Housing Finance Policy Center
Lunchtime Data Talk
Demographic Change in the US and its
Implications for Housing and Mortgage Lending
Rolf Pendall, Urban Institute
Walter Scott, Fannie Mae
Jim Carr, Wayne State University
December 16, 2015
Rental and OwnerOccupied Housing
Demand, 2010-2030
Rolf Pendall
December 16, 2015
Urban Institute
Middle-class housing on Grove Avenue:
https://en.m.wikipedia.org/wiki/West_Hill,_Albany,_New_York#/media/
File%3AAlbany_Houses.jpg UpstateNYer • CC BY-SA 3.0
"Cairo Apartment Building - Washington, D.C." by
AgnosticPreachersKid - Own work. Licensed under CC BY-SA
3.0 via Commons -https://commons.wikimedia.org/wiki/
File:Cairo_Apartment_Building_-_Washington,_D.C..JPG
Presentation outline
A few orientation slides: Demographic changes driving
housing demand
A long look back: Headship and homeownership since the
early 1900s
Future homeownership and headship: How our growing,
aging, diverse population will fuel substantial demand for
housing, especially rentals
Millions
Population, 2014-60: Despite 16 million fewer
whites, all other groups fuel increase of 100 million
450
400
*
350
Two or more
Asian
300
Black
250
Hispanic
**
200
150
100
White
50
0
2014
2024
2034
Source: U.S. Census 2014 National Population Projections
* Native Hawaiian or Pacific Islander
** American Indian or Alaska Native
2044
2054
5
Millions
Over 80 Million Seniors in 2040
400
350
300
250
65+
200
45-64
15-44
150
<15
100
50
0
2015
2020
2030
Source: U.S. Census 2014 National Population Projections
2040
6
A young, diverse population will spur continued
household growth—especially rental housing
Households (millions)
90
80
Other race
70
Hispanic
60
Black
50
White
40
30
20
10
Renters
2030(p)
2020(p)
2010
2000
1990
2030(p)
2020(p)
2010
2000
1990
0
Owners
Source: U.S. Census 1990-2010, Urban Institute Projections (p), average series. Other race includes Asians
and Pacific Islanders, American Indians and Alaska Natives, people of other races, and people of two or
more races.
7
Translating population to housing
Osbornb, Housemates, Santa Barbara, 1603 de la
Vina Street, 1974.
• Every occupied
housing unit has a
household in it
• Every household has a
householder: The
person whose name is
on the lease or
mortgage
• Headship rate: The
number of
householders per
person, usually
expressed for agespecific groups
8
A long look back: Household
formation & homeownership
since the early 1900s
9
For most age groups, headship peaked about 1980
70%
65–74
60%
55–64
50%
45–54
40%
35–44
30%
30–34
25–29
20%
20–24
10%
0%
1930
1950
1970
1990
2010
2030
Sources: Decennial Censuses 1930–2000 and American Community Survey 2007 through 2013, extracted from Steven Ruggles, J.
Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use Microdata
Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
10
Oldest seniors have been gaining steadily in
headship
70%
75-84
85+
60%
50%
40%
30%
20%
10%
0%
1930
1950
1970
1990
2010
2030
Sources: Decennial Censuses 1930–2000 and American Community Survey 2007 through 2013, extracted from Steven Ruggles, J.
Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use Microdata
Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
11
Homeownership rate: Two eras since 1900
80%
70%
60%
50%
40%
30%
20%
10%
0%
1900
1920
1940
1960
1980
2000
2020
Sources: Decennial Censuses 1900–2000 and American Community Survey 2007 through 2013, extracted from Steven Ruggles, J.
Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use Microdata
Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
12
Homeownership also peaked in 1980 for <55
80%
70%
45–54
60%
40–44
35–39
50%
30–34
40%
30%
25–29
20%
10%
0%
1900
1920
1940
1960
1980
2000
2020
Sources: Decennial Censuses 1900–2000 and American Community Survey 2007 through 2013, extracted from Steven Ruggles, J.
Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use Microdata
Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
13
Senior homeownership may be starting to peak
90%
80%
65–74
55–64
70%
75+
60%
50%
40%
30%
20%
10%
0%
1900
1920
1940
1960
1980
2000
2020
Sources: Decennial Censuses 1900–2000 and American Community Survey 2007 through 2013, extracted from Steven Ruggles, J.
Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use Microdata
Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
14
100 Years of Homeownership Transitions
90%
Birth
year
1885
80%
Homeownership
70%
1895
60%
1905
50%
1915
40%
1925
30%
1935
20%
1945
10%
1955
0%
1965
25
2000
1990
1980
1970
1960
1950
1940
1930
1920
1910
35
45
55
Age
65
75
85
1975
Sources: Decennial Censuses 1900–2000, extracted from Steven Ruggles, J. Trent Alexander, Katie Genadek, Ronald Goeken, Matthew
B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use Microdata Series: Version 5.0 [Machine-readable database]. Minneapolis:
University of Minnesota.
15
Looking Ahead: Demographic
Change and Housing Demand
16
Headship and homeownership are like a race
By U.S. Army (Flickr: 2010 Army Ten Miler Start) [CC BY 2.0
(http://creativecommons.org/licenses/by/2.0)], via Wikimedia Commons,
https://upload.wikimedia.org/wikipedia/commons/3/3d/2010_Army_Ten_Miler_Start.jpg
17
%-point change in cohort headship
Young Adults Dominate Headship Transitions
50%
40%
Age at
beginning of
decade
15-19
30%
20-24
20%
25-29
10%
0%
1960s
1970s
1980s
Decade
1990s
2000s
Sources: Decennial Censuses 1960–2000 and American Community Survey 2010, extracted from Steven Ruggles, J.
Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public
Use Microdata Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
18
Projecting Headship: Picking Past Transition Rates
%-point change in cohort headship
rate
40%
White non-Hispanic decennial transition rates, 1990-2010
1990-2000
2000-10
30%
Average
20%
10%
0%
-10%
15-24
25-34
Sources: Decennial Censuses 2000 and 2010.
35-44
45-54
55-64
65-74
75+
19
Projecting Headship: Future transition rates
%-=point change in cohort headship
rate
40%
White non-Hispanic decennial transition rates, projections
Slow
Fast
30%
20%
10%
0%
-10%
15-24
25-34
35-44
45-54
Sources: Decennial Censuses 2000 and 2010, Urban Institute Projections
55-64
65-74
75+
20
Projecting Homeownership: Future transition rates
White non-Hispanic decennial transition rates, projections
Slow
40%
%-point change in cohort
homeownership rate
Fast
30%
20%
10%
0%
-10%
15-24
25-34
35-44
45-54
Sources: Decennial Censuses 2000 and 2010, Urban Institute Projections
55-64
65-74
75+
21
Whites have early homeownership advantage
40%
Fast scenario, decennial transition rates, projections
White
%-point change in cohort
homeownership rate
Black
30%
Hispanic
Other
20%
10%
0%
-10%
15-24
25-34
35-44
45-54
Sources: Decennial Censuses 2000 and 2010, Urban Institute Projections
55-64
65-74
75+
22
Setting the starting point: Finding 2013 cohort rates
60%
50%
White non-Hispanic headship rates, 2010-13 ACS
38-47
28-37
35-44
25-34
40%
30%
18-27
20%
10%
15-24
0%
2010
2011
2012
2013
Source: American Community Survey 2010 through 2013, extracted from Steven Ruggles, J. Trent Alexander,
Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. 2010. Integrated Public Use
Microdata Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota.
23
Setting the starting point: Adjusting ACS to Census
ACS
HVS
70%
65% 66%
65%
Sources: Decennial Censuses, Housing Vacancy Survey, and American Community Survey.
2013
2012
2011
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
60%
2000
Overall homeownership rate
Decennial Census
24
Projecting Homeownership: Transitions 2013-2030
Homeownership
80%
Average of fast and slow scenarios
2010
2020
2030
60%
40%
20%
0%
Age in 2010 15-24 25-34 15-24 25-34 15-24 25-34 15-24 25-34
White
Black
Sources: Decennial Census 2010, Urban Institute Projections
Hispanic
Other
25
Adding it up: Households by race & age, 2010-30
65+
90
80
70
60
50
40
30
20
10
0
45-64
2010
Sources: Decennial Census 2010, Urban Institute Projections
2020
2030
Other
Hispanic
Black
White
Other
Hispanic
Black
White
Other
Hispanic
Black
<45
White
Millions
Average of fast and slow scenarios
26
Change in households by race and age, 2010-2030
Millions
Average of fast and slow scenarios
65+
45-64
<45
Total
8.0
6.0
4.0
2.0
0.0
-2.0
-4.0
2010s
Sources: Decennial Census 2010, Urban Institute Projections
2020s
Other
Hispanic
Black
White
Other
Hispanic
Black
White
-6.0
27
Homeowners by race and age, 2010-2030
65+
70
60
50
40
30
20
10
0
45-64
2010
Sources: Decennial Census 2010, Urban Institute Projections
2020
2030
Other
Hispanic
Black
White
Other
Hispanic
Black
White
Other
Hispanic
Black
<45
White
Millions
Average of fast and slow scenarios
28
65+
45-64
<45
Total
6.0
4.0
2.0
0.0
-2.0
-4.0
2010s
2020s
Other
Hispanic
Black
White
Other
Hispanic
Black
-6.0
White
Millions
Change in homeowners by race and age, 2010-30
29
By the 2020s, Hispanics, blacks, Asians, and other nonwhites will account for all growth in homeownership
Assumes average between the high and low scenarios
Black
Other
White
Other
Black
0%
White
10%
Hispanic
Other
Black
20%
Other
Black
White
Hispanic
40%
30%
Hispanic
Hispanic
50%
White
Percent change in homeowners by
race/ethnicity
60%
-10%
1990–2000
2000–10
Source: U.S. Census 1990-2010, Urban Institute projections.
2010–20
2020–30
2010
2020
2030
Other
Hispanic
Black
White
Other
Hispanic
Black
White
Other
Hispanic
Black
White
Millions
30
Renters by race and age, 2010-2030
30
65+
25
45-64
20
<45
15
10
5
0
31
3.0
65+
45-64
<45
Total
2.5
2.0
1.5
1.0
0.5
0.0
-0.5
2010s
2020s
Other
Hispanic
Black
White
Other
Hispanic
Black
-1.0
White
Millions
Change in renters by race and age, 2010-2030
32
Mapping America’s Futures:
Local demographics, housing markets, and fair housing
Projected population change, 2010-30
33
Thanks!
@rolfpendall
rpendall@urban.org
Mortgage Lending and
Non-Borrower Household
Income
Walter Scott
Senior Economist, Fannie Mae
December 16, 2015
© 2011 Fannie Mae. Trademarks of Fannie Mae.
© 2015 Fannie Mae. Trademarks of Fannie Mae.
34
Disclaimer and Acknowledgments
This research was conducted by Fannie Mae’s Economic and Strategic Research (ESR) and
Credit Portfolio Management Analytics (CPM Analytics) groups. Opinions, analyses, estimates,
forecasts, and other views included in these materials do not necessarily represent views of
Fannie Mae or its management, should not be constructed as indicating Fannie Mae’s business
prospects or expected results, and are subject to change without notice. Although ESR and CPM
Analytics base their opinions, analyses, estimates, and other views on information they consider
reliable, it does not guarantee that the information provided in these materials is accurate,
current, or suitable for any particular purpose. Changes in the assumptions or the information
underlying these views could produce materially different results. The analyses, opinions,
estimates, forecasts, and other views published by ESR and CPM Analytics represent the views
of those groups as of the date indicated and do not necessarily represent the views of Fannie
Mae or its management.
Acknowledgments
The author wishes to thank Hamilton Fout, Patty Koscinski, and Mark Palim at Fannie Mae, as
well as Nayantara Hansel and Saty Patrabansh at FHFA, who reviewed earlier drafts of this
presentation and the associated research paper. Nuno Mota at Fannie Mae provided valuable
advice and assistance in reviewing my data preparation and regression analysis.
December 2015 | Mortgage Lending and Non-Borrower Household Income
35
Introduction
•
An increasing proportion of U.S.
households have shared or extended family
structures.
•
“Doubling up” increased after the 2008
recession, but there are also longer-term
contributing factors:
•
Greater share of minority & immigrant
families
•
Millennial generation slower to marry,
form independent households
December 2015 | Mortgage Lending and Non-Borrower Household Income
36
Introduction
•
Members of these households may share
expenses and household duties (including
childcare) in a variety of ways.
•
But unless they are co-signers of the loan,
the financial contributions of nonborrowers are not counted in, and are
invisible to, the standard mortgage
underwriting process.
•
The borrower may then be denied a loan
due to debt-to-income (DTI) constraints
that underestimate their actual resources.
December 2015 | Mortgage Lending and Non-Borrower Household Income
37
Research Questions
•
We hypothesize that by considering non-borrower
household income in some way, we can expand access to
credit for this group of applicants, without taking on
additional risk relative to comparable borrowers.
•
The goal of this research has been to mine public data
sources to understand:
•
How many and what kinds of households have significant nonborrower income? How do they compare to other homeowners?
•
How stable is this income? What are the risks if the non-borrower
moves out of the home?
•
Can we get at least indirect evidence of the relative loan performance
of these households?
December 2015 | Mortgage Lending and Non-Borrower Household Income
38
Background: Beyond the Nuclear Family
•
Extended family households are more
common among minority groups and
newer immigrants.
•
Pew Foundation: Multi-Generational
families declined from 25% in 1940 to 12%
in 1980 – but then increased to 16% in
2010.
•
Increase seen even within racial/ethnic
groups, driven in part by elderly adults
moving in with kids, but mostly by coresidence with parents/older relatives of
18 to 34 year olds.
December 2015 | Mortgage Lending and Non-Borrower Household Income
39
Background: Beyond the Nuclear Family
•
Surveys of shared households identify
several motivating factors*:
•
Care for a disabled or elderly family member.
•
Mutual social, emotional, and financial support
including help with child care.
•
Cultural norms of extended family patterns.
•
Pooling of resources to afford a home in a
better neighborhood.
•
Response to financial emergency or other life
event.
* See additional research: Bengston (2001) hypothesizes that people are relying more on extended family for support as divorce
and single parenthood increases. Bleemer (NY FRB 2014) finds that co-residence of young adults has a strong geographic
component – concentrated in areas with weak economy and/or high housing costs. Rising student debt is also a likely factor.
December 2015 | Mortgage Lending and Non-Borrower Household Income
40
Background: Beyond the Nuclear Family
•
Shared households come in several
configurations:
•
Vertical Up: parent(s) of head of
household/spouse move in with family
•
Vertical Down: adult children of head of
household continue to live in household
•
Horizontal: siblings and their families live in
same household
Extended households may be at a disadvantage in mortgage
lending if non-borrowers make financial contributions, but do not
make explicit rent payments to the borrowers.
December 2015 | Mortgage Lending and Non-Borrower Household Income
41
Beyond the Nuclear Family
•
Evaluating these families for mortgage
credit is challenging, as origination data
does not include the income of the
additional household members.
•
Leveraged federal government housing
surveys for evidence:
•
American Community Survey (ACS)
•
American Housing Survey (AHS)
December 2015 | Mortgage Lending and Non-Borrower Household Income
42
Extended-Income Households
•
For mortgage origination, extended-income household
(EIH) is defined as one where non-borrower adults have
income that is at least 30% of the borrower(s)’ income.*
•
Most non-borrowers contributing such income are relatives
of the borrower, but some are unmarried partners, friends,
or other roommates.
Percentage of households with mortgages that are EIHs (by type of
non-borrower/ household)
All HH
Relatives
Unmarried Partner
Other Non Relative
Total
Asian
9
3
2
15
Hispanic
Afr-Am
12
3
2
17
14
3
2
20
17
4
2
24
Immigrant
15
3
2
20
Source: 2013 American Community Survey
December 2015 | Mortgage Lending and Non-Borrower Household Income
43
Extended Income Household Members
•
The highest-income non-borrower is most commonly an
adult child or son/daughter-in-law of the borrower or
spouse.
•
Next most common pattern is an unmarried partner of the
borrower – not usually counted in “shared household”
literature but relevant in a mortgage lending context.
Percent of EIHs where highest earning non-borrower has a given relationship to
the borrower(s) by household member, race/ethnicity, immigrant status
Source: 2013 American Community Survey
December 2015 | Mortgage Lending and Non-Borrower Household Income
44
Definitions
•
In this presentation, the universe is limited to homeowners
with mortgages.
•
Shared households are those with at least one adult other
than the borrower or their spouse. This person may be a
relative, unmarried partner, or other non relative.
•
Extended-Income Households are shared households
where non-borrower adult income is > 30% of that of the
borrower + spouse.
•
Non EIH Shared households are those where nonborrower income is < 30% of borrower + spouse income.
• Nuclear households are non-shared households.
December 2015 | Mortgage Lending and Non-Borrower Household Income
45
Extended Income Household Geography
EIHs more likely to be found in bigger MSAs, those with high
minority population, and with greater housing cost.
December 2015 | Mortgage Lending and Non-Borrower Household Income
46
Time Trends in EIHs vs. Other Households
Tracking the numbers of EIHs vs. non-EIH shared households (i.e.,
those where the non-borrowers have relatively little income) indicates
that the crisis years had a much larger effect on the non-EIH
configuration (where home owners provided shelter for dislocated
individuals), than on the levels of EIHs, which grew more erratically.
Percentage of all households with mortgages, and all minority-headed
households with mortgages, that are either shared households (any adult other
than borrower/spouse), and non-borrowers have > 30% or < 30% of borrower
income.
Source: American Community Survey, 2007-13
December 2015 | Mortgage Lending and Non-Borrower Household Income
47
Who Lives in Extended Income Households?
•
Multivariate analysis reveals that borrowers are more
likely to live in an EIH if they:
• Are unmarried
• Have a lower level of education
• Are immigrants or members of a minority group
• Retired or became disabled within the past five years.
December 2015 | Mortgage Lending and Non-Borrower Household Income
48
Drivers for EIH formation
Dependent variable = household is an EIH. Population = 2013 ACS
households with mortgages.
December 2015 | Mortgage Lending and Non-Borrower Household Income
49
Drivers for EIH formation
•
Shared households are more likely to have individuals
who are unemployed or who have left the labor force.
•
But this is generally less true for EIHs in comparison to
other shared households.
December 2015 | Mortgage Lending and Non-Borrower Household Income
50
Evaluating Credit Risk for EIHs: Measures
•
Two income measures are used to compare median
income for three types of households:
• Nuclear Households
• Non EIH Shared
• EIH
• Income measure 1 = Core income (borrower/spouse)
• Income measure 2 = Extended income that includes
Core income plus shadow rent. Shadow rent is 30% of the
adult non-borrower income.
December 2015 | Mortgage Lending and Non-Borrower Household Income
51
Evaluating Credit Risk for EIHs: Shadow Rent
“Shadow rent,” defined as 30% of non borrower adult
income, is used to estimate the actual or potential resources
available to the borrower(s) from non-borrower(s).
• May not be the actual amount of financial contribution
made, but reasonable to expect that the borrower could
draw on this level of support in a crisis.
• 30% is used because it is the affordability standard under
most federal rental housing assistance programs.
December 2015 | Mortgage Lending and Non-Borrower Household Income
52
Evaluating Credit Risk for EIHs: Income Analysis
•
Median income of EIHs is substantially lower than that of
other household types − more likely to be DTI constrained.
•
True resources of EIH borrowers are potentially 1/3 more
than what shows up on a mortgage application when
shadow rent is considered.
Median household income by household type and race/ethnicity/immigrant status
All HH w/mtg
Asian
African-American
Hispanic
Immigrant
Extended Income HH
Core
Extended
Income
Income
$41,900
$54,600
$59,299
$68,100
$36,000
$48,000
$37,000
$49,000
$43,000
$57,100
Non-EIH Shared
Core
Extended
Income
Income
$85,000
$87,350
$100,000 $103,000
$67,000
$69,080
$68,880
$70,420
$86,100
$88,840
Nuclear
Core
Income
$77,000
$105,000
$60,100
$69,130
$85,000
Source: American Community Survey 2007-13
December 2015 | Mortgage Lending and Non-Borrower Household Income
53
Evaluating Credit Risk for EIHs: Loan Performance
•
Created an indirect measure of loan performance using
the American Housing Survey (AHS); no direct
performance data on EIHs is available.
•
AHS is panel of 50,000 households surveyed every two
years.
•
Identified a population of underwater borrowers with MTM
LTV >110% for years 2007, 2009, and 2011.
•
Tracked whether these borrowers then remained in their
homes or moved out two years later. These move-out
events are likely to have been associated with defaults
due to the lack of equity.
December 2015 | Mortgage Lending and Non-Borrower Household Income
54
Evaluating Credit Risk for EIHs: Loan Performance
Summary Statistics, 2007/2009/2011/2013 American Housing Survey,
Household (HH) w/ mortgages
Nuclear
Non EIH
Shared
Extended
Income HH
Borr/spouse income
$
89,690 $
94,788
$
44,642
Income w/ shadow rent
$
89,690 $
96,419
$
56,469
Total household income
$
89,690 $
100,225 $
84,064
2.4%
Gets outside help w/ mtg
Monthly housing cost
$
Estimated mortgage UPB
$
3.0%
$
1,875 $
1,624
150,507 $
148,412 $
126,730
Estimated MTMLTV
67.3
62.4
63.8
Underwater (LTV > 110%)
6.7%
6.7%
6.8%
9.7
12.2
11.7
First time home buyer
Has rental income
41.5%
39.4%
51.8%
6.8%
7.3%
5.7%
Has self-emp income
15.0%
15.7%
10.4%
Has social security
15.3%
10.7%
16.9%
Has retirement income
11.3%
9.2%
9.4%
0.8%
0.7%
1.2%
# of years in home
Has boarder
1,728
2.3%
December 2015 | Mortgage Lending and Non-Borrower Household Income
55
Evaluating Credit Risk for EIHs: Loan Performance
Summary Statistics, 2007/2009/2011/2013 American Housing Survey,
Household (HH) w/ mortgages
Nuclear
Someone aged 65+ in HH
Non EIH
Shared
Extended
Income HH
13.3%
1.7
16.5%
3.0
22.5%
2.9
46.8
50.0
48.9
70.1%
70.1%
31.4%
43.4%
23.6%
45.7%
34.8%
35.1%
37.1%
Borr/spouse is Afr-Am
8.5%
12.8%
12.6%
Borr/spouse is Asian
4.4%
6.8%
6.0%
10.1%
15.1%
17.9%
Number of adults
Borr/spouse age
Borrower is married
Has children in HH
Borr/spouse is minority
Borr/spouse is Hispanic
Sample size (2007-2013)
52,580
14,428
7,532
December 2015 | Mortgage Lending and Non-Borrower Household Income
56
Evaluating Credit Risk for EIHs: Loan Performance
Summary Statistics, 2007/2009/2011/2013 American Housing Survey,
Household (HH) w/ mortgages
Census Region
Nuclear
Non EIH
Shared
Extended
Income HH
Northeast
16.4%
19.2%
19.4%
Midwest
25.1%
22.0%
21.8%
South
36.9%
21.7%
35.6%
23.3%
33.1%
25.7%
West
Max education of
borrower/spouse
a) Less than HS
b) HS Grad
c) Some College
d) BA or Beyond
Avg age of
borrower/spouse
a) Under 35
b) 35 - 49
c) 50 - 64
d) 65 +
Nuclear
6.2%
20.5%
30.3%
43.1%
Nuclear
21.1%
39.2%
28.9%
10.9%
Non EIH
Shared
8.7%
24.7%
31.3%
35.3%
Non EIH
Shared
5.8%
44.2%
43.5%
6.5%
Extended
Income HH
14.4%
30.1%
31.3%
24.2%
Extended
Income HH
19.1%
30.4%
37.6%
12.9%
December 2015 | Mortgage Lending and Non-Borrower Household Income
57
Evaluating Credit Risk for EIHs: Loan Performance
Logistic regression on move-outs between 2011 and 2013, AHS.
December 2015 | Mortgage Lending and Non-Borrower Household Income
58
Evaluating Credit Risk for EIHs: Loan Performance
The first regression is restricted to the 2011−13 period (when the most
accurate measurement of MTMLTV is available).
Borrowers in EIHs were less likely to have moved out after being
underwater than those in non-EIHs, when controlling for other factors
that would be looked at in a mortgage application (credit-related
variables and DTI based on the borrower’s income).
This effect was strong and statistically significant. We also see that EIHs
are less likely to move out even when they have equity and presumably
have a prepay option.
Also, non EIH shared household performance (underwater move-out
rate) is better than that of nuclear households but not as good as EIHs.
December 2015 | Mortgage Lending and Non-Borrower Household Income
59
Evaluating Credit Risk for EIHs: Loan Performance
This suggests two drivers of EIH improved performance:
[1] Willingness to pay when underwater due to greater attachment to
the home
[2] Economic benefit / hedging effect of the additional income
December 2015 | Mortgage Lending and Non-Borrower Household Income
60
Evaluating Credit Risk for EIHs: Loan Performance
Logistic regression on move-outs, 2007-09, 2009-11, 2011-13, AHS.
December 2015 | Mortgage Lending and Non-Borrower Household Income
61
Evaluating Credit Risk for EIHs
Expanded analysis to earlier time periods (2007/09).
AHS did not ask participants for current mortgage balance in 2007 or
2009. Thus the underwater indicator has about a 20% false positive rate,
leading to attenuation bias.
The 2007/09 results for EIHs are weaker than in 2011, particularly under
model A (fewer controls).
Hypothesis: during periods of high unemployment, household members
are less able to compensate for each other’s job loss. This diminishes or
neutralizes some of the EIH benefit on performance.
There is also a negative selection effect occurring – EIHs are
concentrated in lower income groups who were more adversely affected
by the recession.
December 2015 | Mortgage Lending and Non-Borrower Household Income
62
Evaluating Credit Risk for EIHs: Sensitivity Tests
Base
regressions
Alt 1
Alt 2
Alt 3
Alt 4
Alt 5
Alt 6
Alt 7
Underwater
Qualifying Non100% AMI
LTV
Borrowers
threshold filter applied Model A
0.141
Any adults
110%
No
(0.112)
0.439**
Any adults
110%
Yes
(0.134)
0.136
Any adults
100%
No
(0.092)
0.021
Any adults
105%
No
(0.102)
0.216*
Any adults
115%
No
(0.121)
0.148
Any adults
120%
No
(0.131)
Relatives &
0.326***
110%
No
partners
(0.119)
0.560***
Relatives only
110%
No
(0.149)
Model B
0.339***
(0.116)
0.590****
(0.139)
0.328***
(0.096)
0.223**
(0.106)
0.365***
(0.126)
0.321**
(0.136)
0.516****
(0.123)
0.601****
(0.153)
December 2015 | Mortgage Lending and Non-Borrower Household Income
63
Evaluating Credit Risk for EIHs: Sensitivity Tests
• Sensitivity tests indicate that the weaker performance of
EIHs in a stress scenario (high unemployment) could be
mitigated when either:
•
Lending is in the context of a program for low/moderate income
home buyers so that the income portfolio effect of EIHs is
diminished.
•
Non-relatives other than domestic partners are excluded from
qualification.
December 2015 | Mortgage Lending and Non-Borrower Household Income
64
Evaluating Credit Risk for EIHs: Income Stability
To measure income stability, regressions were performed on
the effect of EIH vs. non-EIH status on the risk of an income
loss of various degrees (25% and 50%) between two
consecutive AHS surveys:
• Income includes the shadow rent in any shared
household.
• Households that move away or can’t be contacted two
years later aren’t in the sample.
• Includes all households with mortgages (whether
underwater or not) so that the move-out related censoring
would be less likely to bias the results (as borrowers with
equity might move out due to an upsizing or downsizing).
December 2015 | Mortgage Lending and Non-Borrower Household Income
65
Evaluating Credit Risk for EIHs: Income Stability
Regression includes flags for whether the borrower/spouse
had rental income (i.e., from other housing units), or if they
were self-employed because borrowers with these types of
income are more likely to see drops in total income over a twoyear period, regardless of the threshold of income loss used.
Effect of EIH vs. non-EIH status on the risk of an income loss of various degrees
(25% and 50%) between two consecutive AHS surveys
Source: Fannie Mae analysis of American Housing Survey 2005-13
December 2015 | Mortgage Lending and Non-Borrower Household Income
66
Evaluating Credit Risk for EIHs: Income Stability
•
Income available to the borrower - including shadow rent - in
an EIH is roughly as stable as that of borrowers not in EIHs.
•
Slightly more stable for very large income drops (particularly
when economy is expanding and household members are
able to hedge against borrower income loss).
Effect of EIH vs. non-EIH status on the risk of an income loss of various degrees
(25% and 50%) between two consecutive AHS surveys, and all years.
Source: Fannie Mae analysis of American Housing Survey 2005-13
December 2015 | Mortgage Lending and Non-Borrower Household Income
67
Evaluating Credit Risk for EIHs: Income Stability
2005 – 2013 on average borrower incomes in EIHs tend to rise
while non-borrower incomes fall, reducing mean household
income in an EIH.
Borrower’s resources including 30% shadow rent mutes the
non-borrower income losses, and average EIH borrower’s income
stream goes up in the subsequent two years.
Change in income by household type, by household members
Source: Fannie Mae analysis of American Housing Survey 2005-13
December 2015 | Mortgage Lending and Non-Borrower Household Income
68
Evaluating Credit Risk for EIHs: Income Stability
EIH income stability looks better when the economy is
expanding (particularly 2011−13), because we see a hedging
effect where non-borrower income is more likely to go up when
borrower income is falling, for all periods except for 2007-09.
Relative change in borrower income for EIHs 2005 to 2013
Source: Fannie Mae analysis of American Housing Survey 2005-13
December 2015 | Mortgage Lending and Non-Borrower Household Income
69
Mitigating Credit Risk for EIHs
•
Slightly elevated risk for EIHs of smaller negative income
shocks that may stem from the non-borrower losing their job.
•
Risk is mitigated by capping the amount of non-borrower
income that is taken into account by the lending program
•
Trade-off of non-borrower income “quantity” vs. non-borrower
income stability or “quality.”
•
If the amount of non-borrower income that is evaluated
is no more than 67% of that of the borrower, then this
small-shock risk is neutralized even during periods of
high unemployment.
December 2015 | Mortgage Lending and Non-Borrower Household Income
70
Non Borrower Income in HomeReady™
• Households must document any non-borrower income
through standard income verification procedures.
• Non-borrower income must be at least 30% of that of the
borrower(s) to meet the EIH threshold, and non-borrower
income not considered in DU® risk analysis.
• Allows borrowers with up to 50% DTI (based on the borrower
income only) to be eligible (vs standard 45%).
• Effectively considers no more than 37% of non-borrower
income vs. the 67% cap from research.
December 2015 | Mortgage Lending and Non-Borrower Household Income
71
James H. Carr
Coleman A. Young Endowed Chair
and Professor in Urban Affairs
Wayne State University
And
Senior Fellow, Center for American Progress
At
The Urban Institute
Washington, DC
December 16, 2015
Format for the reviews:
1.
2.
3.
Most significant findings
Most surprising findings
Opportunities for future-related work
Most significant findings:
1.
2.
3.
4.
5.
Between 2010 and 2030 the share of household growth that
will be people of color is estimated to be 77 percent between
2010 and 2020, and 88 percent between 2020 and 2030.
The homeownership rate will decline from 65.1 percent in 2010
to 61.3 percent in 2030.
The absolute growth in the number of new renters will exceed
that of new owners.
The absolute number of households headed by someone age
65 or older will expand by nearly 20 million between 2010 and
2030.
Tight mortgage credit standards may exacerbate tight rental
market.
Most surprising findings:
1.
2.
3.
By 2030, the homeownership rate for African Americans is
projected to be as low as 38 percent to as high a high of 42.
(These estimates are particularly disappointing given that the
homeownership rate for Blacks reached a high of just under 50
percent in 2014.)
By 2030, the homeownership rate for Hispanic/Latinos is
expected to be nearly two percentage points less than their precrisis peak rate of nearly 50 percent.
The homeownership rate for non-Hispanic Whites is also
expected to decline (reinforcing the reality that the housing
finance system is potentially not working optimally for any
group.
Source: http://www.census.gov/housing/hvs/data/histtabs.html (Table 16)
Source: FHA NewsBlog.com
Source: http://www.pewsocialtrends.org/2011/07/26/wealth-gaps-rise-to-record-highs-between-whites-blackshispanics/
Source: Taub, Amy and Catherine Ruetschlin. The Racial Wealth Gap: Why Policy Matters. Washington, DC. March 10,
2015.
Source: Taub, Amy and Catherine Ruetschlin. The Racial Wealth Gap: Why Policy Matters. Washington, DC. March 10,
2015.
Opportunities for future research:
1.
2.
3.
If the homeownership projections are based solely on
demographic changes, adding assumptions related to a
possible improvement in access to credit would be useful.
In particular, what is the potential for homeownership
attainment by race/ethnicity ,by 2030 ,assuming 2001
underwriting guidelines are in place between 2010 and 2030?
Projecting the decline in wealth by the median household and
by race ethnicity between 2010 and 2030 could be a powerful
policy tool to encourage intransigent lawmakers to fix the
housing market.
Most significant findings:
1.
2.
3.
In a growing number of households, a substantial amount of
total income comes from additional adults other than the
homeowner/head of household or spouse (providing us new
insight on the reality of of doubled-up households).
Extended income households (EIHs) are more prevalent in lowincome and minority populations and could benefit (greatly?) by
allowing non-borrower EHI households income to be considered in
the underwriting process (particularly as it pertains to DTI).
Borrowers in EIH households are more likely to remain in their
homes even during housing market turmoil including when
their homes are underwater (But the Great Recession is
hopefully a cyclical event).
Most surprising findings:
1.
2.
3.
4.
Just about half of all shared households meet the 30%
threshold test to qualify as an EIH and the proportion is fairly
consistent across all groups.
Using an inclusive definition, 15% of households with a
mortgage are EIH and 20 % for African Americans and 24% for
Hispanics/Latinos.
Within EHI households, adult children contribute more to
household income than do parents, followed by domestic
partners and other non-relatives.
The complexity of modeling (data and assumptions) required
to determine how best to incorporate EIH households in an
automated underwriting system.
Recap and Release: Not the right path to
affordable mortgages
The government's role in all this
November 16, 2015 Barry Zigas
“…The government’s role should be to ensure a set of critical outcomes:
The broadest possible access to sustainable mortgage credit by the greatest
range of credit worthy borrowers, everywhere in the country
A deep and liquid market for mortgage backed securities to ensure reliable
investment at all times from the broadest possible range of investors at the
lowest possible cost, enabling rate locks for consumers
The choice to have a long term fixed rate mortgages at a reasonable price
for consumers who prefer them…”
Source: Zigas, Barry. Housing Wire.
Three additional needs of the housing finance system:



Leverage the current climate of historically low mortgage
interest rates, vast pools of foreclosed and distressed
properties, and greatly lowered home prices in order to
promote homeownership particularly for historically
underserved borrower groups.
Ensure an adequate supply of credit for rental housing and
pioneer new products for rental as well as rent-to-own,
shared appreciation ownership financing products and
extended-income family (EIH) households.
Provide capital to developers at favorable rates to finance
comprehensive, mixed use, community investment efforts
that include community infrastructure via a community
infrastructure bank.
Source: Humphries, Stan. Zillow.Group. 2015
Opportunities for future research:
1.
1.
2.
3.
To what extent is DTI currently an observable challenge (i.e.
reason for rejection) today?
Are there differences in the reliability of extended family EIH
income versus non-related EIH income. A survey perhaps?
To what extent could additional qualifying income from EIH
households expand homeownership by race/ethnicity?
What are the expectations for the full reach of Fannie Mae’s
Home Ready Program as currently designed?
There is currently no major housing finance
reform legislation being pursued on Congress
But
Important changes can be made and now
• Require loans securitized by Fannie Mae and Freddie Mac and Insured
by FHA to rely on the most predictive credit scoring models available
on the market including both FICO 9 and VantageScore.
• Redefine the GSE underserved markets designation to include
race/ethnicity.*
• Adjust FHA fees to reflect the fact that the agency is no longer
undercapitalized; fees should be calibrated to cover only those risks
posed to the Mutual Mortgage Insurance Fund by future borrowers.
• Clarify buy-back requirements at FHA so as to eliminate that issue as a
justification for excessive lender overlays or banks pulling out
altogether of FHA lending.
• Establish a loss reserve account within the Department of the Treasury
(not recapitalization) to cover potential future losses by Fannie Mae and
Freddie Mac.
*Zonta, Michela. Do the GSEs Meet the Credit Needs of Underserved Communities? Center for America
Progress. Washington, DC. December 2015.
Source: Detroit, MI