Policy Briefs: Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

by Kevin A. Park

339

Cityscape: A Journal of Policy Development and Research • Volume 24, Number 3 • 2022

U.S. Department of Housing and Urban Development • Ofﬁce of Policy Development and Research

Cityscape

Real and Personal: The Effect of

Land in Manufactured Housing

Loan Default Risk

Kevin A. Park

U.S. Department of Housing and Urban Development

The views expressed in this article are those of the author and do not represent the official positions or

policies of the Office of Policy Development and Research, the U.S. Department of Housing and Urban

Development, or the U.S. government.

Abstract

Ownership of manufactured housing is complicated by the distinction between homeownership and

landownership. Roughly two of five manufactured homeowners do not own the underlying land.

Traditional mortgage financing is only available for manufactured homes owned with land as real estate.

Personal loans are available for manufactured homes without land or owned as personal property but are

often more expensive.

The Federal Housing Administration (FHA) provides loan insurance for the purchase or refinance of

manufactured homes owned as either real or personal property. This paper provides an overview of the

Title I loan insurance program and compares the default risk of FHA-insured personal property loans for

the purchase of manufactured homes to similar mortgages for manufactured homes. Landownership, even

when the home is titled as personal property, makes an important difference in risk.

Introduction

Manufactured homes provide an important source of affordable housing in the United States.

The price per square foot of a typical new manufactured home is less than one-half that of a new,

traditionally site-built home, excluding land value ($52.80 vs. $109.14).

Manufactured homes

account for more than one-half of all owner-occupied housing units less than $50,000 (exhibit

Median values for 2019 from the Census Bureau’s Survey of Construction and author tabulations of the Manufactured

Housing Survey.

Park

340 Policy Briefs

1). According to the 2019 American Housing Survey, nearly 6.4 million households in the United

States (roughly one in 20 households) live in manufactured homes. Most of those households own

their home.

Exhibit 1

Owner-Occupied Property Type by Value, 2019

100

$10,000

or Less

$20,000

$30,000

$40,000

$50,000

$60,000

$70,000

$80,000

$90,000

$100,000

More than

$100,000

Property Value

Site-Built

MH House and Land

MH House Only

Other

Share of Owner-Occupied Units (%)

MH = Manufactured Housing.

Note: Value of manufactured homes includes only the value of the housing unit, even if owned with land.

Source: American Housing Survey

However, ownership of manufactured housing is more complicated than ownership of site-built

homes. Nearly one-half of households in manufactured homes own both the housing unit and

the underlying land (exhibit 2). Another 28 percent own the housing unit only and not the lot.

Traditional mortgage financing is only available for real property, meaning the homeowner must

own the land and the home must be fixed to a permanent foundation. Personal loans are available

for households who do not own, or who choose not to encumber, the land but are often more

expensive with shorter terms. According to the 2019 American Housing Survey, only 16 percent

of manufactured-home owners without land report having a loan on the unit, whereas nearly one-

third of manufactured-home-and-landowners and more than 60 percent of site-built single-family-

home owners have loans on their properties.

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

341Cityscape

Exhibit 2

Site-Built and Manufactured Housing Tenure and Mortgage Status, 2019

28.0M Units

2.2M Units

42.7

1.0

1.5

0.3

13.8

1.4

100

Site-built

Manufactured

Share of Households (%)

Own Land and House

With Debt

Own House Only

With Debt

Rent

Source: American Housing Survey

For more than five decades, the Federal Housing Administration (FHA) has provided loan

insurance to facilitate the purchase of manufactured homes not only as real property through

its flagship Section 203(b) program but also as personal property under Title I of the National

Housing Act; however, the latter program has declined in recent decades. This paper provides an

overview of FHA’s Title I manufactured housing loan program and analyzes the performance of

these loans relative to FHA-insured mortgages for the purchase of manufactured homes held as real

property. Land ownership, even when not used to secure a loan as collateral, substantially reduces

the likelihood of default.

Manufactured Housing

Under the National Manufactured Housing Construction and Safety Standards Act of 1974 (Public

Law 93-383), the U.S. Department of Housing and Urban Development (HUD) issues and enforces

standards for the design, construction, and installation of manufactured housing, preempting

state and local laws. Manufactured housing—as opposed to mobile homes, trailers, and modular

homes—is defined as a prefabricated dwelling built on a permanent chassis after June 15, 1976, in

compliance with this “HUD Code.”

Manufactured homes can be owned as either real or personal property. Jointly holding land and

a manufactured home affixed to a permanent foundation makes it nearly indistinguishable from

site-built homes from a legal perspective. Personal or “chattel” property covers only the housing

unit. Most owners of manufactured homes (64 percent) are also landowners. Only 19 percent of

Park

342 Policy Briefs

new manufactured home shipments, however, were titled as real property in 2019, according to the

Manufactured Housing Survey; by contrast, 76 percent were titled as personal property.

Ownership and title affect financing options. Traditional mortgage financing is available only for

real property. Russell et al. (2021) found that almost one-half of borrowers using personal property

loans to purchase a manufactured home leased the land. Another 24 percent lived rent free on land

owned by others, possibly family members. However, more than one-fourth (27 percent) owned

the land but still used personal property loans rather than mortgage financing.

Russell et al. (2021) also found that applications for personal property loans to purchase

manufactured homes are more likely to be denied than manufactured home mortgage

applications, which are more likely to be denied than mortgage applications for a site-built

home, even controlling for credit score. Similarly, the average annual percentage rate (APR) on

personal property manufactured housing loans is 3.6 percentage points higher than the APR for

manufactured home mortgages, which is 1.2 percentage points higher than the APR for site-built

home mortgages.

Chattel financing also is not covered by the same consumer protection laws,

including the Real Estate Settlement Procedures Act (CFPB, 2014).

The Center for Community Capital (CCC, 2020) surveyed borrowers who financed the purchase

of a manufactured home in Texas in 2018. They found that 61 percent of buyers of manufactured

homes owned the underlying land. First-time homebuyers and lower-income, African-American,

and urban homebuyers were less likely to be landowners. Among landowners, 59 percent titled

their home as personal property. Some landowners preferred to avoid encumbering the land, even

if personal loans were associated with higher interest rates. Using a personal property loan was

associated with more knowledge of the loan process and less reliance on lenders and real estate

agents for information but a greater likelihood of applying through or being referred by the seller.

Borrowers using personal loans disproportionately preferred shorter loan terms, whereas borrowers

using mortgage loans preferred lower closing costs and fixed interest rates.

Whether ownership of manufactured homes includes landownership is immensely consequential

for evaluating its effects on wealth. Jewell (2003) and Boehm and Schlottmann (2008) found that

ownership of a manufactured home with landownership is associated with similar average but

more volatile price appreciation compared with site-built homes. Similarly, the Federal Housing

Finance Agency (FHFA, 2018) constructed a repeat-sales index of only manufactured housing

transactions and found price trends similar to those for other forms of housing. Manufactured

housing prices rose 120 percent between 1995 and 2018, compared with 140 percent for other

forms of housing, although manufactured housing prices fell more during the Great Recession.

However, the FHFA index is based on mortgage acquisitions by Fannie Mae and Freddie Mac (i.e.,

manufactured homes owned with land as real property). By contrast, Jewell (2003) and Boehm

and Schlottmann (2008) found that ownership of manufactured homes without landownership is

associated with depreciation relative to house price changes among site-built homes. Manufactured

housing “in which the household does not own the lot is not an investment in any sense. It should

be thought of as a type of consumer durable” (Boehm and Schlottmann, 2008: 200).

Russell et al. (2021) also found substantial bunching at APR spreads just below the thresholds under the Home

Ownership and Equity Protection Act that would require additional disclosure requirements.

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

343Cityscape

The effect of title and landownership is further complicated by the correlation with construction

status. The Home Mortgage Disclosure Act does not identify new construction but, since 2018,

has identified whether a manufactured housing loan is for the house and land or the house alone.

Nearly 225,000 loans for the purchase of manufactured homes were reported in 2018 and 2019, of

which the majority (55 percent) are for the housing unit and land. According to the Manufactured

Housing Survey, this number of purchase loans for manufactured homes and land is several times

the number of new manufactured home sales titled as real property, suggesting that most of those

loans are for existing manufactured homes. For comparison, the number of purchase loans for

manufactured homes without land is close to the number of new manufactured home sales titled

as personal property, suggesting that most of those loans are for new housing units. A premium for

new consumer durables has long been discussed in economic theory, particularly for automobiles

(e.g., Akerlof, 1970; Bond, 1982; Cramer, 1958), and has been empirically extended to new homes

(Coulson, Morris, and Neill, 2019). Any price premium for new durables is lost as soon as they are

bought, appearing as excessive depreciation. Therefore, some of the depreciation associated with a

lack of landownership may be the loss of the premium for new manufactured homes.

The difference in appreciation affects loan performance, given the importance of equity as a

key determinant of default, which is well established in the economic literature on mortgages

(e.g., LaCour-Little, 2008; Quercia and Stegman, 1992). There has been a paucity of research

specifically on the loan performance of manufactured homes (Lawrence, Smith, and Rhoades,

1992). Myers and Forgy (1963) demonstrated the value of developing credit risk scoring systems

using discriminant analysis of a sample of conditional sales contracts on mobile homes. Notable

factors include whether the borrower has a bank account, unsatisfactory credit references, history

of repossessions and bankruptcies, the unpaid balance and downpayment, other terms of the sales

contract, the width of the mobile home, and whether it was new or used. Lawrence, Smith, and

Rhoades (1992) used logistic regression to estimate the likelihood that a sample of loans active

in 1988 would default the following year. Borrowers’ current equity is estimated by assuming a

depreciation rate of 10 percent in the first year and 5 percent in subsequent years. Lower equity,

higher initial payment-to-income ratios, a history of missed payments, smaller loans, older

borrowers, and higher statewide unemployment rates are associated with greater risk.

Loans for manufactured homes experienced a wave of reckless lending and subsequent defaults in

the 1990s that presaged the subprime mortgage crisis. More than 75,000 owners had their homes

repossessed in 2000. Manufactured housing lender Conseco, Inc., which accounted for most loan

originations that year, filed for bankruptcy in 2002 (CFPB, 2014).

Genz (2001) noted that loans

for manufactured homes default at a rate four times higher than conventional loans, which would

be “unthinkable in the world of ‘real housing’ finance, but somehow tacitly accepted for people

who make their home in a factory-built unit” (Genz, 2001: 403). However, there has not been

an explicit and rigorous comparison of loan performance for manufactured housing as real and

personal property.

Conseco was also a major Title I lender in the early 1990s (see exhibit 3).

Park

344 Policy Briefs

Title I

HUD plays an important role in manufactured housing in the United States. In addition to

regulating the construction and safety standards of manufactured homes, HUD helps finance the

purchase and refinance of manufactured homes through FHA. FHA has always insured loans on

manufactured and mobile homes as real property if they meet its Minimum Property Standards

and local building and land use regulations. However, the Housing and Urban Development Act

of 1969 (Public Law 91-152) also authorizes FHA to insure personal property loans secured by

manufactured housing. Borrowers are not required to own the land but must have a land lease that

does not expire for at least 3 years after origination and afterward is renewable on an annual basis.

Insurance is authorized under Title I of the National Housing Act rather than Section 203(b) and

financed through the General Insurance Fund rather than the Mutual Mortgage Insurance Fund.

CCC (2020) reported that FHA accounted for 14 percent of buyers of manufactured homes in

Texas in 2018 using chattel financing, compared with 22 percent using mortgages. However, Title

I volume is a fraction of what it was three decades ago. FHA-insured chattel loan originations

collapsed from more than 26,000 in 1991 to fewer than 225 in 1999 (exhibit 3). Originations

rebounded to nearly 2,000 a year in the early 2000s. Nevertheless, FHA has insured nearly 30

times more loans for purchase or refinance of manufactured housing titled as real estate through

its main 203(b) mortgage insurance program than titled as personal property through Title I since

1998. Two major reasons for the decline are a limited secondary market and low loan limits.

Exhibit 3

Title I Lenders and Loan Volume, 1985–2020

100

7,500

15,000

22,500

30,000

1985

1990

1995

2000

2005

2010

2015

2020

Lenders

Loan Originations

Conseco Finance Corp

Vanderbilt Mortgage Finance

Logan Laws Financial Corp

21st Mortgage Corp

All Others

Lenders

Source: Federal Housing Administration administrative data

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

345Cityscape

Ginnie Mae has facilitated securitization of Title I loans since the 1970s but significantly curtailed

operations after suffering losses. Title I insurance covers only 90 percent of the claim amount,

compared with full coverage under 203(b). In addition, FHA’s liability had been capped at 10

percent of a lender’s aggregate disbursement, known as the reserve account. Those limits became

binding when manufactured housing suffered waves of defaults in the 1980s and 1990s. Nearly

29 percent of Title I loans originated between 1995 and 2000 terminated with a claim (exhibit 4).

Although co-insurance was meant to align incentives between FHA and lenders, it created a moral

hazard problem: “[A]s lenders’ portfolios experienced losses [beyond the 10-percent aggregate

disbursement cap], they were incented to make more loans in order to increase the amount of

claims payments for which they were eligible” (Frenz, 2006: n.d.).

Exhibit 4

FHA Manufactured Housing Claim Rate by Cohort, 1998–2020

1985

1990

1995

2000

2005

2010

2015

2020

Claim Rate (%)

Title I

203b

Source: Federal Housing Administration administrative data

Those losses were pushed onto Ginnie Mae, which guarantees an issuer will make payments

on securities backed by Title I loans. Ginnie Mae placed a moratorium on new manufactured

housing securities issuers after 12 issuers defaulted between 1986 and 1988. Another 10 issuers

defaulted in the 1990s, resulting in at least $514 million in losses for Ginnie Mae (Government

Accountability Office, 2007). The FHA Manufactured Housing Loan Modernization Act of 2008

removed the portfolio cap to provide loan-level insurance coverage similar to the 203(b) program.

Nevertheless, Ginnie Mae requires issuers to have a minimum net worth of $10 million plus 10

percent of outstanding obligations to participate in its Manufactured Home Program, compared

with only $2.5 million plus 0.2 percent of outstanding obligations for the Single-Family Program.

The second major reason for declining Title I origination volume is that loan limits failed to keep

pace with the rising costs of manufactured homes (exhibit 5). Before 1983, there were separate

loan limits for one- and two-section manufactured homes. Congress then raised the limit to

Park

346 Policy Briefs

$40,500 but removed the distinction by number of sections. The Housing and Community

Development Act of 1992 again increased the loan limit for a manufactured home to $48,600,

above the average sales price of two-section homes at the time. However, by 2001 the average

sales price of all new units exceeded the loan limit. The FHA Manufactured Housing Loan

Modernization Act increased the limit to $69,678, roughly the average price of a new unit. The

2008 Act also mandated annual indexing of loan limits. However, the average sales price of new

units fell in the years shortly following the Great Recession. Rather than lower the loan limit in

proportion to the decline in price, FHA kept them unchanged. By 2016, the average sales price had

again risen above the Title I loan limit, but limits were not increased to keep pace.

Exhibit 5

Title I Loan Limit and Manufactured Housing Sales Prices, 1980–2020

$20,000

$40,000

$60,000

$80,000

$100,000

$120,000

1980

1985

1990

1995

2000

2005

2010

2015

2020

Title I Loan Limit or Average Sales Price

All

Single Section

Double Section

Title I Loan Limit

Note: Manufactured housing unit only.

Sources: Public Law 98-181; FHA Title I Letters 424, 480; Manufactured Housing Survey

Title I requires an upfront mortgage insurance premium of 2.25 percent, higher than the 1.75

percent in the 203(b) program, and an annual insurance premium of 1 percent, also generally

higher than the 203(b) program. The maximum loan term for a loan on a manufactured home

is 20 years and 32 days, less than the 30 years common in the 203(b) program. The housing

payment, which includes taxes and lot rent, cannot exceed 33 percent of effective borrower

income, and total debt payments cannot exceed 45 percent.

There is no minimum credit score, but the lender must pull a score if available and examine the

borrower’s overall pattern of credit behavior. In addition, the maximum loan-to-value (LTV) ratio

is lower (90 percent) if the borrower has a credit score lower than 500. Otherwise, the maximum

LTV ratio is 95 percent, less than the 96.5-percent maximum in the 203(b) program. However, the

comparison is not apples to apples. The Title I denominator “value” is the sum of 130 percent of

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

347Cityscape

the wholesale price plus eligible itemized options, sales tax, transportation cost, cost of installing

appurtenance and air conditioning or heat pump, and financeable fees and charges. The upfront

mortgage insurance premium can be financed but counts toward the loan limit, whereas it does not

count in the 203(b) program. Secondary financing is not permitted in Title I.

The Title I program has gradually suffocated from the lack of a secondary market, failure to

increase loan limits with the rising cost of manufactured homes, and antiquated paper-based

program procedures. Fewer than 35 loans were originated under the program in 2020, and roughly

8,000 loans were still active at the end of 2021.

Data

This study used administrative data from FHA’s 203(b) and Title I loan insurance programs to

analyze the performance of personal property loans relative to comparable mortgages for the

purchase of manufactured homes originated between 2012 and 2018. Roughly 2.5 percent of

observations were dropped due to incomplete information, mostly credit score. The resulting

sampling frame consists of nearly 127,000 observations, of which roughly 3,900 (3 percent) are

Title I loans. Exhibit 6 provides descriptive statistics of the data.

Exhibit 6

Descriptive Statistics (1 of 2)

Item Title I

203(b)

Matched All

Observations 3,944 9,190 122,831

Weighted 3,944 3,944

Estimated Title I Probability (%) 18.4 15.4 2.6

Loan Status (%)

Active 83.9 77.8 74.6

Default-Claim 7.8 4.6 2.1

Prepaid 8.3 17.7 23.3

Sales Price ($2021) 57,910 144,756 151,129

(14,139) (44,291) (96,124)

New Construction (%) 91.9 91.9 12.6

Loan Amount ($2021) 55,725 133,647 144,938

(13,698) (41,291) (69,797)

LTV Ratio (%) 96.8 92.7 96.5

(9.2) (7.7) (5.2)

Loan Term (Months) 239 359 357

(6) (11) (21)

Interest Rate (%) 7.3 4.5 4.5

(0.5) (0.5) (0.6)

Rate Spread 3.3 0.5 0.3

(0.5) (0.5) (0.5)

Credit Score 660 654 683

(55) (42) (48)

Income ($2021) 3,547 4,255 4,915

(1,568) (1,823) (2,442)

Park

348 Policy Briefs

Exhibit 6

Descriptive Statistics (2 of 2)

Item Title I

203(b)

Matched All

Housing Ratio (%) 21.7 25.4 24.0

(8.6) (8.0) (8.8)

Debt Ratio (%) 38.3 40.5 39.0

(14.1) (8.7) (9.4)

Coborrower (%) 25.2 31.7 34.0

Age 42.9 43.5 39.9

(17.3) (14.0) (13.8)

First-Time Buyer (%) 60.3 51.6 79.3

Race

White 72.5 67.1 76.9

African-American 9.7 13.3 3.3

Hispanic 4.3 3.4 11.2

Other 2.4 2.8 1.6

Not Available 11.0 13.4 6.9

Rural-Urban (%)

Urban Center 32.5 26.4 33.5

Urban Commuting 28.8 29.3 31.1

Micropolitan Area 19.7 21.5 19.6

Small Town 12.0 14.5 9.1

Rural 7.1 8.2 6.7

Year (%)

2012 14.4 14.4 9.7

2013 11.7 11.7 9.7

2014 10.4 10.4 11.1

2015 16.0 16.0 14.9

2016 20.6 20.6 16.5

2017 16.8 16.8 18.2

2018 10.1 10.1 19.9

State (%)

Other 12.9 12.9 65.6

Alabama 9.2 9.2 1.2

Arkansas 4.4 4.4 0.8

Kentucky 7.2 7.2 2.6

Louisiana 9.9 9.9 2.6

Mississippi 3.7 3.7 0.4

North Carolina 9.7 9.7

6.7

Oklahoma 3.3 3.3 1.6

South Carolina 6.0 6.0 2.6

Tennessee 7.1 7.1 3.6

Texas 17.8 17.8 8.5

Virginia 4.6 4.6 2.7

West Virginia 4.2 4.2 0.9

LTV = loan-to-value.

Note: Standard deviations shown in parentheses.

Source: Federal Housing Administration administrative data

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

349Cityscape

Selection

Before comparing loan performance, this study analyzed differences between manufactured home

mortgages and personal property loans. The results of this analysis were used to reduce differences

that might confound estimating the relative risk of Title I loans.

Methodology

This study used a binomial logistic regression to estimate the likelihood of using Title I as opposed

to 203(b) insurance.

Borrower level covariates X used to predict program and property type include:

Credit Score The decision credit score of the borrower (i.e., the median of the scores from the

three credit bureaus, or minimum if fewer than three are available). If multiple

borrowers are present, then the lowest decision score is used.

Income The natural logarithm of the inflation-adjusted monthly income (with

Winsorization

to limit the influence of outliers) used in underwriting.

Co-Borrower A binary indicator of whether more than one borrower is on the loan.

Age The age of the primary borrower.

Race Categorical variables reflecting the race and ethnicity of the primary borrower.

First-Time Buyer A binary indicator of whether the borrowers are first-time homebuyers.

Housing market conditions Z at the time of closing include the following:

Unemployment The monthly county unemployment rate reported by the U.S. Bureau of Labor

Statistics.

MH Share The number of manufactured housing units as a share of the total housing stock

estimated in the most recent 5-year American Community Survey before loan

origination.

RUCA The rural-urban commuting area (RUCA) 2010 classification of the property

ZIP Code developed by the U.S. Department of Agriculture. The classification is

condensed into five categories: Urban Center (RUCA code 1), Urban Commuting

Area (2–3), Micropolitan Area (4–6), Small Town (7–9), and Rural (10).

Logistic regression estimated using the logit command in Stata/SE 15.0.

Winsorization refers to top- and bottom-coding values at given percentiles. In this case, the top and bottom one percent of

income values are replaced with 99

and 1

percentiles, respectively.

Park

350 Policy Briefs

Fixed effects are included for the year the loan was closed

and the 12 states with the most Title I

manufactured housing loans

. All other states are grouped in the reference category.

Findings

The first column of exhibit 7 presents the results of the logistic regression estimating the likelihood

of using a Title I personal property loan for the purchase of a manufactured home relative to

a 203(b) mortgage. Borrowers with lower income and lower credit scores were more likely to

use Title I. Older homebuyers were also more likely to use Title I, whereas Hispanic and first-

time buyers were less likely. Buyers in the most urbanized areas were more likely to use Title I,

and 203(b) mortgages were more common within micropolitan areas and commuting zones of

metropolitan areas.

Exhibit 7

Selection Model (1 of 2)

(1) (2) (3)

New Construction 66.390*** 76.810***

(4.5420) (7.3890)

Credit Score 0.988*** 0.993*** 0.993***

(0.0005) (0.0005) (0.0007)

Income (Log) 0.145*** 0.082*** 0.089***

(0.0073) (0.0049) (0.0074)

Co-Borrower 1.017 0.749*** 0.791***

(0.0458) (0.0361) (0.0527)

Age 1.012*** 1.007*** 1.008***

(0.0013) (0.0015) (0.0021)

Race

African-American 1.395*** 0.750*** 0.699**

(0.0900) (0.0557) (0.0766)

Hispanic 0.560*** 0.608*** 0.612***

(0.0470) (0.0557) (0.0720)

Other 2.022*** 1.539** 1.021

(0.2440) (0.2260) (0.2490)

Not Available 1.810*** 1.324*** 1.649***

(0.1060) (0.0893) (0.1430)

First-Time Buyer 0.342*** 0.511*** 0.461***

(0.0130) (0.0236) (0.0294)

Housing Market

Unemployment Rate 1.071*** 1.087*** 1.019

(0.0104) (0.0130) (0.0213)

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

351Cityscape

Exhibit 7

Selection Model (2 of 2)

(1) (2) (3)

MH Share of Stock 1.019*** 1.006* 0.999

(0.0024) (0.0027) (0.0054)

Land Value (Log) 76.810***

(7.3890)

Land Share of Value 1.610***

(0.2320)

Home Sales Rate 0.019***

(0.0174)

Change in House Prices 0.985***

(0.0022)

Mortgage Delinquency Rate 0.989

(0.0097)

RUCA

Urban Commuting 0.592*** 0.549*** 0.548***

(0.0285) (0.0305) (0.0375)

Micropolitan Area 0.824*** 0.692*** 0.651***

(0.0444) (0.0431) (0.0594)

Small Town 0.918 0.714*** 0.650*

(0.0602) (0.0562) (0.1140)

Rural

0.992 0.708*** 0.586*

(0.0781) (0.0659) (0.1250)

Observations 126,775 126,775 82,399

χ² 6793*** 8103*** 4543***

AIC 25,423 17,260 9,136

AIC = Akaike information criterion. MH = manufactured housing. RUCA = rural-urban commuting area codes.

Notes: State and year ﬁxed effects not shown. Statistically signiﬁcant at the * 0.050 ** 0.010 *** 0.001 level. Robust standard errors shown in parentheses.

Source: Federal Housing Administration administrative data

The second column in exhibit 7 includes a binary indicator of construction status. Most (88

percent) Title I loans were for the purchase of new manufactured homes, whereas most (81

percent) 203(b) mortgages for manufactured homes were for the purchase of existing homes

(exhibit 8). Even controlling for other characteristics, purchasing a new rather than existing

manufactured home increased the odds of using Title I by a factor of 66.

Park

352 Policy Briefs

Exhibit 8

FHA Manufactured Housing Lending by Construction Status, 1998–2020

123,087 New

19,858 New

527,143 Existing

2,589

Existing

100

203(b)

Title I

Share of Loan Originations (%)

Source: Federal Housing Administration administrative data

The unemployment rate and manufactured share of the housing stock were correlated with

increased use of Title I. However, the third column of exhibit 7 adds additional housing market

conditions, which causes both of those variables to lose statistical significance. The additional

factors were not available for more than one-third of observations, reflecting the general lack of

data in many rural areas. For those observations with complete information, higher land values

were associated with increased use of Title I, whereas hot housing markets (home sales and house

price appreciation) were associated with more 203(b) mortgages.

Propensity Score Matching

This study used propensity score matching to control for observable differences in homebuyers

titling their manufactured homes as personal and real property. The propensity score is the log

odds of using a Title I derived from the first logistic regression specification shown in exhibit 7.

This paper stratified the matching process by year, state, and construction status. The analysis

matched each Title I loan for a new manufactured home to five 203(b) mortgages for new

manufactured homes in the same state and year using nearest-neighbor matching with

replacement.

In addition, the study matched Title I loans for existing manufactured homes to 10

203(b) mortgages for existing manufactured homes.

Matching reduced the sample size to 13,134 loans but substantially improved the overlap in

borrower characteristics. Exhibit 6 shows how matching and weighting observations reduced

differences in borrower characteristics and location, which allowed any differences in default risk

Propensity score matching executed using the psmatch2 command in Stata from Leuven and Sianesi (2003).

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

353Cityscape

to be more specifically associated with the type of loan. However, differences remain that may

affect relative loan performance. For example, Title I borrowers continued to be lower income

than 203(b) borrowers. Matching did not meaningfully change the higher interest rate associated

with Title I loans: the raw difference of 2.83 percentage points narrows slightly to 2.76 percentage

points after matching but remains statistically significant.

Default

Default in this analysis is defined as an insurance claim dated to the start of the delinquency

episode. This outcome definition addresses differences in the claim and property disposition

processes between the two programs. Nearly 8 percent of Title I loans originating in the study

period have terminated in a claim, compared with roughly 2 percent of 203(b) manufactured home

mortgages. After propensity score matching, the claim rate of 203(b) mortgages increases to nearly

5 percent, still below the Title I default-claim rate.

Methodology

A Cox proportional hazard model estimates the additional default risk associated with Title I loans as

where λ

indicates an unspecified baseline hazard.

Loan performance is censored at the end of

2019 to avoid problems arising from the COVID-19 pandemic.

Exhibit 9 shows the cumulative

hazard of default by loan type before and after propensity score matching. The results show that

the matching reduces the higher default hazard associated with Title I loans.

The coefficient of interest δ captures the difference in default risk associated with Title I loans

relative to 203(b) mortgages. Borrower covariates X and housing market conditions Z are the

same first specification of the selection model. The study analyzes the effects associated with the

following additional risk factors:

Housing Ratio Total housing payments, including the amount of lot rent for Title I borrowers,

relative to borrower income. Often referred to as the “front-end” debt-to-

income (DTI) ratio.

Debt Ratio Total fixed payments, including housing and all other debt, relative to

borrower income. Often referred to as the “back-end” DTI ratio.

New Construction A binary indicator of whether the housing unit is new construction.

Rent Lot A binary indicator of whether a Title I borrower pays lot rent.

Cox hazard is estimated using the stcox command in Stata/SE 15.0.

Wong (2021) reports that owners of manufactured homes were more likely to be behind in their housing payments during

the COVID-19 pandemic. Most personal property loans for manufactured homes were not covered by CARES Act relief

provisions.

Park

354 Policy Briefs

Land Tenure A categorical variable indicating the type of land tenure, including (1)

ownership, (2) leased private property, (3) leased park community, or (4) other.

Housing and debt ratios are also Winsorized to limit the influence of outliers.

Exhibit 9

Cumulative Default Hazard

0.00

0.05

0.10

0.15

0.20

Cumulative Default Hazard

Months Since Origination

Title I

203(b) Matched

203(b)

Source: Federal Housing Administration administrative data

Findings

Exhibit 10 presents the results of the Cox proportional hazard model. The first specification shows

that the baseline difference is default risk before propensity score matching and with no covariates.

Title I loans are associated with a risk of claim-default nearly three times higher than 203(b)

mortgages. That difference narrows after propensity score matching. The second column of exhibit

10 shows that Title I loans are associated with a 56-percent increase in the risk of default relative to

203(b) mortgages with similar borrower characteristics.

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

355Cityscape

Exhibit 10

Default Hazard Ratios

(1) (2) (3) (4) (5) (6) (7)

Title I 2.995*** 1.557*** 1.430*** 1.671*** 1.687*** 1.465** 1.172

(0.179) (0.161) (0.146) (0.205) (0.212) (0.203) (0.213)

Credit Score 0.995*** 0.996*** 0.996*** 0.996*** 0.996***

(0.001) (0.001) (0.001) (0.001) (0.001)

Income (Log) 0.772* 1.199 1.232 1.090 1.187

(0.085) (0.213) (0.227) (0.204) (0.224)

Co-Borrower 0.762* 0.680* 0.682* 0.689* 0.692*

(0.101) (0.113) (0.113) (0.114) (0.114)

Age 1.016*** 1.016*** 1.016*** 1.016*** 1.017***

(0.004) (0.004) (0.004) (0.004) (0.004)

Race

African-American 0.683* 0.653** 0.654** 0.662** 0.666*

(0.108) (0.105) (0.105) (0.106) (0.107)

Hispanic 0.482** 0.462** 0.462** 0.457** 0.472**

(0.133) (0.129) (0.129) (0.128) (0.132)

Other 0.444* 0.422* 0.423* 0.438* 0.429*

(0.166) (0.158) (0.158) (0.163) (0.160)

Not Available 0.785 0.782 0.784 0.780 0.786

(0.189) (0.192) (0.192) (0.191) (0.193)

First-Time Buyer 1.536*** 1.524*** 1.521*** 1.491*** 1.456**

(0.183) (0.182) (0.182) (0.178) (0.174)

Housing Ratio 1.024** 1.026** 1.020* 1.023*

(0.009) (0.009) (0.009) (0.009)

Debt Ratio 1.005 1.005 1.005 1.004

(0.005) (0.005) (0.005) (0.005)

New Construction 0.864 0.879 0.860

(0.127) (0.130) (0.128)

Rent Lot 1.309*

(0.162)

Land Tenure

Leased Park

Community

1.794**

(0.336)

Leased Private

Property

1.458*

(0.237)

Other 1.586

(0.916)

Observations 126,775 13,134 13,134 13,134 13,134 13,134 13,134

χ² 338*** 18*** 190*** 204*** 204*** 212*** 228***

AIC 64,812 8,173 8,120 8,113 8,114 8,112 8,110

AIC = Akaike information criterion.

Note: Unemployment rate, manufactured share of housing stock, and state and year ﬁxed effects not shown.

Source: FHA administrative data

Park

356 Policy Briefs

The third column further includes the borrower characteristics used in the selection model as

covariates. Higher credit scores, higher income, and co-borrowers are associated with a lower risk

of default. Minority borrowers are associated with a lower risk of default, whereas older and first-

time homebuyers are associated with higher risk. Including those characteristics reduces the hazard

ratio associated with Title I to 43 percent; that is, approximately two-thirds of the baseline risk

associated with Title I loans (first column) can be explained by the characteristics of the borrowers

that program serves. Exhibit 9 illustrates how the cumulative default hazard of 203(b) mortgages

increases after matching to Title I borrowers. The fourth column of exhibit 10 introduces new risk

factors not included in the selection model. The ratio of required housing payments to income

is associated with an increase in the risk of default, but the overall debt-to-income ratio is not

statistically significant. The indicator of new construction (fifth column of exhibit 10) is also not

significant. Including those risk factors increases the risk associated with Title I.

The sixth column shows the results of including the indicator of paying rent. Roughly one-fourth

of Title I borrowers own their land, 46 percent are on leased private property, and only 23 percent

are in a mobile home park (exhibit 11). Overall, nearly two-thirds do not pay a lot rent. Title I

borrowers without a lot rent payment are associated with a 46-percent increase in the likelihood of

default relative to a 203(b) mortgage borrower. Title I borrowers who pay a lot rent are associated

with an additional 31-percent increase in default risk, which is statistically significant at the

5-percent level, or 92 percent higher than the risk associated with 203(b) mortgage borrowers.

Notably, that estimated effect exists while controlling for the amount of the lot rent, which is

incorporated into the housing payment ratio.

Exhibit 11

Title I Land Tenure and Rent

Own

Leased Private

Property

Leased Park

Community

Other

Share of Title I Loans (%)

No Rent

Land Rent

Source: FHA administrative data

The lot rent is the amount at time of underwriting. It does not reﬂect subsequent changes in lot rent.

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

357Cityscape

The final column of exhibit 10 replaces the lot rent indicator with the land tenure classification. The

reference group in the typology is Title I borrowers who own the underlying land. Those borrowers

are associated with a 17-percent increase in the likelihood of claim-default relative to 203(b)

mortgage borrowers; however, the analysis cannot reject the null hypothesis that the default risk is

equivalent to a manufactured home purchase mortgage. Title I borrowers who are on leased private

property are associated with a 46-percent increase in the default risk relative to Title I landowners (71

percent relative to 203(b) mortgages), which is significant at the 5-percent level. Title I borrowers in

leased park communities are associated with a 79-percent increase in default risk (more than twice

the risk of 203(b) mortgages). Exhibit 12 shows the results of an additional specification with select

combinations of construction status, lot rent payment, and land tenure type. The results illustrate that

Title I landowners are not significantly higher risk than similar homebuyers using 203(b) mortgages,

whereas Title I borrowers renting lots in mobile home parks are substantially higher risk.

Exhibit 12

Claim Hazard Ratios

0.5

1.0

1.5

2.0

2.5

3.0

No Rent

Rent

203(b)

New

Title I

Existing

Own Lot

Leased Private Property

Leased Park

Community

Other

Title I New

Hazard Ratio

Claim

Claim with Prepayment Risk

Note: Error bars indicate 95-percent conﬁdence interval.

Source: FHA administrative data

The study used a Cox proportional hazard model to understand the causal effect of property and

program type on loan performance (Allison, 2018; Austin, Lee, and Fine, 2016). However, exhibit

13 shows the results of comparable Fine-Gray subhazard models that treat prepayment (i.e.,

termination without insurance claim) as a competing risk.

Title I loans are substantially less likely

to prepay, possibly reflecting the difficulty in refinancing personal property loans on manufactured

homes (Goodman and Neal, 2021; Russell et al., 2021)—a circumstance that prolongs the

exposure of Title I loans to the risk of default. Therefore, the estimated risk associated with Title

I loans in the Fine-Gray models shown in exhibit 13 is higher than the comparable estimates

FHA requires a lease of at least 3 years for Title I loans in leased park communities. However, an additional speciﬁcation

(not shown) did not ﬁnd a statistically signiﬁcant dierence in hazard ratios for such loans before and after 3 years.

Fine-Gray model estimated using the stcrreg command in Stata/SE 15.0.

Park

358 Policy Briefs

in the Cox models shown in exhibit 10. Nevertheless, the final specification of exhibit 13 and

the additional specification in exhibit 12 confirm that Title I landowners are not associated with

significantly higher risk than 203(b) mortgage borrowers.

Exhibit 13

Default Hazard with Prepayment Risk

(1) (2) (3) (4) (5) (6) (7)

Title I 3.387*** 1.701*** 1.561*** 1.805*** 1.821*** 1.571** 1.264

(0.202) (0.176) (0.161) (0.226) (0.233) (0.222) (0.233)

Credit Score 0.995*** 0.996*** 0.996*** 0.995*** 0.996***

(0.001) (0.001) (0.001) (0.001) (0.001)

Income (Log) 0.746** 1.106 1.134 1.001 1.090

(0.082) (0.195) (0.208) (0.187) (0.205)

Co-Borrower 0.746* 0.667* 0.668* 0.675* 0.679*

(0.100) (0.111) (0.111) (0.112) (0.112)

Age 1.016*** 1.016*** 1.016*** 1.015*** 1.017***

(0.004) (0.004) (0.004) (0.004) (0.004)

Race

African-American 0.708* 0.680* 0.681* 0.689* 0.694*

(0.112) (0.109) (0.109) (0.110) (0.111)

Hispanic 0.495* 0.478** 0.478** 0.472** 0.489*

(0.137) (0.133) (0.133) (0.132) (0.137)

Other 0.457* 0.439* 0.439* 0.456* 0.445*

(0.170) (0.163) (0.163) (0.169) (0.166)

Not Available 0.796 0.791 0.794 0.789 0.792

(0.194) (0.196) (0.196) (0.195) (0.196)

First-Time Buyer 1.535*** 1.526*** 1.523*** 1.494*** 1.459**

(0.184) (0.183) (0.183) (0.180) (0.176)

Housing Ratio 1.021* 1.022* 1.016 1.020*

(0.009) (0.009) (0.009) (0.009)

Debt Ratio 1.005 1.005 1.005 1.005

(0.005) (0.005) (0.005) (0.005)

New Construction 0.873 0.888 0.869

(0.129) (0.132) (0.130)

Rent Lot 1.323*

(0.164)

Land Tenure

Leased Park Community 1.799**

(0.336)

Leased Private Property 1.449*

(0.235)

Other 1.662

(0.962)

Observations 126,775 13,134 13,134 13,134 13,134 13,134 13,134

χ² 419*** 26*** 221*** 233*** 233*** 242*** 258***

AIC 65,634 14,574 14,429 14,416 14,417 14,412 14,406

AIC = Akaike information criterion.

Notes: Unemployment rate, manufactured share of housing stock and state and year ﬁxed effects not shown. Statistically signiﬁcant at the * 0.050 ** 0.010

*** 0.001 level. Robust standard errors shown in parentheses.

Source: Federal Housing Administration administrative data

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

359Cityscape

Conclusion

Manufactured housing is often seen as a technological solution to the affordable housing problem.

The Manufactured Housing Improvement Act of 2000 (Public Law 106-569) states, “[M]anufactured

housing plays a vital role in meeting the housing needs of the Nation, and manufactured homes

provide a significant resource for affordable homeownership and rental housing accessible to all

Americans.”

FHA is specifically noted as an instrument for improving access to manufactured

housing. The 2000 Act calls for a review of FHA’s manufactured housing programs and “developing

any changes to such programs to promote the affordability of manufactured homes, includes

changes in loan terms, amortization periods, regulations and procedures.”

Eight years later, the

FHA Manufactured Housing Loan Modernization Act aimed to “modernize the FHA title I insurance

program for manufactured housing loans to enhance participation by Ginnie Mae and the private

lending markets.” However, a Government Accountability Office (GAO) review in 2014 found

that “HUD has not yet examined or researched the effectiveness of these loan programs because its

research has focused on other priorities” (GAO, 2014: 29).

This paper compares the performance of personal property insured under Title I with similar

mortgages for the purchase of manufactured homes insured under FHA’s flagship 203(b) mortgage

insurance program. Title I loans are more than three times more likely to default than 203(b)

mortgages. However, two-thirds of that difference is because Title I disproportionately serves older,

lower-income borrowers with lower credit scores.

The remaining difference is mostly due to land tenure. One-fourth of Title I borrowers own the

land on which their manufactured home rests. Those borrowers have approximately the same

default risk as 203(b) mortgagors. A plurality of Title I borrowers are on leased private property,

and most of them do not pay land rent. Less than one-fourth rent a lot in a mobile home park.

FHA requires land leases in park communities to have initial terms of at least 3 years, annually

renewable, with 180 days written notice before expiration if the borrower is required to move.

Nevertheless, renting land is associated with an increase in the likelihood of default, particularly if

in a mobile home park.

There are many challenges to reinvigorating FHA’s 50-year-old program of personal property loan

insurance for manufactured homes. The maximum loan amount of $69,678 in the Title I program

is less than the average cost of a new manufactured home ($81,900 in 2019), not including the

costs of transportation and installation. Ginnie Mae places additional requirements for Title I

securities issuers compared with issues of securities based on mortgages on single-family homes,

limiting the secondary market. Title I continues to rely on manual underwriting and processing of

paper case binders.

Title I loans carry a significantly higher interest rate, roughly 2.76 percentage points above the

average rate on FHA-insured manufactured home mortgages for otherwise similar borrowers.

Higher rates may be expected given the higher risk associated with personal property loans not

secured by land. However, FHA provides nearly the same loan-level insurance coverage to Title I

42 USC § 5401(a).

42 USC § 5407(a).

Park

360 Policy Briefs

loans as to 203(b) mortgages. Investors are reimbursed for 90 percent of losses, and the portfolio

limit on claims was removed by the FHA Manufactured Housing Loan Modernization Act of 2008.

Therefore, the difference may reflect a more limited secondary market for these loans.

The Manufactured Housing Improvement Act of 2000 encourages the government-sponsored

enterprises “to actively develop and implement secondary market securitization programs for the FHA

manufactured home loans and those of other loan programs, as appropriate, thereby promoting the

availability of affordable manufactured homes to increase homeownership for all people in the United

States.” The policy goal was reiterated in the Housing and Economic Recovery Act of 2008 (Public

Law 110-289), which states that Fannie Mae and Freddie Mac have a “duty to serve” manufactured

housing, including developing “loan products and flexible underwriting guidelines to facilitate a

secondary market for mortgages on manufactured homes for very low-, low- and moderate-income

families.” The Act also singles out rural housing markets, where manufactured homes are a greater

share of the housing stock. Fannie Mae and Freddie Mac have proposed pilot programs to securitize

personal property loans on manufactured homes (Fannie Mae, 2021; Freddie Mac, 2020). However,

they currently levy higher fees on manufactured home mortgages; for example, Fannie Mae charges

a 50-basis-point upfront charge to purchase mortgages on manufactured homes. Private mortgage

insurance can add another 18 to 60 basis points to the monthly cost.

Homeownership without landownership is akin to buying down rent. The upfront cost of buying

the housing unit may lower subsequent housing expenses. The potential benefits of this buydown,

however, depend on the terms used to finance the purchase and the stability of tenure after

purchase. FHA was instrumental in increasing homeownership by popularizing the long-term

amortizing mortgage that dominates the American housing finance system. It has an opportunity to

play a similar role increasing access to affordable housing by increasing the availability of personal

property loans with appropriate risk management and consumer protections.

Acknowledgments

The author would like to thank Patricia McBarron for vital assistance in understanding the Title

I program and data and Caitlin Gorback, Lariece Brown, Jessica Russell, Nora O’Reilly, William

Reeder, Kurt Usowski, and Adam Hoffberg for helpful comments and suggestions.

Author

Kevin A. Park is an economist in the Housing Finance Analysis Division of the U.S. Department of

Housing and Urban Development’s Office of Policy Development and Research. He can be reached

at kevin.park@hud.gov.

Real and Personal: The Effect of Land in Manufactured Housing Loan Default Risk

361Cityscape

References

Akerlof, George A. 1970. “The Market for ‘Lemons’: Quality Uncertainty and the Market

Mechanism,” The Quarterly Journal of Economics 84 (3): 488–500.

Allison, Paul. 2018. “For Causal Analysis of Competing Risks, Don’t Use Fine & Gray’s

Subdistribution Method,” Statistical Horizons, March 24. https://statisticalhorizons.com/for-causal-

analysis-of-competing-risks.

Austin, Peter C., Douglas S. Lee, and Jason P. Fine. 2016. “Introduction to the Analysis of Survival

Data in the Presence of Competing Risks,” Circulation 133 (6): 601-609.

Boehm, Thomas P., and Alan Schlottmann. 2008. “Is Manufactured Housing a Good Alternative for

Low-Income Families? Evidence from the American Housing Survey,” Cityscape 10 (2): 159–224.

Bond, Eric W. 1982. “A Direct Test of the ‘Lemons’ Model: The Market for Used Pickup Trucks,”

The American Economic Review 72 (4): 836–40.

Center for Community Capital (CCC). 2020. The Loan Shopping Experiences of Manufactured

Homeowners: Survey Report. Tysons Corner, VA: Freddie Mac. https://sf.freddiemac.com/resources/

manufactured-homeowners-survey-and-report-on-loan-shopping-experiences.

Consumer Financial Protection Bureau (CFPB). 2014. Manufactured-Housing Consumer Finance in

the United States. Washington, DC: Consumer Financial Protection Bureau.

https://files.consumerfinance.gov/f/201409_cfpb_report_manufactured-housing.pdf.

Coulson, Edward N., Adele C. Morris, and Helen R. Neill. 2019. “Are New Homes Special?” Real

Estate Economics 47 (3): 784–806.

Cramer, Jan Salomon. 1958. “The Depreciation and Mortality of Motor-Cars,” Journal of the Royal

Statistical Society: Series A (General) 121 (1): 18–46.

Fannie Mae. 2021. “Duty to Serve Underserved Markets Plan for the Manufactured

Housing Market.” https://www.fhfa.gov/PolicyProgramsResearch/Programs/Documents/

FannieMaeDTSPlan_2018-2021.pdf.

Federal Housing Finance Agency (FHFA). 2018. “HPI Highlights 2Q2018: Manufactured

House Price Indexes.” https://www.fhfa.gov/DataTools/Downloads/Documents/HPI_Focus_

Pieces/2018Q2_HPIFocus_N508.pdf.

Freddie Mac. 2020. “Duty to Serve Underserved Markets Plan: 2022–2024.”

https://sf.freddiemac.com/content/_assets/resources/pdf/marketing-materials/freddie-mac-

underserved-markets-plan-2018-2021.pdf.

Frenz, Michael J. 2006. “Statement of Michael J. Frenz, Executive Vice President & Chief

Operating Officer, Government National Mortgage Association (Ginnie Mae), Before the U.S.

Senate Committee on Banking, Housing, and Urban Affairs, Subcommittee on Housing and

Transportation.” https://archives.hud.gov/testimony/2006/test040406b.cfm.

Park

362 Policy Briefs

Genz, Richard. 2001. “Why Advocates Need to Rethink Manufactured Housing,” Housing Policy

Debate 12 (2): 393–414.

Goodman, Laurie, and Michael Neal. 2021. “It’s Difficult for Manufactured Home Borrowers

to Reap the Benefits of Historically Low Interest Rates,” Urban Institute September 17.

https://www.urban.org/urban-wire/its-difficult-manufactured-home-borrowers-reap-benefits-

historically-low-interest-rates.

Government Accountability Office (GAO). 2014. “Manufactured Housing: Efforts Needed to

Enhance Program Effectiveness and Ensure Funding Stability.” GAP-14-410. Washington, DC:

GAO. https://www.gao.gov/assets/gao-14-410.pdf.

———. 2007. “Federal Housing Administration: Agency Should Assess the Effects of Proposed

Changes to the Manufactured Home Loan Program.” GAO-07-879. https://www.gao.gov/products/

GAO-07-879.

Jewell, Kevin. 2003. “Manufactured Housing Appreciation: Stereotypes and Data.” Consumers

Union. https://advocacy.consumerreports.org/wp-content/uploads/2013/04/Appreciation.pdf

LaCour-Little, Michael. 2008. “Mortgage Termination Risk: A Review of the Recent Literature,”

Journal of Real Estate Literature 16 (3): 295–326.

Lawrence, Edward C., L. Douglas Smith, and Malcolm Rhoades. 1992. “An Analysis of Default Risk

in Mobile Home Credit,” Journal of Banking & Finance 16 (2): 299–312.

Leuven, Edwin, and Barbara Sianesi. 2003. “PSMATCH2: Stata Module to Perform Full

Mahalanobis and Propensity Score Matching, Common Support Graphing, and Covariate

Imbalance Testing.” Statistical Software Components S432001, Boston College Department of

Economics. https://ideas.repec.org/c/boc/bocode/s432001.html.

Myers, James H., and Edward W. Forgy. 1963. “The Development of Numerical Credit Evaluation

Systems,” Journal of the American Statistical Association 58 (303): 799–806.

Quercia, Roberto G., and Michael A. Stegman. 1992. “Residential Mortgage Default: A Review of

the Literature,” Journal of Housing Research 3 (2): 341–79.

Russell, Jessica, Nora O’Reilly, Karl Schneider, Nicolas Melton, Nick Schwartz, and Sam Leitner.

2021. Manufactured Housing Finance: New Insights from the Home Mortgage Disclosure Act Data.

Washington, DC: Consumer Financial Protection Bureau. https://files.consumerfinance.gov/f/

documents/cfpb_manufactured-housing-finance-new-insights-hmda_report_2021-05.pdf.

Wong, Kristin. 2021. Housing Insecurity and the COVID-19 Pandemic. Consumer Financial

Protection Bureau. https://files.consumerfinance.gov/f/documents/cfpb_Housing_insecurity_and_

the_COVID-19_pandemic.pdf.