
l.aS
avag
e (1
999)
Num
ber
Num
ber
Num
ber
Num
ber
Hou
se-P
rice
Cat
egor
y ($
)(in
Tho
usan
ds)
%(in
Tho
usan
ds)
%(in
Tho
usan
ds)
%(in
Tho
usan
ds)
%
Less
than
20,
000
21,3
08
82.7
17
,345
81.0
17,8
00
69.1
14
,977
69.9
20,0
00–2
9,99
962
7 2.
4 67
03.
11,
364
5.3
1,58
97.
430
,000
–39,
999
587
2.3
438
2.0
1,22
8 4.
8 89
54.
240
,000
–49,
999
337
1.3
351
1.6
806
3.1
532
2.5
50,0
00–5
9,99
938
6 1.
5 35
91.
760
9 2.
4 51
92.
460
,000
–69,
999
282
1.1
282
1.3
513
2.0
417
1.9
70,0
00–7
9,99
926
4 1.
0 20
91.
046
5 1.
8 32
11.
580
,000
–89,
999
271
1.1
265
1.2
421
1.6
363
1.7
90,0
00–9
9,99
918
3 0.
7 11
30.
535
4 1.
4 17
00.
810
0,00
0–12
4,99
955
4 2.
2 32
21.
563
7 2.
5 44
12.
112
5,00
0–14
9,99
925
1 1.
0 23
71.
146
0 1.
8 30
21.
415
0,00
0–19
9,99
930
7 1.
2 32
81.
550
7 2.
0 52
22.
420
0,00
0 or
mor
e40
6 1.
6 50
62.
460
1 2.
3 37
41.
7To
tal
25,7
63
100.
0 21
,425
100.
025
,765
10
0.0
21,4
2210
0.0
*A
pplie
s Li
stok
in e
t al
.’s d
ata
set
of 4
,109
ren
ter
fam
ilies
fro
m t
he 1
993
SIP
P t
o C
UP
R’s
mor
tgag
e pa
ram
eter
s, u
sing
the
Mor
tgag
e M
odel
.
EXPLANATION OF THE FINDINGS: WHY ARE RENTERS LIMITED
IN THEIR ABILITY TO REALIZE HOMEOWNERSHIP?
Why are so many renters, especially minority renters and LMI renters, unserved inthe mortgage market? Recall that in the current study lender bias is not a barrierbecause every renter is treated fairly; that is, all renters are treated solely on thebasis of their financial resources. This, however, is just the issue. Renters in general,and minority and LMI families in particular, have very modest financial resources,and this condition limits their home-buying capacity.
These financial limitations were noted earlier. To recap, renters as a group had a me-dian family income of only $18,786—about two-thirds of the $28,710 median incomeof all families (renters and homeowners) as reported in the SIPP. Black and Hispanicrenters trailed their majority counterparts considerably, with median family incomesof $13,770 and $15,314, respectively (table 10).
Disparities are further apparent when assets and debts are considered (table 10).White renters have mean assets of $11,368, and recent immigrants have mean assetsof $9,076; black and Hispanic renters have fractions of these resources, with meanassets of $1,601 and $2,000, respectively. Indeed, all renters have very limited assets,as indicated by the fact that the median level of assets for every subgroup consideredhere is about $500 or less.
With such a trace level of assets, even a 100 percent LTV mortgage will not facilitatehomeownership because of the resources required to meet substantial closing costs.Given the combination of modest income and assets, it is no wonder that home-buyingcapacity is a distant dream for many renter families, particularly minority families(Haurin, Hendershott, and Wachter 1997).
Which of the dual hurdles to homeownership is more significant—modest income ormodest assets? One way of addressing that issue is to first calculate the maximum-priced housing unit that could be purchased by traditionally underserved rentersgiven their actual characteristics. Next, one determines how much more these renterscould afford were they endowed, in turn, with the higher income and then the largerassets of their more advantaged white counterparts. We effect such an analysis below,focusing on a comparison of non-Hispanic white and black renters. The analysis ap-plies the GSE Standard Mortgage and the financial-demographic profiles of rentergroups detailed in table 10.
A black renter with an adjusted family income of $18,154, $1,601 in assets, and anaverage of $2,443 of debt could afford only a $50 maximum-priced house with a GSEStandard Mortgage. By contrast, the white renter counterpart, with an adjusted fam-ily income of $25,153, assets of $11,368, and debt of $4,247, could afford a $60,750maximum-priced home. The $50 versus $60,750 home-buying capacities reflect thevastly different existing financial conditions of white and black renters.
Explanation of the Findings 65
The Potential and Limitations of Mortgage Innovation
What if the black renter could hypothetically draw on, in turn, the enhanced incomeand assets of his or her white renter counterpart? Interestingly, tapping into the whiterenter’s greater income alone while retaining other existing financial conditions (e.g.,assets, debt) would not boost our example black renter’s home-buying capacity above$50. Income has no effect since the black renter’s assets are so limited. In contrast, ifthe black renter could retain the profile group’s average existing income and debt butdraw on the white renter’s greater assets, he or she would dramatically increase home-buying capacity from $50 to $52,650. The thousandfold increase from $50 to $52,650shows how important assets are to a renter’s ability to buy. Yet the fact that the blackrenter, even with the white renter’s assigned assets, can afford only a $52,650 house,as opposed to the $60,750 house affordable to the white renter, shows that otherfinancial constraints (e.g., lower income) play a role, albeit a less significant one, inthe home-buying equation.
Using the different mortgage instruments, we can more specifically trace the linksbetween the above financial attributes, the demands of purchasing various reference-priced homes, and constraints on homeownership. Table 24 attempts such an analysisby applying a GSE Standard Mortgage to a modestly priced (25th percentile) home.We begin with this combination because it is the one previously considered by Savage(1997, 1999).
Given the limited assets of renters, one would anticipate that a down payment con-straint—insufficient cash resources to cover the down payment and closing costs—would be a major hurdle to homeownership. Modest renter income also suggests anincome constraint—insufficient income to cover the monthly mortgage costs. Becauserenters are both income constrained and asset constrained, we anticipate a reasonablyhigh incidence of dual down payment and income constraints.
These anticipated findings are borne out in table 24. We find that 28 percent of allrenters who cannot afford the modestly priced home are constrained only by downpayment costs and that an additional 68 percent are dually constrained by the downpayment and income. Only 4 percent of all renters are limited solely by income. Thesefindings resemble those of Savage. The constraints by renter subgroup are furtherdetailed in table 24.
We began the constraints analysis with the modestly priced house/GSE Standardcombination because that was the scenario considered by Savage and we wished tocompare our findings with his. Yet renters aspire to their “dream house”—that is,our target-priced house. What are the homeownership constraints when the targetunit is the benchmark? These findings are shown in table 25 for two combinations:the target-priced house with a GSE Standard Mortgage and the target-priced housewith a CHBP mortgage. As anticipated, the number and percentage of renters ableto purchase the target house increase with the more liberal financing terms offeredby the CHBP. This comports with the findings traced earlier in this study. We can also
66 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
observe the constraints to affordability. A constraint by income alone is relatively rare.A down payment constraint alone or a dual down payment and income constraintpresents a greater hurdle. The same is true when the constraints to purchasing themodestly priced house are examined (see table 26).
Explanation of the Findings 67
The Potential and Limitations of Mortgage Innovation
Table 24. Constraints on Renter Group’sa Ability to Buy a Modestly Priced Home Using a GSE Standard Mortgage
Savage Listokin et al.
(1999) RecentAll All White Black Hispanic Other Immigrant
Constraint Families Families Families Families Families Families Families
Down paymentb 27.9 28.2 31.5 26.3 19.6 23.2 25.5Income 2.1 4.4 6.6 0.8 1.2 2.0 3.2Down payment and income 69.9 67.5 62.0 72.9 79.2 74.8 71.3Totalc 100.0 100.0 100.0 100.0 100.0 100.0 100.0
Source: Savage 1999 and authors’ analysis of 1993 SIPP data.Note: The table categorizes those who cannot afford the 25th percentile home in their census region according toconstraint. There are three reasons why a household might not be able to afford the home: (1) insufficient income tocover the monthly mortgage costs; (2) insufficient cash resources to cover the down payment and closing costs; and(3) too much debt. Savage (1997) conflates down payment and excessive debt into a single category (insufficientdown payment), and we mostly follow that approach in reproducing his results.
To determine if a household has too little income to buy the reference home, we took the monthly housing expendi-ture (monthly income times the front-end ratio for the given mortgage instrument) and determined the net presentvalue (NPV) of that payment over 30 years (using an interest rate of 8.05 percent for all instruments). If the NPV wasless than 95 percent of the value of the reference house, we determined that the household was income constrained.This calculation was performed without regard to the creditworthiness of the household or the availability of cash tocover the down payment and closing costs.
To determine if a household had insufficient cash to cover the down payment and closing costs, we took the family’sassets and subtracted the down payment (5 percent times the value of the house as given by the maximum LTV ratio)and the estimated closing costs for the housing market. This differs slightly from Savage’s analysis in that our closingcosts are estimated based on the particulars of the housing market (at least for the eight MSAs for which we researchedlocal closing costs.) Cash available for a down payment was estimated without regard to a family’s debts. For instance,if a family had $5,000 in assets and $2,500 in debt, we used $5,000 (and not assets less debt) as the cash availablefor down payment.
a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,they sum to all renter families. Recent immigrant families, those entering the United States after 1984, overlap withthe previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted, whites,blacks, and other groups are non-Hispanic.
b Savage also includes those constrained by excessive debt in this category.c Figures may not add to indicated totals because of rounding.
68 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 25. Comparison of the GSE Standard Mortgage with the Fannie Mae CHBP with Two Policy Options, Target-Priced House
(Number and Percentage of Renter Families Who Can Afford a Target-Priced House with the Indicated Mortgage)
CHBP with $100Monthly Housing CHBP with
Fannie Mae Income $5,000 DownAffordability of a Target House GSE Standard CHBP Supplement Payment Voucher
Can afford to buy, number (%) 1,288,186 (5.0) 1,655,782 (6.4) 1,862,248 (7.2) 3,538,367 (13.7)
Constrained by income only, 1,487,092 (5.8) 1,399,314 (5.4) 1,192,850 (4.6) 5,177,856 (20.1)number (%)
Constrained by down payment 3,989,196 (15.5) 5,041,003 (19.6) 6,648,642 (25.8) 3,277,084 (12.7)only, number (%)
Constrained by income and 18,998,464 (73.7) 17,666,839 (68.6) 16,059,070 (62.3) 13,769,630 (53.4)down payment, number (%)
Total, number (%) 25,762,939 (100) 25,762,939 (100) 25,762,939 (100) 25,762,939 (100)Sample size, number 4,109 4,109 4,109 4,109
NPV of policy option 858.5a 9,413b
(in $ millions)
Note: Figures may not add to indicated totals because of rounding.a The NPV of the income supplement assumes a supplement of $100 per month for 48 months at a 6.0 percent dis-
count rate and that the supplement is applied to only the additional qualifying applicants.b The NPV of the down payment voucher assumes that a lump sum payment of $5,000 is given to only the additional
qualifying applicants.
Table 26. Comparison of the GSE Standard Mortgage with the Fannie Mae CHBP with Two Policy Options, Modestly Priced House
(Number and Percentage of Renter Families Who Can Afford a Modestly Priced House with the Indicated Mortgage)
CHBP with $100Monthly Housing CHBP with
Affordability of a Fannie Mae Income $5,000 DownModestly Priced House GSE Standard CHBP Supplement Payment Voucher
Can afford to buy, number (%) 2,381,156 (9.2) 2,835,612 (11.0) 3,057,030 (11.9) 7,560,159 (29.3)
Constrained by income only, 1,027,692 (4.0) 895,107 (3.5) 673,649 (2.6) 5,339,348 (20.7)number, (%)
Constrained by down payment 6,582,168 (25.5) 7,486,227 (29.1) 9,418,673 (36.6) 2,971,919 (11.5)only, number (%)
Constrained by income and 15,771,923 (61.2) 14,545,992 (56.5) 12,613,535 (49.0) 9,891,513 (38.4)down payment, number (%)
Total, number (%) 25,762,939 (100) 25,762,939 (100) 25,762,939 (100) 25,762,939 (100)Sample size, number 4,109 4,109 4,109 4,109
NPV of policy option 920.7a 23,623b
(in $ millions)
Note: Figures may not add to indicated totals because of rounding.a The NPV of the income supplement assumes a supplement of $100 per month for 48 months at a 6.0 percent dis-
count rate and that the supplement is applied to only the additional qualifying applicants.b The NPV of the down payment voucher assumes that a lump sum payment of $5,000 is given to only the additional
qualifying applicants.
POLICY CONSIDERATIONS
This section examines various policies to enhance the capacity of renters to achievehomeownership. In doing so, we recognize that some researchers assert that there isalready too much support for, and subsidization of, homeownership (Krueckeberg1999). Further, even those who are generally supportive of homeownership may verywell propose that this tenure is not suitable for everyone and that by emphasizinghomeownership we may very well not pay enough attention to other pressing hous-ing needs, such as the need for new rental housing and maintenance of the existingrental housing stock. One could also argue that the government’s job in the marketshould be limited to guarding against discrimination as opposed to equalizing the con-sumption of homeownership. We recognize these very valid points. For the sake ofdiscussion, however, we shall proceed under the assumption that homeownership isan important good deserving of public support.
A critical strategy to foster homeownership is to address renters’ financial constraints(Eller and Fraser 1995; Engelhardt and Mayer 1994). Because so many renters areconstrained by down payment costs, whether solely or in conjunction with income con-straints, one reasonable intervention would be to supplement the assets available torenters (Mayer and Engelhardt 1996). This could take the form of a voucher to beapplied for down payment and closing costs. An income supplement, say a voucher topay for mortgage costs, would be a way of compensating for renters’ modest incomes.The cost of both of these programs would depend on the magnitude and duration ofthe supplements.
Table 25 poses hypothetical examples of both of these approaches to addressing finan-cial constraints to homeownership. One is a $5,000 down payment voucher to be usedfor down payment or other up-front costs (e.g., closing expenses). The other is a hous-ing payment supplement of $100 per month, or $1,200 yearly. It is anticipated thatthis voucher would be in force for four years, amounting to $4,800 in cumulative pay-ments. Although the voucher would cease after four years, it is anticipated that therenter’s income would rise sufficiently over this period to compensate for the end ofthe voucher supplement. (The same reasoning underlies the concept of the adjustablerate mortgage, where the initial rate is pegged below the market rate.)
The different approaches address different constraints. As shown in table 25, the downpayment voucher has the biggest effect on reducing the share of renters who are con-strained by down payment costs with respect to purchasing the target-priced house;the income voucher targets the income constrained. Since down payment constraintsloom so large as a barrier to renters, it is not surprising that the down payment vouch-er dramatically enhances home-buying capacity. Approximately 1.7 million renterfamilies (6.4 percent of all renters) can afford the target-priced house with the CHBPmortgage. When the down payment voucher is added to the CHBP financing, 3.5million renter families (13.7 percent of all renters) can afford target-priced home-ownership. The increase in affordability is much lower with the income voucher/CHBPcombination, which brings about 1.9 million renter families, or approximately 7.2percent of the total, to homeownership.
Policy Considerations 69
The Potential and Limitations of Mortgage Innovation
Not surprisingly, these outcomes have different costs. The down payment voucher isa potent tool. Yet its cost in net present value terms—approximately $9.4 billion—far exceeds the approximate $0.86 billion expenditure of the income voucher. Thisdisparity in cost is due to two factors. First, the down payment voucher moves manymore renters into homeownership than the income voucher does. Second, the formeris an up-front payment, making it more expensive in time-value terms than the lat-ter, which is spread over four years.
Similar results are observed when the policy options are applied to the modestlypriced house (table 26). Approximately 2.8 million renters (11.0 percent of the total)can afford this home with the CHBP mortgage alone. That number increases to ap-proximately 3.1 million renters (11.9 percent of the total) with the income supplementand 7.6 million renters (29.3 percent of the total) with the down payment assistance.Clearly, supplementing assets is a more powerful way to enable renters to afford themodestly priced house. However, in this instance, the present value of the asset sup-plement, $23.6 billion, is more than 25 times greater than the present value cost ofthe income supplement ($0.9 billion).
In practice, an asset supplement approach underlies the design of the 3/2 Option inthe GSE Affordable Mortgages. Thus far, we have discussed the affordability outcomesusing Fannie Mae’s CHBP and Freddie Mac’s AG without consideration of each pro-gram’s 3/2 Option (i.e., 3 percent minimum borrower contribution to the down pay-ment; 2 percent additional amount allowed from gifts and grants). We maintainedthis restriction because we wanted to determine the renters’ homeownership poten-tial based solely on their own resources. What happens if the 2 percent gift or grantis considered? The gift or grant is, in effect, an asset supplement and should increasethe attainment of homeownership. We simulate the impact of the 3/2 Option in tables12 through 20. This option increases the absolute and the reference home-buyingcapacities. For instance, while 11.4 percent of all renter families (2.945 million) canafford the modestly priced house with an AG mortgage without the 3/2 Option acti-vated, this proportion rises to 12.9 percent (3.323 million renter families) with the3/2 Option in force (table 17).
Given that an asset transfer can boost homeownership, it is important to measurethe aggregate asset gap. Notwithstanding renters’ income constraints, what is thetotal amount of assistance needed to address the shortfall in resources required forthe down payment and closing costs? We calculate this shortfall, which we term theasset gap, by mortgage type for the four reference-priced houses (table 27). The assetgap for the target house and GSE Standard Mortgage is approximately $182 billion.This figure decreases to approximately $105 billion when the same mortgage is usedbut applied to the low-priced house. The asset gap for the target house changes whendifferent mortgages are applied: the asset gap is approximately $287 billion with theHistorical Mortgage; $182 billion with the GSE Standard Mortgage; as low as $121billion with the GSE Affordable Mortgages33 (including the Emerging GSE Affordable
70 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
33 With the 3/2 Option activated.
loans); and less than $90 billion for the very aggressive Portfolio Affordable and Gov-ernmental Affordable Mortgages. The above amounts are the resources needed toaddress the asset shortfall. Even with that supplement, renters could remain incomeconstrained, yet meeting the shortfall in assets would go a long way toward enablingrenters to become homeowners.
In short, although mortgage innovation expands homeownership potential, otherstrategies, such as income and asset supplements, are needed to address the financialconstraints faced by the vast majority of renters. In fact, there is a larger universe ofpolicy options that we illustrate for the modestly priced house and the target-pricedhouse in tables 28 and 29, respectively. Each of these tables starts with a baseline: thepercentage of renters who can afford the reference house with the GSE StandardMortgage. Each table then presents four generic approaches to improving the home-ownership potential. The first, mortgage term adjustment, relaxes the mortgage terms
Policy Considerations 71
The Potential and Limitations of Mortgage Innovation
Table 27. Asset Gap by Mortgage Type and Reference-Priced House
Asset Gapa (in $ Billions)
50th 25th 10thTarget Percentile Percentile Percentile
Loan Name House House House House
Historical Mortgage 287.4 319.4 221.7 155.1
GSE Standard Mortgage (Fannie Mae 182.3 201.6 144.1 105.2 and Freddie Mac)
GSE Affordable MortgagesFannie Mae
CHBP; CHBP, 3/2 Option (NA)b 161.5 181.2 124.5 86.3 CHBP, 3/2 Option (A)c 121.2 135.4 94.3 66.7 Fannie 97 131.7 145.9 104.2 76.0
Freddie MacAG; AG, 3/2 Option (NA)b 161.5 181.2 124.5 86.3 AG, 3/2 Option (A)c 121.2 135.4 94.3 66.7 AG 97 134.9 149.0 107.2 78.8
Emerging GSE Affordable MortgagesFannie Mae
Flex 97 133.7 147.9 106.1 77.8 Freddie Mac
Community Gold 136.1 150.2 108.3 79.9
Portfolio Affordable MortgagesBank of America Zero Down 87.3 93.2 73.4 60.0 Bank of America Credit Flex 136.5 150.6 108.7 80.3 Portfolio Composite 121.4 135.7 94.5 66.8
Governmental Affordable MortgagesFHA 203(b)–prior 87.5 102.9 65.9 41.2 FHA 203(b)–current 103.3 115.0 80.9 58.0
Source: Authors’ analysis of 1993 SIPP data.a Notwithstanding income constraints, the total amount of assets needed to address the shortfall in resources required
to buy the respective reference-priced houses.b NA = 3/2 Option is not activated (i.e., 2 percent outside funding is not forthcoming).c A = 3/2 Option is activated (i.e., 2 percent outside funding is forthcoming).
by (1) reducing the interest rate from the market level, (2) reducing the required downpayment, and/or (3) reducing the PMI requirement.
The second generic approach, transaction-carrying cost adjustment, reduces thetransaction-carrying cost of homeownership by allowing for (1) closing-cost savings,(2) property tax savings, and/or (3) hazard insurance savings. In all instances, the
72 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 28. Effects of Policy Options by Renter Group:a Percentage of RentersNationwide Who Can Afford a Modestly Priced House
RecentAll White Black Hispanic Other Immigrant
Renters Renters Renters Renters Renters RentersBaseline/Option (%) (%) (%) (%) (%) (%)
BaselineModestly priced house, 9.2 13.1 2.7 1.8 6.1 8.4
GSE Standard Mortgage
OptionsI. Mortgage term adjustment
A. Interest rate reductionb
5% 10.0 14.3 2.7 2.1 6.1 9.12.5% 10.6 15.1 2.7 2.1 6.7 9.90% 11.2 16.0 3.0 2.1 7.3 10.6
B. Down payment reduction3% down 9.8 13.9 2.8 2.0 6.1 8.40% down 11.3 15.6 4.0 2.4 10.2 11.2
C. Reduced PMIc
0% 9.4 13.3 2.7 2.0 6.1 9.1II. Transaction-carrying cost adjustment
A. Closing-cost savingsLowest closing costs 9.6 13.6 2.8 2.0 6.6 8.4
B. Property tax savingsLowest property tax 9.4 13.3 2.7 1.8 6.1 8.4
C. Reduced hazard insuranceLowest insurance 9.3 13.1 2.7 2.0 6.1 8.4
III. House price adjustment–10% 9.9 14.1 2.7 2.0 6.1 9.1
IV. Borrower financial supplementA. Annual income supplement
+$1,000 9.9 14.2 2.7 2.0 6.1 9.1+$5,000 12.2 17.3 3.3 2.6 8.0 11.4+$10,000 13.1 18.6 3.5 3.0 8.0 11.4
B. Asset supplement+$1,000 9.9 13.8 3.0 2.3 6.7 8.4+$5,000 16.2 21.6 8.5 3.8 12.9 16.5+$10,000 35.6 41.8 29.8 20.1 24.7 33.2
Source: Authors’ analysis of 1993 SIPP data.a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,
they sum to all renter families. Recent immigrant families—those entering the United States after 1984—overlapwith the previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted,whites, blacks, and other groups are non-Hispanic.
b Current contract interest rate is 8.05 percent.c For periodic or recurring (not up-front) PMI payment.
savings are based on the lowest closing costs, property taxes, and hazard insuranceexpenses in the eight metropolitan areas for which we obtained such data.
The third generic area of adjustment, housing price, simply reduces the cost of thehousing unit. This is an extension of the reference-priced gradation considered inthis study.
Policy Considerations 73
The Potential and Limitations of Mortgage Innovation
Table 29. Effects of Policy Options by Renter Group:a Percentage of RentersNationwide Who Can Afford a Target-Priced House
RecentAll White Black Hispanic Other Immigrant
Renters Renters Renters Renters Renters RentersBaseline/Option (%) (%) (%) (%) (%) (%)
BaselineTarget-priced house, 5.0 7.0 1.5 1.0 3.9 4.6
GSE Standard Mortgage
OptionsI. Mortgage term adjustment
A. Interest rate reductionb
5% 5.9 8.3 1.6 1.1 4.4 5.62.5% 6.9 9.8 1.7 1.3 6.0 7.50% 7.6 10.8 1.7 1.4 6.6 8.3
B. Down payment reduction3% down 5.6 7.8 1.8 1.3 3.9 4.60% down 6.3 8.6 2.5 1.7 5.4 4.6
C. Reduced PMIc
0% 5.2 7.4 1.5 1.0 4.4 4.6II. Transaction-carrying cost adjustment
A. Closing-cost savingsLowest closing costs 5.4 7.5 1.6 1.3 4.4 4.6
B. Property tax savingsLowest property tax 5.3 7.5 1.5 1.0 3.9 4.6
C. Reduced hazard insuranceLowest insurance 5.0 7.1 1.5 1.0 4.4 4.6
III. House price adjustment–10% 5.7 8.1 1.6 1.0 4.4 5.6
IV. Borrower financial supplementA. Annual income supplement
+$1,000 5.5 7.8 1.5 1.0 4.4 4.6+$5,000 8.2 11.6 1.7 1.7 7.3 9.0+$10,000 9.9 14.0 2.3 2.1 7.9 10.5
B. Asset supplement+$1,000 5.4 7.6 1.8 1.2 3.9 4.6+$5,000 8.4 11.1 3.7 2.8 8.3 6.8+$10,000 19.4 22.3 14.5 13.6 18.1 20.8
Source: Authors’ analysis of 1993 SIPP data.a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,
they sum to all renter families. Recent immigrant families—those entering the United States after 1984—overlapwith the previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted,whites, blacks, and other groups are non-Hispanic.
b Current contract interest rate is 8.05 percent.c For periodic or recurring (not up-front) PMI payment.
The fourth approach, borrower financial supplement, supplements the borrower’s fi-nances by adding income or adding assets. In the former option, we assume a varyingannual income supplement of $1,000, $5,000, or $10,000. In the latter option, we fac-tor a one-time asset transfer of $1,000, $5,000, or $10,000.
The improvements in homeownership potential from the respective baselines are in-dicated by the rising percentage of renters who can afford the modestly priced andtarget-priced homes. In all instances, the greatest gains are realized by the borrowerfinancial supplement approach. That is followed by the mortgage term adjustmentapproach. A more modest improvement results from adjusting the transaction-carry-ing costs and the house price.
To illustrate, 9.2 percent of all renters can afford a modestly priced house with theGSE Standard Mortgage. That share rises to 12.2 percent with a $5,000 income sup-plement; 16.2 percentage with a $5,000 asset supplement; and 11.2 percent with amortgage contract interest rate reduction from 8.05 percent to 0 percent (table 28).The baseline share of black renters who can afford a modestly priced home with theGSE Standard Mortgage is 2.7 percent. The only way to bring that share up appre-ciably is through an asset supplement. A $10,000 annual income supplement boostsblack renters’ ability to realize modestly priced homeownership by about 1 percent(from 2.7 to 3.5 percent); however, a $10,000 asset supplement would allow 29.8 per-cent of black renters to purchase the modestly priced home. This would essentiallybring black renters to parity with all renters, although they would still trail theirwhite counterparts.
In a similar vein, the affordability rate for Hispanics for a modestly priced home canbe increased to 20.1 percent (from a baseline share of 1.8 percent) with a $10,000asset supplement. Other options have a measurable but less significant impact onHispanic renters. For instance, a $10,000 income supplement allows only 3.0 percentof Hispanic renters to purchase the modestly priced unit. The impact of the variousoptions on the affordability rate follows a similar pattern for recent immigrants. The$10,000 asset transfer would bring approximately 33 percent of recent immigrants tohomeownership; a $5,000 asset supplement to this group would have half that impact(16.5 percent would be able to afford the modestly priced house). Finally, note thatthe results are similar when the options are applied to a target-priced house. A$10,000 income supplement would allow approximately 10 percent of all renters topurchase a target-priced home (doubling the 5 percent reach of the unsubsidized GSEStandard Mortgage); a $10,000 asset supplement would allow approximately 20 per-cent of all renters to realize target-priced homeownership (table 29).
We also can calculate how the policy options can affect absolute home-buying capac-ity (table 30). Using the GSE Standard Mortgage as the baseline, all renters haveapproximately $388 billion in aggregate purchasing power. With the lowest closingcosts, the absolute home-buying capacity is boosted by $25 billion (to approximately$413 billion). A $10,000 income supplement to each renter increases buying power toapproximately $631 billion; a $10,000 asset supplement boosts the purchasing abilityby about $890 billion (to approximately $1.280 trillion). The same $10,000 asset sup-
74 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
plement to each black renter would increase purchasing power about ninefold, from$17 billion (baseline) to $150 billion (table 30).
Tables 28 through 30 show the impact of item-by-item changes on homeownership af-fordability. Combining the items will magnify the increase in affordability and pur-chasing power. The three tables thus provide an informative matrix for policy consid-eration, but they do not exhaust the universe of possible changes. For example, we
Policy Considerations 75
The Potential and Limitations of Mortgage Innovation
Table 30. Effects of Policy Options by Renter Group:a
Absolute Home-Buying Capacity, Nationwide (in $ Millions)Recent
All White Black Hispanic Other ImmigrantRenters Renters Renters Renters Renters Renters
Baseline/Option (%) (%) (%) (%) (%) (%)
BaselineGSE Standard Mortgage 387,695 347,340 16,702 13,422 10,231 10,724
OptionsI. Mortgage term adjustment
A. Interest rate reductionb
5% 435,440 390,564 18,604 14,534 11,739 12,078 2.5% 487,135 435,904 21,015 17,030 13,186 13,057 0% 546,219 488,388 23,578 19,008 15,245 14,230
B. Down payment reduction3% down 414,703 370,026 19,289 14,872 10,516 11,929 0% down 472,154 409,383 27,865 17,960 16,945 12,908
C. Reduced PMIc
0% 395,112 354,105 16,935 13,629 10,442 10,879 II. Transaction-carrying cost adjustment
A. Closing-cost savingsLowest closing costs 413,291 369,075 19,018 14,377 10,821 11,420
B. Property tax savingsLowest property tax 408,611 366,486 17,232 14,335 10,559 11,663
C. Reduced hazard insuranceLowest insurance 390,833 350,143 16,802 13,581 10,307 10,790
III. House price adjustment–10% 437,327 391,979 18,847 14,914 11,587 12,616
IV. Borrower financial supplementA. Annual income supplement
+$1,000 416,130 372,711 17,734 14,584 11,101 11,875 +$5,000 528,221 473,407 22,750 18,037 14,027 14,203 +$10,000 630,669 566,109 25,818 21,790 16,952 16,376
B. Asset supplement+$1,000 409,772 366,447 18,251 14,444 10,630 11,748 +$5,000 712,743 578,016 65,976 43,441 25,310 19,592 +$10,000 1,280,300 942,716 149,667 138,432 49,499 34,775
Source: Authors’ analysis of 1993 SIPP data.a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,
they sum to all renter families. Recent immigrant families—those entering the United States after 1984—overlapwith the previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted,whites, blacks, and other groups are non-Hispanic.
b Current contract interest rate is 8.05 percent.c For periodic or recurring (not up-front) PMI payment.
observe from our mortgage simulations that products with a back-end ratio but nofront-end ratio have greater potential to expand homeownership. Therefore, a policythat eliminates the front-end ratio (while keeping the back-end ratio) would enablemore renters to buy a house.
There are numerous policy options to foster homeownership that go beyond the mort-gage innovations and resource supplements considered thus far. A HUD (1995b) studylisted 100 actions to further the National Homeownership Strategy, and other litera-ture has considered a range of policies (DiPasquale 1990). For example, homeowner-ship counseling can guide households in repairing credit problems and assemblingassets to finance down payments and closing costs. Savings vehicles, such as indi-vidual development accounts, can be part of a process that helps accumulate assetsfor homeownership. Outreach is also important so that the traditionally underservedavail themselves of the more flexible mortgage instruments and do not fall victim topredatory home lending practices.
Assessments of policy options must also consider costs. The net present value (NPV)expense of the respective options is a critical consideration. Of the different policiespresented in tables 28 through 30, the most potent in terms of enhancing the absoluteand relative home-buying capacities were the income and asset supplements, and,therefore, we should consider their costs in NPV terms. In making this calculation, weassume that the $1,000, $5,000, and $10,000 asset supplements are one-time infusionsand that the $1,000, $5,000, or $10,000 income supplements are given annually andare maintained for varying periods of time, from 4 to 12 years, depending on theamount (a shorter duration for the lesser amounts and a longer period for the largersupplements).34 After the 4 to 12 years, it is assumed that the income supplementwould cease but that the renter’s income would rise sufficiently over this period tocompensate for the termination of the outside assistance.
Table 31 shows the NPV costs of the $1,000, $5,000, and $10,000 income and assetsupplements, based on the assumptions described above. The costs vary by the indi-cated home price and mortgage combinations for several reasons, including the factthat the number of renters able to realize homeownership varies with each option.
In short, supplementing income and, especially, assets can dramatically enhancehome-buying capacity. Both options are expensive, and the more potent asset sup-plement (i.e., $5,000 and $10,000) is especially expensive. Policy makers must makedecisions about the trade-offs. Providing a $10,000 asset supplement would raise theshare of renters able to afford a modestly priced house with a GSE Standard Mort-
76 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
34 We make the following assumptions: the $1,000 annual income supplement is provided for 4 years; the $5,000annual income supplement is provided for 8 years; and the $10,000 annual income supplement is provided for12 years. We assume an 8-year period for the $5,000 income supplement, as opposed to the 4-year period usedwhen we previously calculated the effect of a $5,000 voucher with the CHBP (tables 25 and 26) because here theincome supplements are added specifically to the earnings of traditionally underserved renters who have mod-est incomes. Consequently, it is reasonable to assume that it will take these renters a considerably longer periodof time to increase their incomes by the indicated amounts (e.g., $5,000) through their own resources.
gage from 9.2 percent to 35.6 percent. The cost of this intervention, however—a stag-gering $68 billion—must be balanced against the economic and noneconomic benefits.These benefits include the interest on larger mortgages paid to investors, the increaseddemand for real estate and home improvement services, and a variety of alleged soci-ological benefits associated with stable neighborhoods in which owners have a long-term stake. Scholars and policy makers remain deeply divided on the question ofwhether public policy should attempt to equalize homeownership. We would simplypoint out that the economic and sociological benefits are routinely invoked to justifythe current subsidy of middle-class homeownership. Our calculations simply indicatean explicit price tag if a decision is made to expand these benefits to all renters, espe-cially the underserved.
Many other costs need to be considered; for example, the innovative mortgage productshave long-term loan performance implications (Calem and Wachter 1996; Quercia andStegman 1992). Although FHA qualifies many people for homeownership because ofits generous mortgage terms, its delinquency rate has historically been much higherthan that of more restrictive conventional mortgages (Berkovec et al. 1997). Innova-tive mortgage products also carry greater risk because they allow increasing sharesof income to be applied for housing purposes. A high apportionment of income forhousing may not leave much leeway to meet household medical and other unexpectedexpenses. A high housing debt ratio may also be imperiled by future changes, suchas rising utility costs or a downturn in the economy that raises unemployment ratesand creates other adverse effects. For these and other reasons, the innovative mort-gage products may encounter higher delinquency rates than those encountered byconventional mortgages—which is surely a cost. Fully quantifying the costs of themany possible policy options, however, is beyond the scope of the current investigation.
In considering policy options, it is also instructive to consider the potential and costsof bringing to homeownership only those renters who are most “suitable” for or “ori-
Policy Considerations 77
The Potential and Limitations of Mortgage Innovation
Table 31. NPV Cost of Income and Asset Supplement Policy Options(Applied to All Renters)
Modestly Priced Target HouseHouse with a with a
GSE Standard GSE StandardMortgagea Mortgageb
Policy Option ($ Millions) ($ Millions)
Annual income supplement+$1,000 154 866+$5,000 2,980 3,881+$10,000 7,008 6,987
Asset supplement+$1,000 164 109+$5,000 8,939 4,366+$10,000 67,920 37,124
a See table 28 for details of the effects of each policy option.b See table 29 for details of the effects of each policy option.
78 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
ented” to that tenure (hereinafter referred to as homeowner-oriented renters). To thispoint in the analysis, we have looked at the ability and costs of all renters to realizehomeownership because that is an instructive benchmark. In fact, homeownership isnot sought by all renters; there are lifestyle renters and others who rent by choice(Goodman 1999; Varady and Lipman 1994). In addition, homeownership is an expen-sive good that is often viewed as far beyond the means of the most financially con-strained. Removing the lifestyle and the very poor renters from the all-renters groupyields as a remainder the homeowner-oriented renters.
It is difficult to precisely demarcate the homeowner-oriented target population. Varadyand Lipman (1994), Goodman (1999), and others have observed that both the youngand the old often prefer to rent. Varady’s lifestyle and elderly life-cycle renters, for ex-ample, had average ages of 55.5 and 60.8, respectively. Researchers also note thatlifestyle renters are frequently the most affluent of renters; they have the means tobuy but opt to rent. Goodman’s lifestyle market renters had a mean income of almost$65,000. Renters drawn to reviving urban locations in the New York metropolitan areaoften have six-figure incomes (Roseland Property Co. 1999). In addition, although thereis little consensus as to who is too poor to buy, it is not unrealistic to assume a $15,000to $20,000 income as the realistic earnings floor for achieving homeownership.
Based on these considerations, we demarcate the homeowner-oriented renters asthose between 25 and 55 years of age with family earnings of at least $15,000 butless than $80,000. We recognize that this delineation is rough at best, but it likelydifferentiates on an order-of-magnitude basis those renters who are more desirous of,and more suitable for, homeownership. What then is the home-buying capacity ofthe homeowner-oriented renters?
We shall not repeat the entire simulation for the homeowner-oriented group, but wecan obtain a glimpse of the results in tables 32 and 33. Comparing the data on thehomeowner-oriented group (table 32) with the data on the all-renters group (table 12)indicates that, overall, a higher share of the homeowner-oriented renters can purchasethe reference-priced homes. The increase is due to the deletion of renters earningbelow $15,000. The gain is slight, however, because the homeowner-oriented grouphas also dropped renters with earnings above $80,000.
The policy option results (table 33) indicate that for homeowner-oriented renters, asfor all renters (table 28), the most dramatic gains in homeownership potential areachieved through income and asset supplements, especially the latter. For example,3.8 percent of black homeowner-oriented renters can afford a modestly priced housewith a GSE Standard Mortgage. With the assistance of the $5,000 and $10,000 assetsupplement, the percentage of black homeowner-oriented renters who can afford amodestly priced home increases to 15.5 percent and 64.6 percent, respectively. Asnoted before, however, the NPV costs for these policy options are quite high. Table 34shows the NPV costs for the income and asset supplement policy options where ap-plied to the homeowner-oriented renters group.
Policy Considerations 79
The Potential and Limitations of Mortgage Innovation
Tabl
e 32
.Im
pac
t o
f A
lter
nat
ive
Mo
rtg
age
Inst
rum
ents
:A
bso
lute
an
d R
efer
ence
Ho
me-
Bu
yin
g C
apac
ity
for
Ho
meo
wn
er-O
rien
ted
Ren
ters
,Nat
ion
wid
e
Ref
eren
ce H
ome-
Buy
ing
Cap
acity
(Per
cent
age
of R
ente
rs W
ho C
ould
Abs
olut
e H
ome-
Buy
ing
Cap
acity
Affo
rd th
e In
dica
ted
Hou
se)
Tota
l Nat
iona
lA
vera
ge/M
edia
n50
th25
th10
thH
ome-
Buy
ing
Cap
acity
aH
ome-
Buy
ing
Cap
acity
bTa
rget
Per
cent
ileP
erce
ntile
Per
cent
ileLo
an N
ame
(in $
Bill
ions
)M
ean(
$)M
edia
n($)
Hou
seH
ouse
Hou
seH
ouse
His
toric
al M
ortg
age
133.
811
,646
0
3.7
5.4
7.8
10.8
GS
E S
tand
ard
Mor
tgag
e (F
anni
e M
ae
175.
215
,248
0
5.3
6.6
10.2
13.7
and
Fred
die
Mac
)
GS
E A
fford
able
Mor
tgag
esFa
nnie
Mae
C
HB
P;C
HB
P, 3
/2 O
ptio
n (N
A)c
224.
719
,553
0
7.2
8.6
12.7
17.0
CH
BP,
3/2
Opt
ion
(A)d
261.
522
,763
0
8.4
10.3
14.9
19.9
Fann
ie 9
720
9.4
18,2
25
0 6.
57.
712
.716
.9Fr
eddi
e M
acA
G;A
G, 3
/2 O
ptio
n (N
A)c
240.
520
,932
0
7.8
9.2
13.2
17.1
AG
, 3/2
Opt
ion
(A)d
280.
124
,377
0
9.2
10.9
15.6
20.0
AG
97
241.
521
,016
0
7.9
9.3
13.4
16.8
Em
ergi
ng G
SE
Affo
rdab
le M
ortg
ages
Fann
ie M
aeFl
ex 9
723
0.9
20,0
98
0 7.
79.
113
.416
.8Fr
eddi
e M
acC
omm
unity
Gol
d25
1.9
21,9
25
0 8.
69.
713
.816
.7
Por
tfolio
Affo
rdab
le M
ortg
ages
Ban
k of
Am
eric
a Ze
ro D
own
248.
221
,606
0
8.4
10.1
14.1
17.9
Ban
k of
Am
eric
a C
redi
t Fle
x25
5.0
22,1
90
0 8.
79.
813
.816
.6P
ortfo
lio C
ompo
site
284.
724
,778
0
9.6
11.2
15.8
20.1
Gov
ernm
enta
l Affo
rdab
le M
ortg
ages
FHA
203
(b)–
prio
r31
1.5
27,1
11
0 9.
211
.418
.226
.1FH
A 2
03(b
)–cu
rren
t28
2.8
24,6
13
0 9.
011
.016
.121
.8
Sou
rce:
Aut
hors
’ana
lysi
s of
199
3 S
IPP
data
.a
The
agg
rega
te v
alue
of
hous
es t
hat
coul
d be
pur
chas
ed b
y cu
rren
t re
nter
s ab
ove
a lo
w-p
riced
hou
se t
hres
hold
.b
For
all
curr
ent
rent
ers,
not
just
ren
ters
who
can
affo
rd a
t le
ast
a lo
w-p
riced
hou
se.
cN
A =
3/2
Opt
ion
is n
ot a
ctiv
ated
(i.e
., 2
perc
ent
outs
ide
fund
ing
is n
otfo
rthc
omin
g).
dA
= 3
/2 O
ptio
n is
act
ivat
ed (
i.e.,
2 pe
rcen
t ou
tsid
e fu
ndin
gis
fort
hcom
ing)
.
80 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
In considering policies to further homeownership, we must also acknowledge some ofthe attendant risks and limitations. As noted earlier, the new homeowners may bevulnerable to worsening economic conditions that may affect their mortgage-payingability. To illustrate, we simulated the effect of a $2,000 downward “shift” in financesof the renters realizing homeownership via mortgage innovation. That “shift” couldconsist of income dropping by $2,000 or nonhousing expenses rising by that sameamount (e.g., for medical expenses)—two scenarios that could readily occur. In both
Table 33. Effects of Policy Options by Renter Group:a Percentage of Homeowner-OrientedRenters Nationwide Who Can Afford a Modestly Priced House
RecentAll White Black Hispanic Other Immigrant
Renters Renters Renters Renters Renters RentersBaseline/Option (%) (%) (%) (%) (%) (%)
BaselineModestly priced house, 10.2 13.2 3.8 3.1 10.7 10.9
GSE Standard Mortgage
OptionsI. Mortgage term adjustment
A. Interest rate reductionb
5% 11.0 14.3 3.8 3.7 10.7 10.92.5% 11.5 15.1 3.8 3.7 12.0 10.90% 11.8 15.3 4.3 3.7 13.3 10.9
B. Down payment reduction3% down 11.3 14.9 3.8 3.4 10.7 10.90% down 14.1 17.8 6.8 4.3 16.4 17.4
C. Reduced PMI (renewal)0% 10.3 13.3 3.8 3.4 10.7 10.9
II. Transaction-carrying cost adjustmentA. Closing-cost savings
Lowest closing costs 10.8 14.0 3.8 3.4 10.7 10.9B. Property tax savings
Lowest property tax 10.3 13.4 3.8 3.1 10.7 10.9C. Reduced hazard insurance
Lowest insurance 10.2 13.2 3.8 3.4 10.7 10.9III. House price adjustment
–10% 10.7 14.0 3.8 3.4 10.7 10.9IV. Borrower financial supplement
A. Annual income supplement+$1,000 10.7 13.9 3.8 3.4 10.7 10.9+$5,000 11.9 15.4 4.0 3.7 13.3 10.9+$10,000 12.4 16.1 4.0 4.0 13.3 10.9
B. Asset supplement+$1,000 11.3 14.5 4.4 4.2 11.9 10.9+$5,000 21.8 26.8 15.5 6.3 22.7 22.6+$10,000 57.1 60.3 64.6 39.9 43.1 51.5
Source: Authors’ analysis of 1993 SIPP data.a Non-Hispanic white, non-Hispanic black, Hispanic, and other non-Hispanic renter families are discrete. Combined,
they sum to all renter families. Recent immigrant families—those entering the United States after 1984—overlapwith the previously listed racial/ethnic groups (e.g., non-Hispanic whites and blacks). Finally, if not otherwise noted,whites, blacks, and other groups are non-Hispanic.
b Current contract interest rate is 8.05 percent.
Policy Considerations 81
The Potential and Limitations of Mortgage Innovation
situations, there would be $2,000 less available to pay for the homeownership-relatedoutlays of principal, interest, taxes, and insurance (PITI). We calculated what shareof renters who qualify for homeownership would become unable to meet the annualPITI mortgage commitment (i.e., become delinquent) with the $2,000 annual familyfinancial change. For example, for renters in the United States able to purchase amodestly priced home with a GSE Standard Mortgage, a $2,000 change in their fi-nances would result in a 9.3 percent delinquency in their ability to meet PITI costs;for renters able to buy a target-priced home with a GSE Standard Mortgage, a $2,000financial shift would result in their being 17.2 percent delinquent in covering PITIobligations. In an economic downturn, a $2,000 change in fortune is very likely andcarries with it the risk of new homeowners becoming delinquent.
Another type of “risk” is that mortgage innovation may expand homeownership oppor-tunities but not in locations with the better school systems and other sought-afterneighborhood traits that support future house-price appreciation. In many Americanmetropolitan areas, that often means homeownership in newer suburbs as opposedto inner-ring suburbs or inner-city locations. The 1999 State of the Nation’s Housingreports that “large and growing shares of both minority and low-income householdsare buying homes in the ‘suburbs’ ” (Joint Center for Housing Studies 1999, 18). None-theless, we still do not have a clear picture of the geography of housing opportunityoffered by mortgage innovation.
To further our knowledge in that regard, we applied the mortgage simulation in atest application. For illustrative purposes, we considered one of the more potent inno-vative mortgages, the Freddie Mac AG with 3/2 Option, and simulated the maximumhome-buying capacity with this product in the South Atlantic region. The medianhouse affordable to all renters in the South Atlantic region with the AG with 3/2Option falls just short of $109,000 in 1996. If we further assume that a sufficientnumber of homes would be offered for sale in this price range in all areas, then it ispossible to map a best-case scenario for the “geography of opportunity” (Galster and
Table 34. NPV Cost of Income-Asset Supplement PolicyOptions Applied to Homeowner-Oriented Renters
Modestly PricedHouse with a
GSE StandardMortgage*
Strategy ($ Millions)
Annual income supplement+$1,000 48+$5,000 742+$10,000 1,740
Asset supplement+$1,000 132+$5,000 6,698+$10,000 53,923
* See table 33 for details of the effects of each policy option.
Killen 1995) associated with mortgage market intervention. We apply this geographi-cal analysis in the Atlanta, metropolitan area, which is contained within the SouthAtlantic region. Results are shown in figure 2. Not surprisingly, flexible underwritingopens vast opportunities for ownership in Atlanta’s central city neighborhoods witholder, low-priced housing stock. Mortgage innovation also affords opportunities inAtlanta’s suburbs, yet it affords little access in the vibrant growth corridors of upper-middle-class Atlanta’s suburban housing submarkets located north and northeast of
82 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Figure 2. Geographical Variations in Housing Affordability with Flexible Underwriting
Source: Based on synthetic underwriting model for AG with 3/2 Option and Housing Consumption Model derivedfrom 1993 SIPP. Median house value for renter families able to afford a home at least as expensive as the 10th per-centile home in the South Atlantic Region, adjusted for inflation, was $108,667 in 1996. This value was divided bytract median house value in 1996 (estimated by the Atlanta Regional Commission) to obtain the values shown onthe map.
0 10 Miles
21%–34.9%35%–49.9%50%–74.9%75%–99.9%100%–199.9%200% or more
Maximum affordable houseas a percentage of tract
median home value, 1996
Atlanta City Boundary
Central BusinessDistrict
Policy Considerations 83
The Potential and Limitations of Mortgage Innovation
Figure 2. Geographical Variations in Housing Affordability with Flexible Underwriting (continued)
Less than 10 years10–19.920–29.930–39.940–44.945–49.950–57
Median Housing Age, 1996
Less than 15.0%15.0%–24.9%25.0%–34.9%35.0%–49.9%50.0%–88.0%
Poverty Rate, 1990
Source: Claritas™, Inc. (Real EstateSolutions series, 1996–97 Release.
Source: U.S. Bureau of the Census, 1990Census of Population and Housing.
the city. Again, we emphasize that figure 2 presents a best-case scenario, and limita-tions are no doubt far more severe for LMI families searching for available homes inlower price ranges. Efforts to funnel mortgage credit to underserved borrowers orneighborhoods, therefore, may support inner-city revitalization if pursued on a suffi-cient scale, but these efforts may fall short in broadening access to the well-fundedschool districts, expanding employment, and healthy price appreciation of favoredsuburban markets.
84 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
SUMMARY: THE ACHIEVEMENTS AND LIMITATIONS
OF MORTGAGE INNOVATION
Our simulations show the following:
1.Compared with their predecessors, current mortgage instruments are much morepotent in furthering homeownership. Total national absolute home-buying capac-ity, which was approximately $314 billion under the Historical Mortgage,35 hasincreased to a range of $500 billion to $600 billion under today’s more liberal mort-gages (table 12). Affordability also has increased. Under the Historical Mortgage,only 3.8 percent of renter families (1.0 million) could afford to purchase their“dream house” (i.e., the target-priced house). With today’s more aggressive mort-gage instruments, approximately 8 percent (2.0 million) of renter families can doso (tables 12 and 15). Expressed another way, the delta, or gain, in home-buyingcapacity from the Historical Mortgage to today’s more liberal mortgage products isabout $300 billion. When measured against the historical baseline, approximately1 million more renters can potentially buy the target house with the currentlyavailable mortgage instruments considered here. (Appendix F reports all of therespective delta gains from mortgage innovation.)
2.The more liberal mortgage instruments offer incrementally rising home-buyingopportunities. The national home-buying capacity is about $388 billion for the GSEStandard Mortgage. This absolute figure increases to a range of $445 billion to$584 billion for the GSE Affordable Mortgages,36 including the Emerging category,and it is in the $511 billion to $601 billion range for the Portfolio Affordable andGovernmental Affordable Mortgages (table 12).
3.The FHA loans are some of the most potent mortgage products offered. Becausethese products have, in fact, been offered for several decades, FHA must be cred-ited as a trailblazer of contemporary mortgage innovation.
4.Major hurdles remain. Even the most aggressive mortgage (FHA included) allowsonly 2.0 million renter families to realize target-priced homeownership, leavingapproximately 24 million renters unserved at the target level (table 15).
5.Lowering housing expectations helps, but a large gap remains. Even the mostliberal mortgage products considered in this report leave the vast majority ofrenter families unserved at any level. The most aggressive product leaves at least
Summary 85
The Potential and Limitations of Mortgage Innovation
35 The mortgage simulation can be extended to give further historical perspective. For example, the 1999 FannieMae Foundation Housing Conference selected FHA as one of the most significant housing achievements of thiscentury. This is borne out by the Mortgage Model, which indicates that pre-FHA (i.e., before 1934) loan instru-ments would allow approximately $220 billion of national home-buying capacity among today’s renters. Thatpre-FHA capacity is much lower than the roughly $600 billion capacity afforded by the ultimately adopted FHAmortgage (i.e., FHA 203(b)–prior).
36 With the 3/2 Option activated.
21 million renter families—or about 80 percent of the total—unable to enter home-ownership at even the low-priced threshold. Less advantaged subgroups (e.g.,blacks, Hispanics) fare even worse (table 18).
6.The reference gap—the difference between the price of the four reference houses(target priced, median priced, modestly priced, and low priced) and the value thateach renter can afford (capped at the four reference points)—further attests to thebenefits and limitations of mortgage innovation. For example, the reference gapwith the Historical Mortgage is approximately $2.2 trillion for the target-pricedhouse. The shortfall drops noticeably with contemporary mortgage products; how-ever, even with the most flexible products analyzed in this report (i.e., Flex 97,Community Gold), the reference gap for the target-priced house remains approx-imately $2.0 trillion (table 20).
7.Although many renters cannot shift their tenure status solely through mortgageinnovation, the incremental gains provided by innovative mortgage products areimportant, especially among minority populations. Absolute home-buying capaci-ty for black and Hispanic renters, about $11 billion and $10 billion, respectively,under the Historical Mortgage, increases to approximately $42 billion and $28billion, respectively, under today’s more liberal loan products considered in thisreport (table 13). Under the Historical Mortgage, the ability of blacks and Hispan-ics to afford a modestly priced house was barely measurable (i.e., approximately2 percent of these renter groups); the most aggressive mortgage products consid-ered here would allow approximately 4 percent to 6 percent of these minoritygroups to achieve homeownership of the modestly priced house. Recent immigrantrenters also gain from mortgage innovation (table 17).
8.The challenge of realizing homeownership is daunting, given renters’ scant finan-cial resources. Renters generally face both income and asset constraints, with thelatter posing a major challenge. Traditionally underserved renters in particularare financially constrained. Black and Hispanic renters have an average familyincome of less than $20,000 and average assets of about $2,000 (table 10). Nearlyall of the renters in these groups, therefore, lack the resources to buy a target-priced home at a cost of almost $100,000.
9.The outlook is somewhat brighter when we focus on the home-buying capacity ofhomeowner-oriented renters only rather than the home-buying capacity of allrenters. Without financial supplements, however, the gains for this latter groupare slight. For example, approximately 10 percent of the homeowner-orientedrenters can afford to purchase the target-priced house with today’s more liberalmortgage products (a slight increase from 8 percent for all renters).
10. If we accept it as an important societal goal, expanding homeownership will re-quire layered interventions. These include furthering mortgage innovation alongthe continuum sketched here (e.g., lowering down payment costs, raising front-endand back-end ratios); reducing the price of housing (analogous to lowering the“housing bar” from the target-priced house to the lower-cost benchmarks); reduc-
86 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
ing transaction costs and carrying costs; and offering various asset and incomesubsidies. The most potent intervention is the provision of income and, especially,asset supplements. These supplements are especially potent when focused on thehomeowner-oriented renters. A $10,000 asset infusion could potentially allowalmost 60 percent of all homeowner-oriented renters (and 40 percent of Hispanicand 65 percent of black homeowner-oriented renters) to realize modestly pricedhomeownership (table 33). The income and asset supplements, especially the latter,can be very expensive (e.g., the $10,000 asset supplement for the homeowner-oriented renters costs about $54 billion). All of these interventions are costly andproblematic.
In attempting to expand homeownership opportunities through mortgage innovation,we must recognize some of the attendant risks. Traditionally underserved householdsthat attain homeownership may be challenged to meet their mortgage obligations inan economic downturn. We also must work to expand the geography of housing oppor-tunity so that mortgage innovation broadens access to favored suburban markets.
These pessimistic conclusions from the mortgage simulation must be tempered, how-ever, with substantial real-world progress in expanding homeownership. We contrastsimulation results with observed real-world progress in the next section.
Summary 87
The Potential and Limitations of Mortgage Innovation
EVALUATION OF SIMULATION RESULTS
Our simulations indicate that relatively few renter families and hardly any minorityrenters can realize homeownership. That finding parallels very closely the conclusionsreached by Savage (1997, 1999). It is important to understand the limitations of suchan analysis, especially in light of empirical evidence of vigorous gains in homeowner-ship, particularly among the traditionally underserved.
There are approximately 35 million renter households in the United States, includingabout 12.6 million black households. Our mortgage simulations suggest that, evenwithout factoring credit underwriting, only 9.2 percent of all renter households and2.7 percent of black renter households37 could afford the modestly priced house38
(table 17). That would translate into about 3.2 million total and 0.35 million blackrenter households having the capacity to become homeowners.
The above figures are “stock” rather than “annual flow” figures; that is, they are basedon an application of the mortgage simulation model to all renters and represent thestock of renters who might qualify for homeownership at a given point in time. Wewould expect the stock figures to exceed that of the flow of renters actually movinginto homeownership in any given year.
In reality, actual net flow numbers on homeownership far exceed our estimated stockfigures. Between 1993 and the first quarter of 1999, there was a net increase of 7.8million homeowners in the United States (Joint Center for Housing Studies 1999).Over that six-year period, there was a net gain of 1.2 million black homeowners. Thesenet flows are multiples of our simulation estimate of the number of renters who couldpotentially afford modestly priced houses: 3.2 million total renters and 0.35 millionblack renters.
The same incongruity applies to Savage’s study. In his latest research (1999), he findsthat 9.9 percent of all renters and 3.2 percent of black renters could afford to buy amodestly priced house. Even if we assume, as above, that all renters, as well as allblack renters, bought only modestly priced units (an improbable scenario), Savage’spercentages would indicate that approximately 3.5 million total renter householdsand 0.4 million black renter households could realize homeownership. This is a farcry from the net addition of 7.8 million total homeowners and 1.2 million black home-owners in the last few years.
Further, in both our research and Savage’s research, credit underwriting was notperformed. In other words, the mortgage simulations represent a best-case scenario.In a real-world scenario, credit surely is considered and would dampen affordability.We obtain a glimpse of that in appendix B, where we incorporate a credit underwrit-ing submodel. Application of the credit underwriting submodel indicates that the num-
Evaluation of Simulation Results 89
The Potential and Limitations of Mortgage Innovation
37 This applies in a gross fashion our family-based simulations to households.
38 This is with the GSE Standard Mortgage, which is likely to be the most common mortgage product.
ber of black renters able to achieve homeownership (375,000, averaging our resultsand Savage’s results) would be considerably lower in a real-world credit-scoring envi-ronment. However, actual figures show that there was a net increase of 1.2 millionblack homeowners from 1993 to 1999.
Other statistics also suggest more progress in real-world homeownership than thatsuggested by our simulation results. Home Mortgage Disclosure Act (HMDA) statis-tics, for example, show the changes indicated in table 35 in the recipients of homepurchase loans. While the HMDA data include home purchase loans made to existingas well as new homeowners, the above gains in home purchase loans made to blacksand Hispanics contradict the results of our simulation and Savage’s study, which in-dicate that very low numbers of black and Hispanic renters can purchase a home. Tocomport the above HMDA data with our data, one would have to assume that almostall of the minorities receiving a home purchase loan were move-up minority homebuyers, as opposed to first-time purchasers, and that is unlikely.
Longitudinal analysis of the SIPP also poses incongruities. Of the 4,565 total familiesin the 1993 SIPP that we examined, 4,109 remained renters and 456 (4,565 – 4,109)became owners over the course of the panel interviews. This tenure change allows aretroactive assessment. Based on the mortgage simulation model, we can predict themaximum home affordable to the 456 families as of the time they were still renters.This, after all, is exactly what we do for the 4,109 renter families who remainedrenters. However, for the 456 families that changed tenure, we can compare our pre-dicted maximum home-buying capacity with the price of the home actually bought.One would not expect that the actual purchase values would mirror the modeledprices because some fluctuation around the latter is to be expected (e.g., renters mayelect to buy a house that is priced at less than their absolute maximum home-buy-ing capacity). However, within that fluctuation, we should not find many instanceswhere families are consuming much more than what our mortgage simulation indi-cates is the ceiling.
In fact, the retroactive assessment summarized in table 36 shows that 93 percent ofthe 456 families are buying houses that are priced higher than the affordable levels
90 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Table 35. Home Purchase Loans by Recipient Group,1990 and 1997
MortgageRecipient Group 1990 1997
Whites 1,733,981 2,997,069Blacks 94,624 257,233Hispanics 100,022 254,382Other* 87,496 133,123Total 2,016,123 3,641,807
Source: HMDA statistics for 1990 and 1997 were provided by theWoodstock Institute (1999).* Asian and Native American.
forecast by the mortgage simulation model. The gap between the modeled purchasingpower and the actual home resource consumed is quite large. We find that almost 88percent of the 456 families purchased a home that is at least 50 percent more expen-sive than the price we model as affordable. This gap must give considerable pause tomortgage simulation research that is based on the SIPP.
These disparate statistics suggest that renters—especially traditionally underservedrenters—have a home-purchasing capacity that is greater than that suggested bySIPP-based mortgage simulations. That does not mean that the SIPP-based simula-tions are wrong but rather that we must understand the limitations of the exercise.
There are numerous reasons why observed homeownership gains may exceed whatwe and Savage predict. First, SIPP data may understate the level of resources avail-able to renters; respondents may underreport informal income or the “hidden wealth”associated with intergenerational or other wealth transfers (Gale and Scholz 1994;Gyourko, Linneman, and Wachter 1997). The mortgage industry is recognizing thisunderreporting phenomenon. For example, one Bank of America portfolio product,
Evaluation of Simulation Results 91
The Potential and Limitations of Mortgage Innovation
Table 36. Comparison of the Price of the Actual Home Purchasedwith the Simulated Purchasing Capacity ofSIPP Renters Who Became Homeowners*
Price of Percentage ofPurchased House as SIPP Rentersa Percentage of the Who BecamePurchasing Capacity Homeowners
More expensive>50 87.7440–50 0.4130–40 1.1520–30 1.0210–20 0.930–10 1.77
Subtotal 93.03
Less expensive(–10)–0 0.92(–20)–(–10) 1.44(–30)–(–20) 0.79(–40)–(–30) 0.49(–50)–(–40) 1.53<(–50) 1.80
Subtotal 6.97
Total 100.00
Note: Figures may not add to indicated totals because ofrounding.* The 1993 SIPP contained a sample of 4,565 renter families,
of which 456 moved from renting to owning between the firstand seventh waves of the survey. Simulated purchasingcapacity was computed using the assets in wave 1 (beforethe house was bought) and income in wave 7.
Zero Flex™, will allow up to $600 a month in undocumented income. However, whenwe conduct simulations based on the income statistics in the SIPP, we are applying avery stringent standard. The simulations might be improved by obtaining guidancefrom lenders, community groups, and others who deal with renter populations (espe-cially the traditionally underserved) on the “hidden” financial resources that may beavailable to renters from relatives and other sources.
We must also recognize that the SIPP renter data, even if fully accurate, representsa fixed point in time and that renters electing to become homeowners can bootstrapthemselves financially. For example, a head of a household can take a second job, anonemployed spouse can enter the labor force, or discretionary spending can be re-duced. Given the low LTVs of today’s mortgages and, therefore, the tremendouslylessened down payment requirements, the actions just noted can quickly yield resultsfor renters.
Researchers are increasingly recognizing the endogenous relationship of savings andthe decision to purchase (Haurin, Hendershott, and Wachter 1997). Thus, renters mayvery well raise their savings rate just before buying a home. Traditionally underservedrenters, many of whom have the income but not the down payment to buy a house,may simply delay saving until they are ready to choose homeownership.
Earlier, we examined the tremendous homeownership-enhancing capacity of a $5,000public grant to renters. We observed that the multi-billion-dollar cost of this grantmade it unlikely to be widely adopted. But if renters could access $5,000 (e.g., fromrelatives), or if they could save this sum, the results of the SIPP-based mortgage sim-ulation would be dramatically altered.
Thus, we must be careful in the conclusions drawn from an SIPP-derived model. Sim-ilar cautions would apply to others doing similar work with other data sources. If weapply the simulation model with the data as given, we are likely to underestimaterenters’ true home-buying capacity. The results we, Savage, and others have obtainedare informative; however, these results also convey gross orders of magnitude or rel-ative order of accomplishment across mortgage instruments as opposed to literal po-tential market penetrations. For example, we find that 1.0 percent of black renterfamilies (43,113) could have afforded a target-priced home with a Historical Mortgage.The figures increase with today’s loan products: 1.5 percent of black renter families(66,321) can afford the target-priced house with a GSE Standard Mortgage; 2.1 per-cent of black renter families (95,041) can be served with the CHBP mortgages; and2.1 percent of black renter families (95,041) can be served with the AG mortgages.Product managers at the GSEs should not take these figures literally as representingmaximum market captures. Instead, the lessons are that the GSE Affordable productsoffer a slight gain in affordability over the Historical Mortgage, and that the FreddieMac mortgages offer a slight gain in affordability relative to their Fannie Mae equiv-alents. (The delta gains from mortgage innovation are detailed in appendix F.) Finally,the simulations indicate that although commendable progress has been made fromthe historical baseline, much remains to be done to expand homeownership opportu-nities to traditionally underserved populations.
92 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
Refining the results of the mortgage simulations will require better data on and un-derstanding of renters. We need more insight into the relationship between savingsand the decision to purchase. Access to credit score data is especially critical for im-proving the simulations (see appendix B). In addition, the simulations do not considerhousing availability, which is another component of the affordability equation worthyof future study.
Finally, even with enhanced data, the mortgage simulations report only the mathe-matical calculation of home-buying capacity based on the confluence of the renter’sfinancial profile and the loan terms. As the mortgage industry changes the latter,home-buying capacity expands. This study documents the significant gain providedby mortgage innovation as of the late 1990s. Yet mortgage innovation is only a firststep to a more comprehensive approach to expanding homeownership.
The more encompassing approach adds other components to the very real contribu-tions of mortgage innovation. Subsidies from the public sector, foundations, interme-diaries, and other sources help bridge the financial gap. Partnerships of varioustypes—such as lenders joining forces with other for-profit entities, government, andthe nonprofit sector—are also commonly used to address the multiple barriers tohomeownership that arise at different stages of the home-buying process. To reachthe underserved, for example, a lender will typically join forces with a church or neigh-borhood group. Counseling to address credit issues and education of prospective homebuyers often involve similar alliances. Ensuring that traditionally underserved house-holds remain successful borrowers over the long term also often requires a joint effortby the lender and a nonprofit counselor.
A companion study to the current investigation presents detailed case studies of thecomprehensive approach described above (Listokin et al. 2000). One example is theLittle Haiti Housing Association (LHHA), which is bringing homeownership to one ofthe most disadvantaged populations—very low income Haitian immigrants living inan impoverished neighborhood (“Little Haiti”) in Miami, Florida. LHHA is currentlyqualifying households earning approximately $18,000 to $20,000 to purchase homescosting about $80,000. That is roughly the price of the median-priced home in thecensus division (South Atlantic) that contains Florida, and the Haitians’ $18,000 to$20,000 average income comports very closely with that of the average black renter.Applying mortgage innovation alone, mortgage simulations find that few black renterscould afford a median-priced house. Yet LHHA is making homeownership affordableto the residents of Little Haiti. How LHHA accomplishes that feat and addresses themany barriers to attracting, qualifying, and retaining disadvantaged homeowners isdetailed in the companion study. We offer a synopsis below.
To bring the Haitians to homeownership, LHHA engages in a broad array of activities(Harder 1998). Outreach is important, for few Haitians believe they can afford a homeand many fear dealing with banks. (These apprehensions are widespread among im-migrant populations [Ratner 1997].) Because a large number of Haitians have blem-ished credit records, LHHA works on addressing credit issues. Furthermore, LHHA
Evaluation of Simulation Results 93
The Potential and Limitations of Mortgage Innovation
helps Haitian renters overcome the tremendous financial barriers to homeownership(the underpinning of our and Savage’s findings) through the following:
1. A low 5 percent down payment by the purchaser
2. A market-rate, no points, modest-size first mortgage granted by a lender
3. A large, soft second mortgage (i.e., with minimal repayment requirements) fund-ed by Miami-Dade County or the federal government
4. A modest-size, soft third mortgage from the Federal Home Loan Bank’s Afford-able Housing Program
The 5 percent down payment typically amounts to about $4,000. LHHA notes thatHaitian households often do not have these funds immediately available. These house-holds, however, garner the down payment from relatives’ gifts, informal lending ar-rangements (e.g., “sous-sous”), and, most frequently, savings. Counselors at LHHArecount numerous instances of Haitian households earning as little as $15,000 to$20,000 a year who manage to save $3,000 to $5,000 of their income. Haitians livedvery frugally, often growing food on empty lots or forgoing a car purchase. The Haitianexperience more generally points to the ability of renters to save once they decide tobecome homeowners. In other words, the low assets of renters might be a transitoryrather than a permanent condition.
The 5 percent down payment from the LHHA home buyers still leaves about $75,000in financing to be secured on the $80,000 homes currently being sold. A $75,000 “hard”(i.e., fully payable) mortgage would be beyond the reach of the Haitian renters. Inresponse, LHHA assembles substantial “soft” financing so that the loan repaymentcan be sustained. Thus, of the $75,000 total, approximately $25,000 would be carriedas a hard first mortgage and the remaining $50,000 would be partially repaid.
In sum, LHHA clientele benefit from mortgage innovation. The Haitian home buyersare making small down payments—5 percent of the house value. Such a small invest-ment would be impermissible under the Historical Mortgage. In addition, LHHAfinancing permits the buyer to fold the closing costs into the home purchase price—another strategy not allowed by the Historical Mortgage. Haitian home buyers arealso gaining from the underwriting reforms of mortgage innovation. The HistoricalMortgages’ underwriting requirements of a strong formal credit record, steady docu-mented employment at the same job/occupation, strict asset verification, and a middle-income “neighborhood standard” would disqualify almost all Haitians attempting toobtain a home loan in Little Haiti.
Mortgage innovation alone would not enable the Haitians to become homeowners.They need LHHA’s outreach to assuage deep apprehensions about dealing with insti-tutional lenders. LHHA’s intensive homeownership and credit counseling is also aprerequisite and helps Haitians become “bankable.” The layered subsidies packaged byLHHA are essential to the strategy, enabling very low income Haitians to purchase
94 David Listokin, Elvin K. Wyly, Brian Schmitt, and Ioan Voicu
Fannie Mae Foundation Research Report
an $80,000 home. We calculate a subsidy of about $30,000 for each LHHA unit (Lis-tokin et al. 1999). In addition, the intensive postpurchase support provided by LHHAcontributes to the zero mortgage delinquency rate observed thus far.
All of the above LHHA activities are examples of what we term the comprehensivepartnership strategy. That strategy, coupled with mortgage innovation, is contributingto the impressive recent gains in homeownership in the United States.
Evaluation of Simulation Results 95
The Potential and Limitations of Mortgage Innovation