Shan

2011 V39 4: pp.
743–768
DOI: 10.1111/j.1540-6229.2011.00310.x
REAL ESTATE
ECONOMICS
Reversing the Trend: The Recent

Expansion of the Reverse Mortgage Market
Hui Shan∗
Reverse mortgages allow elderly homeowners to tap into their housing wealth
without having to sell or move out of their homes. However, very few eligible
homeowners used reverse mortgages to achieve consumption smoothing un-
til recently, when the reverse mortgage market in the United States witnessed
substantial growth. In this article, I examine 1989–2007 loan-level reverse mort-
gage data and conduct three sets of analyses to better understand the demand
for reverse mortgages among elderly homeowners. First, I study the ZIP code
characteristics correlated with reverse mortgage originations. Second, I show
that recent reverse mortgage borrowers are significantly different from earlier
borrowers in many respects. Third, I investigate the reasons why the reverse
mortgage market experienced substantial growth in the mid-2000s. Combin-
ing the reverse mortgage data with county-level house price data, I find that
higher house prices lead to more reverse mortgage originations. Specifically,
the increases in house prices account for about one-third of the overall growth
in the reverse mortgage market from 2003 to 2007.
Introduction
A reverse mortgage is a loan to elderly homeowners that converts home equity
into cash income with no repayment due until the borrower dies, the home is sold
or the borrower permanently moves out of the home. Despite its potential appeal
to house-rich but cash-poor elderly homeowners who want to access home
equity while continuing to live in their homes, the use of reverse mortgages
was extremely rare among elderly homeowners prior to the early 2000s. The
reverse mortgage market then grew substantially until the financial crisis in
2008. No research thus far has rigorously examined what contributed to the
remarkable growth of the reverse mortgage market in recent years.
In this article, I examine loan-level data on reverse mortgages originated from

1989 to 2007. I first use ZIP code–level analyses to show the characteristics of
ZIP codes that are correlated with reverse mortgage take up rates. These analy-
ses shed light on who reverse mortgage borrowers are and where they are from.
∗
Federal Reserve Board of Governors, Washington, DC 20551 or hui.shan@frb.gov.

C 2011 American Real Estate and Urban Economics Association
15406229, 2011, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1540-6229.2011.00310.x by Technical University Ostrava, Wiley Online Library on [11/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
744 Shan
I then compare reverse mortgages originated in earlier years (1989–2002) to

those originated in recent years (2003–2007) with an emphasis on the signifi-
cant changes over time in borrower and loan characteristics. Lastly, I investigate
factors that may have contributed to the rapid growth of the reverse mortgage
market in the mid-2000s. My findings suggest that higher house prices induced
higher demand for reverse mortgages among elderly homeowners.
This article contributes to the literature on three fronts. First, I study a longer
sample period and conduct analyses at a more detailed level. Using a data
set spanning 18 years with ZIP code identifiers, I am able to paint a more
comprehensive picture on who reverse mortgage borrowers are and where
they are from. Second, although previous research has used loan-level data to
study the reverse mortgage market, this is the first study that systematically
shows the significant changes in borrower and loan characteristics over time.
Third, combining the reverse mortgage data with county-level house price
data, this article is also the first to empirically test the link between house price
appreciation and the demand for reverse mortgages.
The results shown in this article have important economic and policy impli-
cations for the field of aging and housing. Housing wealth is often the largest
nonpension wealth component of elderly homeowners. For example, the 2007
Survey of Consumer Finances (SCF) data suggest that, for 6.5 million home-
owners aged 62 or above, or a quarter of all homeowners in this age group,
housing wealth represents at least 80% of their total wealth. As life expectancy
continues to rise and millions of baby-boomers approach retirement age, it is
crucial for us to better understand the reverse mortgage market and the extent to
which reverse mortgages may facilitate consumption smoothing in retirement.
The rest of this article proceeds as follows. The next section reviews the existing
literature and describes how reverse mortgages work. I then introduce the data
used in this article. The following three sections show the empirical analyses
conducted in this article. The last section concludes.
Background
Overview of Existing Studies

The question of whether Americans are financially prepared for retirement has
inspired heated debates in the literature.1 When evaluating the adequacy of
retirement saving, economists and financial planners have to decide whether
1
See Skinner (2007) for a review on this topic.
The Recent Expansion of the Reverse Mortgage Market 745
housing equity should be included as consumable wealth. Because housing

is both a consumption good and an investment good, the correct treatment
of housing equity may not be obvious in the retirement saving context. For
example, although Mitchell and Moore (1998) add housing equity to household
net worth, Bernheim et al. (2000) exclude it in their calculation. More recently,
Sinai and Souleles (2008) suggest that the fraction of “consumable housing
equity” ranges from 60% to 99% for elderly homeowners depending on their
age.
The degree to which we should consider housing equity as retirement savings

depends on the degree to which elderly homeowners are willing and able
to consume their housing wealth. It is well known that many seniors prefer
staying in their homes for as long as they can.2 A series of studies by Venti
and Wise (e.g., 1989, 1990 and 2004) show that elderly homeowners do not
reduce their housing wealth in the absence of precipitating events such as the
death of a spouse or a move to a nursing home. If elderly homeowners have
strong psychological attachments to their homes, then reverse mortgages, which
generate additional income for elderly homeowners while allowing them to
continue living in their homes, may be welfare-improving for many households.
A number of studies have estimated the potential demand in the United States
for reverse mortgages.3 Although the estimates of different researchers vary
between 0.8 and 6.7 million households, the actual size of the reverse mortgage
market prior to the early 2000s was much smaller than even the lower bound of
these estimates. A wide range of factors may have prevented reverse mortgages
from becoming more popular. On the demand side, bequest motives, the desire
to self-insure against large health expenses, mobility risks and other behavioral
and psychological hurdles may be important.4 On the supply side, lenders
face various obstacles due to the complexity of reverse mortgage products and
regulatory requirements. In addition, reverse mortgage markets may suffer from
adverse selection and moral hazard problems.5
In summary, most of the previous studies estimate the potential demand for
reverse mortgages and study why the size of the market is much smaller than
2
For example, in a survey sponsored by the American Association of Retired Persons
(AARP), 95% of persons 75 and older agreed with the statement “What I’d really like
to do is stay in my current residence as long as possible” (see Bayer and Harper 2000).
3
These studies include Venti and Wise (1991), Mayer and Simons (1994), Merrill, Finkel
and Kutty (1994), Rasmussen, Megbolugbe and Morgan (1995) and Kutty (1998).
4
See Munnell, Soto and Aubry (2007), Davidoff (2010) and Michelangeli (2008).
5
See Davidoff (2006) and Davidoff and Welke (2007).
746 Shan
expected using public survey data. Among the few papers that examine loan-
level reverse mortgage data, Davidoff and Welke (2007) and Szymanoski,
Enriquez and DiVenti (2007) mainly focus on modeling the termination out-
comes of reverse mortgages; Case and Schnare (1994) and Rodda, Herbert and
Lam (2000) analyze reverse mortgages originated in earlier years. This article
complements the existing literature by looking at how recent borrowers differ
from early borrowers and what may have caused the rapid growth of the reverse
mortgage market in the mid 2000s.
Background on the Home Equity Conversion Mortgage (HECM) Program

The most common type of reverse mortgage product in the United States
is the HECM, which constitutes over 90% of the reverse mortgage market.6
The HECM program is administered by the Department of Housing and Urban
Development (HUD), and the first HECM loan was made in 1989. To be eligible
for a HECM loan, borrowers must be 62 years of age or older and occupy the
property as their principal residence. Borrowers must either own their homes
outright or have a small mortgage balance. A HECM loan is a “nonrecourse”
loan, meaning that, if the borrower (or estate) does not pay the full balance
when due, the property will be foreclosed and the borrower (or estate) will not
be personally liable for any deficiency resulting from the foreclosure.
HECM loans differ from traditional home equity loans or home equity line
of credit (HELOC) in two ways. First, a HECM loan does not have a fixed
maturity date. The loan becomes due and payable only after the borrower dies,
the property is sold or the borrower permanently moves out.7 Second, although
home equity loans and HELOCs generally require borrowers to have sufficient
income and creditworthiness, HECM loans do not have such underwriting
requirements.8 Therefore, house-rich but cash-poor elderly homeowners who
cannot obtain home equity loans may find HECM loans particularly attractive.
The amount of cash income that the borrower can receive from a HECM
loan is calculated in three steps. The first step is to determine the Maximum
6
Other reverse mortgage products include the Home Keeper program offered by Fannie
Mae and jumbo reverse mortgage loans offered by private lenders. Because of the recent
financial crisis, the private reverse mortgage market has evaporated so that HECM loans
represent nearly 100% of newly originated reverse mortgages.
7
Technically, the loan also becomes due and payable when the borrower refinances into a
new HECM loan. Although the number of refinanced HECM loans increased after 2004
when HUD reduced the mortgage insurance premium charged on an HECM refinance,
refinancing a HECM loan is still costly and refinanced loans remain a small fraction of
total HECM originations.
8
However, borrowers who have been delinquent or defaulted on federal debt may not
be eligible for HECM loans.
Claim Amount (MCA)—the lesser of the appraised value of the property or the
county-specific Federal Housing Administration (FHA) mortgage limit for a
one-family residence under section 203 (b) of the National Housing Act.9 The
second step is to determine the Initial Principal Limit (IPL) by multiplying the
MCA by a factor between zero and one. The magnitude of the factor depends
on the age of the borrower and the “expected interest rate” at the time of loan
closing.10 The expected interest rate, a proxy for the future interest rate, equals
the sum of the 10-year Treasury rate and the lender’s margin.11 The lender’s
margin is typically between 100 and 200 basis points. The principal limit factor
increases with the borrower’s age and decreases with the expected interest rate.
For example, the factor equals 0.281 for a 65-year old at a 10% expected interest
rate, and it equals 0.819 for an 85-year old at a 5% expected interest rate.12
The third step is to calculate the Net Principal Limit (NPL), which is the amount
the borrower can take as a lump sum in cash at closing, by subtracting from
the IPL the upfront costs associated with HECM loans and a set-aside for a
monthly servicing fee. The upfront costs include the initial Mortgage Insurance
Premium (MIP), origination fee and other closing costs. The initial MIP is set
at 2% of the MCA. The origination fee is capped at $2,000 or 2% of the MCA,
whichever is greater.13 The servicing fee set-aside is the present value of the
monthly servicing fee charged by the lender. The typical monthly servicing fee
is $30 or $35. Both the upfront costs and servicing fee are financed rather than
paid by the borrower out-of-pocket. Figure 1 summarizes the steps described
above in calculating the net principal limit.
9
HECM allows properties with one to four units. However, the FHA limit on MCA is
based on the Section 203 (b) limit for a one-unit dwelling for all eligible properties.
The FHA mortgage limits are usually set at 95% of the median sales prices for any
given county in a given year, although MCAs are subject to both ceiling and floor
levels, creating nationwide maximum and minimum values. For example, the ceiling
was $362,790 and the floor was $200,160 in 2007. The HECM limit on MCA was raised
to a nationwide limit by the Housing and Economic Recovery Act of 2008.
10
For married couples, only the age of the younger borrower is taken into consideration.
11
A HUD policy change in 2007 allowed adjustable-rate HECM loans to be indexed to
the London Interbank Offered Rate (LIBOR). The expected interest rate for a LIBOR-
indexed HECM loan is the 10-year U.S. dollar denominated LIBOR swap rate plus the
lender’s margin.
12
The limit factors were designed to be actuarially sound for FHA, and under certain
assumptions the loan balance would reach the MCA near the time of the borrower’s life
expectancy. See Szymanoski (1994) for detailed discussions. In 2009, HUD announced
a new set of principal limit factors for the HECM program, which lowered the principal
limits by 10%.
13
The Housing and Economic Recovery Act of 2008 established new limits on the loan
origination fee for HECM loans. The limit is the greater of $2,500 or 2% of the first
$200,000 of the MCA, plus 1% of the portion of the MCA that is greater than $200,000.
The total amount of loan origination fee may not exceed $6,000.
748 Shan
Figure 1 Calculating the amount of payment one can receive from a HECM loan.
Given the NPL, HECM borrowers have numerous options for how they draw
their loan proceeds. Under the Tenure plan, the borrower will receive equal
monthly payments from the lender for as long as the borrower lives and contin-
ues to occupy the property as his principal residence.14 Under the Term plan,
the borrower will receive equal monthly payments from the lender for a fixed
period of months selected by the borrower.15 Under the Line of Credit plan,
the borrower will receive the mortgage proceeds in unscheduled payments or
in installments, at times and in amounts of the borrower’s choosing, until the
line of credit is exhausted. In addition, the Modified Tenure and Modified Term
plans allow the borrower to combine a line of credit with a tenure plan and
a term plan, respectively. Borrowers may change their payment plan at any
time for a small administrative fee. Table 1 shows the principal limit factor, net
principal limit, the monthly loan proceeds under a tenure plan and a 10-year
term plan for a hypothetical borrower.
Finally, it is worth emphasizing that HECM loans are insured by the FHA
insurance program. Under this program, HUD insures the borrower against the
risk that the lender can no longer make the contracted payments. The insurance
program also insures the lender against the risk that the loan balance exceeds
14
This payment plan is also called a “reverse annuity mortgage” in the literature due to
its resemblance to an annuity product.
15
Note that, even though payments stop at the end of the selected term, the loan is not
due until the borrower dies or moves out of his or her home permanently.
Table 1 HECM proceeds to a hypothetical borrower.
Expected Rate Age = 65 Age = 75 Age = 85

A. Principal limit factor
0.05 0.649 0.732 0.819
0.07 0.489 0.609 0.738
0.09 0.336 0.472 0.636
B. Net principal limit
0.05 $114,188 $131,492 $150,112
0.07 $83,323 $107,715 $134,344
0.09 $67,200 $94,400 $127,200
C. Tenure plan monthly receipt
0.05 $610 $804 $1,221
0.07 $558 $791 $1,238
0.09 $436 $702 $1,184
D. 10-Year term plan monthly receipt
0.05 $1,234 $1,421 $1,622
0.07 $983 $1,271 $1,585
0.09 $687 $1,039 $1,467
Note: Assume Maximum Claim Amount (MCA) = $200,000, initial Mortgage Insurance
Premium (MIP) = $4,000, origination fee = $4,000, closing costs = $2,000, monthly
servicing fee = $30.
the property value.16 To pay for this insurance program, HUD charges a MIP.
The initial MIP, as mentioned before, is set at 2% of the MCA. The monthly
MIP is set at an annual rate of 0.5% and is charged on the outstanding balance
of an HECM loan.
Data Description
The data analyzed in this study are loan-level administrative HECM data pro-
vided by HUD. In this data set, I observe all HECM loans made from 1989 to
2007, a total of 387,999 records in the raw data. The data set has information
on the age of the borrower, the age of the co-borrower, the gender and mari-
tal status of the borrower, the appraised value of the property at origination,
the location of the property (i.e., state and ZIP code), the MCA, the expected
interest rate, the IPL, the chosen payment plan, the monthly payment amount,
the loan origination date, the loan termination date, the loan assignment date,
whether a claim was filed to HUD by the lender and the nature of the claim. In
16
For example, lenders can assign loans to HUD when the loan balance reaches 98% of
the MCA. In the event that the proceeds from the sale of the property are not sufficient
to pay the outstanding loan balance, lenders who have not assigned the mortgage to
HUD can submit a claim for insurance benefits up to the MCA.
750 Shan
Figure 2 Growth in HECM loans, 1989–2007.
100,000
80,000
Number of Loans Originated
60,000
40,000
20,000
10,000
0
1989 1991 1993 1995 1997 1999 2001 2003 2005 2007
Year
the process of data cleaning, I dropped borrowers from places other than the
continental United States because the FHA mortgage limits for these places
are very different from the rest of the country. The final sample has 384,935
observations.
Figure 2 shows the number of HECM loans originated in each year from 1989
to 2007. In the early years, only a small number of homeowners took out HECM
mortgages. In contrast, HECM loan originations grew substantially since the
early 2000s. Also, approximately 82% of the borrowers in the analysis sample
chose the Line of Credit payment plan. Only about 10% of the borrowers chose
the Tenure or Modified Tenure plans, suggesting that most of HECM loans do
not have an “annuity” component.
Because reverse mortgage borrowers are a very small fraction of the general
population, they are rarely captured by nationally representative public surveys.
This makes administrative data indispensable in studying the reverse mortgage
market. Although administrative data tend to be more accurate than most public
surveys, there are a couple of limitations associated with the data analyzed in
this study. First, I do not know very much about these borrowers beyond their
characteristics that are used in the HECM pricing model. For example, I do not
observe income or financial wealth of these borrowers, nor do I observe their

demographic characteristics such as race, education and number of children.
Second, this data set is a snapshot of the HECM loans at the end of 2007. For
borrowers who chose the Line of Credit payment plan at origination, I do not
observe when and by how much they drew down their line of credit. Moreover,
because borrowers are allowed to change their payment plans at a small cost,
it is possible that the payment plan indicated in the data set is not the original
payment plan chosen by the borrower.17
ZIP Code Characteristics and HECM Loan Origination

In lieu of studying the characteristics of the borrowers directly, I merge the
2000 Census ZIP code level data to the HECM data to study the ZIP code
characteristics that are correlated with reverse mortgage originations. The ZIP-
code-level cross-sectional regression analyses shown in this section address
who reverse mortgage borrowers are and where they are from. To the best of my
knowledge, no previous research has examined reverse mortgage originations
at the ZIP code level. Specifically, I estimate the following regression model:

HECM Originations 1995–2005
= α + βXz + z ,
Housing Units with Elderly Owners z
where each observation is a ZIP code. Because the ZIP code characteristics are
from the 2000 Census, I limit the HECM data to loans originated from 1995
to 2005 to ensure that the Census measures are reasonable proxies for the ZIP
code characteristics at the time of loan origination.18 For each ZIP code, the
dependent variable is the number of HECM originations from 1995 to 2005 as
a percentage of the number of housing units owned by people aged 65 or older.
Thus, the dependent variable approximates the reverse mortgage take up rate in
each ZIP code. I drop ZIP codes with fewer than 100 housing units owned by
elderly homeowners to reduce measurement error in the dependent variable.19
The regressors are a set of demographic and social-economic measures in-

cluding education, race/ethnicity, fraction of population in urban areas, the
logarithm of the median household income, the logarithm of the median house
17
According to industry specialists, most of the payment plan changes add a line-of-
credit option to existing term or tenure policies. Because HUD does not keep records
on the payment plan history of HECM loans, such assertions cannot be verified.
18
In robustness checks not shown, the results are similar if I restrict the data to loans
originated from 1997 to 2003.
19
In robustness checks not shown, the results are similar if I change the cutoff point to
50 housing units.
752 Shan
Table 2 Summary statistics of ZIP code characteristics.
Mean Median Std. Dev.

Number of HECM loans 9 2 19
originated from 1995 to 2005
Number of owner-occupied units 836 469 911
age 65+ in 2000
Number of 1995–2005 loans as a 0.78 0.48 1.04
percent of age 65+ units
Fraction of high school graduates 0.33 0.34 0.10
Fraction of college graduates 0.47 0.44 0.16
Fraction of blacks 0.09 0.01 0.16
Fraction of Hispanics 0.07 0.02 0.13
Fraction of population in urban 0.49 0.57 0.43
areas
Median household income ($) 42,562 38,431 16,429
Median house value ($) 116,082 91,500 86,637
Median owner costs as a 17.8 17.6 2.9
percentage of income (%)
Median credit score in 2005 723 732 77
Note: N = 19,847. Each observation is a ZIP code. Median credit score refers to the
median VantageScore among people in the ZIP code.
value and median owner costs as a percentage of household income.20 In ad-

dition to these census variables, I also control for the median credit score of
residents in each ZIP code in 2005.21 The final analysis sample has 19,847
observations (i.e., ZIP codes) and about 30% of these ZIP codes had no HECM
loan originations from 1995 to 2005. Table 2 shows the summary statistics
of this sample. In an average ZIP code, less than 1% of elderly homeowners
took out an HECM loan from 1995 to 2005. There is also a great deal of
heterogeneity in reverse mortgage take-up rate across ZIP codes.
Column (1) of Table 3 presents the ordinary least squares (OLS) estimation re-
sults. The numbers shown in brackets are the effect of a one-standard-deviation
increase in the explanatory variable on the dependent variable. The results sug-
gest that reverse mortgage take-up rates are higher in ZIP codes where residents
are better educated and in ZIP codes with higher fractions of blacks, Hispanics
and urban residents. Reverse mortgages are also more likely to originate in ZIP
codes with lower household income but higher house prices and in ZIP codes
20
According to the census definition, owner costs include payments for mortgages,
property taxes, hazard insurance on the property, utilities, fuels and condominium fees.
21
The credit score is the VantageScore provided by Equifax, one of three major credit
bureaus. A VantageScore of 700 is approximately equal to a FICO score of 660. I do not
have credit score data by ZIP code for years prior to 2005. Excluding the credit score
control does not affect the coefficients on other explanatory variables.
Table 3 ZIP-code-level characteristics and HECM loan originations from 1995 to

2005.
(1) (2) (3)

OLS Tobit Q50
Fraction of high school graduates 1.07∗∗ 2.18∗∗ 0.31∗∗
(8.3) (10.7) (3.3)
[0.11] [0.22] [0.03]
Fraction of college graduates 1.22∗∗ 2.09∗∗ 0.82∗∗
(12.2) (14.4) (12.3)
[0.20] [0.34] [0.13]
Fraction of blacks 0.52∗∗ 0.73∗∗ 0.07∗
(7.4) (8.5) (2.3)
[0.08] [0.12] [0.01]
Fraction of Hispanics 1.47∗∗ 1.84∗∗ 0.94∗∗
(18.0) (18.4) (22.4)
[0.19] [0.24] [0.12]
Fraction of population in urban areas 0.32∗∗ 0.69∗∗ 0.27∗∗
(18.8) (28.0) (22.0)
[0.14] [0.30] [0.12]
Log median household income −0.18∗∗ −0.34∗∗ −0.08∗∗
(4.4) (6.6) (3.1)
[−0.06] [−0.12] [−0.03]
Log median house value 0.22∗∗ 0.40∗∗ 0.18∗∗
(8.5) (11.0) (10.9)
[0.12] [0.22] [0.10]
Median owner costs as a percentage of income 0.12∗∗ 0.14∗∗ 0.07∗∗
(31.6) (29.8) (36.9)
[0.34] [0.41] [0.21]
Median credit score in 2005 (in 100s) −0.04∗∗ −0.05∗∗ −0.01
(4.5) (3.6) (1.9)
[−0.03] [−0.04] [−0.01]
Adjusted R2 0.328
N 19,847 19,847 19,847
Note: Each observation is a ZIP code. The dependent variable is the number of
1995–2005 HECM loans originated as a percent of housing units owned by house-
holders age 65 and over. Numbers in parentheses are t statistics. Standard errors in
the ordinary least squares (OLS) and Tobit regressions are robust to heteroskedasticity.
Numbers in brackets are the effects of a one standard deviation increase in the explana-
tory variable on the dependent variable. ∗ indicates significance at the 0.05 level. ∗∗
indicates significance at the 0.01 level.
where owner costs are high relative to income. These results are consistent with
the perception that cash-poor but house-rich elderly homeowners benefit from
reverse mortgages. Interestingly, ZIP codes where residents have lower credit
scores tend to have somewhat higher reverse mortgage take-up rates, perhaps
because the HECM program does not require credit worthiness and elderly
homeowners with higher credit scores prefer other types of loans to reverse
mortgages.
754 Shan
Table 4 Compare HECM borrowers to elderly homeowners in the SCF.
Single Male Single Median

Median Age (%) Female (%) Couple (%) HV (Real)
HECM SCF HECM SCF HECM SCF HECM SCF HECM SCF
1989 75 70 14 7 56 40 30 52 159,797 100,307
1992 75 71 14 13 56 42 30 45 159,575 103,428
1995 75 72 13 10 57 40 30 50 144,865 115,620
1998 75 72 13 10 57 39 30 51 137,352 120,819
2001 75 71 13 15 52 30 35 55 166,215 143,975
2004 74 73 16 14 48 36 36 49 208,507 164,611
2007 72 71 21 11 45 39 34 51 222,000 175,000
Note: Median house values are in 2007 dollars. HV = Housing Value; HECM = Home
Equity Conversion Mortgage; SCF = Survey of Consumer Finances.
All of these estimated coefficients are statistically significant, although eco-

nomically some are more important than others. For example, owner costs as a
percentage of income seem to have the largest correlation with reverse mortgage
originations in a ZIP code. A one-standard-deviation increase in this variable
is correlated with a 0.34-percentage-point increase in the reverse mortgage
take-up rate, raising it to 44% higher from the average of 0.78%. In contrast,
a one-standard-deviation increase in the median credit score is correlated only
with a 0.03-percentage-point decrease in the reverse mortgage take-up rate.
Because about 30% of the ZIP codes in the analysis sample have no reverse
mortgage originated from 1995 to 2005 and the dependent variable equals zero
for these observations, I also estimate a Tobit model and a median regression
model as robustness checks. Columns (2) and (3) display the estimation re-
sults of these two alternative models. All estimated coefficients have the same
sign as in the OLS model. The marginal effects of the explanatory variables
appear to be larger in the Tobit model and smaller in the median regression
model when compared to the OLS results. Overall, the above analyses show
that reverse mortgage take-up rates vary considerably across ZIP codes. A sig-
nificant portion of the variation can be explained by differences in ZIP code
characteristics.
Changing Characteristics of Reverse Mortgage Borrowers

In this section, I compare HECM loans originated in earlier years to those
originated in recent years to see if the composition of HECM borrowers changed
when the reverse mortgage market grew rapidly in the mid 2000s. Table 4
shows the characteristics of HECM borrowers at the time of loan origination
Figure 3 Distribution of HECM borrower age.

5
4
3
Percent of Loans Originated
2
1
5
4
3
2
1
62 65 70 75 80 85 90 95 100
Age at Origination
by year.22 It also tabulates the characteristics of elderly homeowners in the

general population, as indicated by the SCF data, in corresponding years. The
SCF is a cross-sectional survey of U.S. households that is conducted once every
3 years. Comparing HECM borrowers to elderly homeowners in the general
population informs us of whether potential changes in the characteristics of
HECM borrowers were driven by trends in the general population.
The first two columns of Table 4 show that HECM borrowers tend to be older
than other elderly homeowners, although the difference began to narrow in
recent years.23 Figure 3 displays the age distribution of HECM borrowers
22
Because there are only 11 loans originated in 1989, the 1989 HECM statistics shown
in Table 4 refer to loans originated in 1989 and 1990.
23
For HECM borrowers who are couples, the age of the younger spouse determines the
principal limit factor. To make the HECM and SCF statistics comparable, I use the age
of the younger spouse for couples in the SCF as well.
756 Shan
who took out loans before 2003 and those who took out loans from 2003 to
2007, respectively. Two features of this figure are worth mentioning. First, the
distribution of borrower age shifts to the leftover time, meaning that recent
borrowers are younger than early borrowers at the time of loan origination.
Second, there is a spike at age 62 in the distribution for recent borrowers but not
in the distribution for early borrowers. Such a spike suggests that homeowners
younger than age 62 may want to purchase reverse mortgages if allowed.
The middle six columns of Table 4 show the gender and marital status compo-
sitions of HECM borrowers and elderly homeowners in the SCF. Relative to
married couples, single elderly homeowners are more likely to purchase reverse
mortgages. Although the SCF data show no apparent trend in the gender and
marital status composition from 1989 to 2007, the HECM data suggest that the
fractions of single males and couples increased notably in recent years.24 Single
females, still the largest among the three groups, constitute a declining fraction
of all reverse mortgage borrowers. The last two columns of Table 4 compare the
median house value of HECM borrowers at the time of loan origination to that
of other elderly homeowners. In all years, the median HECM borrower tends
to have higher house values than the median elderly homeowner in the general
population. Real house values of elderly homeowners in the general population
increased steadily from 1989 to 2007. Interestingly, real house values of HECM
borrowers declined in the 1990s before rising in the 2000s.
It has been argued in the literature that reverse mortgages may not be attractive
to potential borrowers because the loan proceeds are limited to a fraction of
the house value (Sinai and Souleles 2008). In Figure 4, I plot the distribution
of the ratio of the IPL to house value for early borrowers and recent borrowers,
respectively. Recall that the IPL represents the present value of all loan proceeds
that may be received by the borrower plus the upfront costs and a servicing
fee set-aside. All else equal, a higher IPL-to-house-value ratio means that the
borrower can access a larger fraction of her or his housing equity at the time of
loan origination. The IPL is the product of MCA and the principal limit factor
which increases with age and decreases with expected interest rates. Figure 4
shows that, for early borrowers, the average IPL is about 60% of the house
value. For recent borrowers, the average IPL-to-house-value ratio increased to
about 65% despite the fact that recent borrowers tend to be younger than early
borrowers. The increase in the IPL-to-house-value ratio is driven by the lower
interest rates in recent years.
In Figure 5, I compare the reverse mortgage take-up rate before 2003 to that
from 2003 to 2007 by state. As before, the take-up rate is approximated by the
24
The fluctuations in the SCF statistics from year to year are likely due to SCF sampling
error.
Figure 4 Distribution of the initial principal limit as a percentage of house value.

7
6
5
4
Percent of Loans Originated
3
2
1
0
7
6
5
4
3
2
1
0
30 40 50 60 70 80 90
Initial Principal Limit as a Percent of House Value
ratio of HECM loans originated to housing units owned by householders aged

65 or older. During the first 13 years of the HECM program, very few elderly
homeowners took out reverse mortgages. For example, the District of Columbia,
Colorado, Utah and Rhode Island are the only states where over 1% of elderly
homeowners took out reverse mortgages. From 2003 to 2007, however, 30
states had a take up rate above 1%. The top five states with the highest take-up
rates in recent years are the District of Columbia, Nevada, California, Colorado
and Utah. In both time periods, elderly homeowners along the east coast and
the west coast are more likely to take out reverse mortgages than those living in
the middle of the country. However, the particularly large increases in take-up
rates in Nevada, California, Arizona and Florida in recent years suggest that
house price appreciation may have played a role in the growth of the reverse
mortgage market. I look into this link in more detail in the next section.
The results shown in this section suggest that the characteristics of HECM bor-
rowers changed notably when the reverse mortgage market grew substantially
758 Shan
Figure 5 Reverse mortgage originations as a percentage of owner-occupied units

with householders aged 65 or older by state, 1989–2002 and 2003–2007.
Early Years (1989-2002)
Recent Years (2003-07)
in the mid 2000s. For example, recent borrowers are younger and more likely
to be single males or couples. In addition, they can access a larger fraction of
their housing equity. Nevertheless, reverse mortgage borrowers still appear to
be a very select group of elderly homeowners who are older, are more likely to
be single, are relatively house-rich and tend to be from coastal states.
Figure 6 Fraction of indebted homeowners aged 62 or above, 1989–2007.
.6
.5
.4
.3
.2
1989 1992 1995 1998 2001 2004 2007
Credit Card Home Equity Any Debt
Note: Data are from the survey of consumer finances.
Explaining the Recent Expansion of the Reverse Mortgage Market

As shown before, the number of loans originated each year was relatively small
and stable during the first decade of the HECM program. Since the early 2000s,
the reverse mortgage market experienced substantial growth. In this section, I
explore factors that may have contributed to the growth of the reverse mortgage
market. In particular, I use panel data to study the degree to which house price
appreciation affects the demand for reverse mortgages.
One of the potential explanations for the significant growth of the reverse
mortgage market is that elderly homeowners have become more comfortable
with the idea of cashing out home equity and taking on debt in general. To
test the plausibility of this explanation, I use the 1989–2007 SCF data and
plot in Figure 6 the fractions of homeowners aged 62 or above who have
credit card debt, debt secured by their primary residences or any type of debt,
respectively. The fraction of elderly homeowners with credit card debt trended
up slightly over the 18-year sample period. The fraction of elderly homeowners
with mortgages, home equity loans or home equity lines of credit increased
from 22% in 1989 to 39% in 2007. The fraction of elderly homeowners with
760 Shan
Figure 7 10-Year Treasury bill interest rate and Case-Shiller nominal house price
index, 1989–2007.
10 200
190
9 180
170
8 160
150
7
140
130
6
120
5 110
100
4 90
80
3 70
1989 1992 1995 1998 2001 2004 2007
any type of debt rose from 44% to 57%. Further analysis of the SCF data shows
that the debt to income ratio among elderly homeowners almost tripled during
the sample period (from 0.28 to 0.81), while their median income only rose
by about 15% in real terms. Even though the time trends shown in Figure 6
are likely confounded by other time-varying factors, the pattern is consistent
with the hypothesis that an increasing proportion of elderly homeowners are
comfortable taking on debt secured by their homes. However, this potential
shift in financial attitude is much smaller than the strikingly large increase in
reverse mortgage originations.
Other factors may be important in explaining the growth of the HECM program.
Figure 7 shows the movements of two macroeconomic factors from 1989 to
2007—interest rates and house prices. The yield on the 10-year Treasury bill,
which is used to calculate the “expected interest rate” on HECM loans, declined
from between 7% and 8% in the early 1990s to between 4% and 5% in the mid-
2000s. Accordingly, the expected interest rate on HECM loans fell from around
9% to between 5% and 6%. Lower interest rates imply that borrowers can cash
out a larger fraction of their home equity at the time of loan origination. For
example, Table 1 shows that a 75-year-old homeowner with a $200,000 house

may pocket only $94,400 when the expected interest rate is 9%. However,
when the expected interest rate is 5%, the however, pocket $131,492 at the time
of loan origination. Therefore, declining interest rates may have made reverse
mortgages more attractive to elderly homeowners.25
In addition, house prices experienced unprecedented increases during this pe-

riod. Figure 7 shows that the Case-Shiller house price index almost doubled
between 2000 and 2006 in nominal terms. When house prices go up, elderly
homeowners can borrow more against their homes. For example, Table 1 shows
that a 75-year-old homeowner with a $200,000 home can pocket $107,715 at
the time of loan origination, assuming an expected interest rate of 7%. Suppose
house prices in the area rise by 25% and the homeowner’s home is now worth
$250,000. All else equal, he or she can now take home $137,165. The addi-
tional $30,000 loan proceeds for the same house would likely make reverse
mortgages more appealing.
Because yields on Treasury bills do not vary across locations, the effect of
lower interest rates on the demand for reverse mortgages cannot be separately
identified. However, the link between house prices and reverse mortgage orig-
inations can be examined using panel data because of the great heterogeneity
in house price movements across geographical areas. The remainder of this
section aims at identifying the effect of house price appreciation on the demand
for reverse mortgages.
To construct a regression analysis sample, I collapse the loan-level HECM data

into a panel data set where each observation is a county-year combination. The
dependent variable is the county-level growth rate of HECM loan originations in
percentage terms. For example, if a county had 100 loans originated last year and
150 this year, the growth rate would be 50%. To make sure that the dependent
variable is measured meaningfully, I drop counties with less than 30 reverse
mortgages originated in the previous year.26 The key independent variable, the
25
Even though the effect of interest rates cannot be separately identified from other
time-varying factors such as elderly homeowners becoming more familiar with reverse
mortgages and more comfortable with borrowing against their home, it is illustrative
to regress HECM origination growth rate on house price appreciation, the change
in real per capita income, the change in unemployment rate, the change in the 10-
year Treasury rate and a linear time trend. The estimated coefficient on the change
in interest rate is negative and statistically significant, which is consistent with the
hypothesis that declines in interest rate in recent years may have contributed to the
rising demand for reverse mortgages. The estimated coefficient on the linear time trend
is positive and statistically significant, which is consistent with the hypothesis that
elderly homeowners have become more familiar with reverse mortgages and more
comfortable with borrowing against their home.
762 Shan
county-level house price appreciation rate, is constructed using monthly county-

level single-family house price indexes purchased from CoreLogic. The house
price index data cover 999 counties in the United States from January 1976 to
December 2009.
Because a county has to have at least 30 HECM loans originated in the previous
year to be included, the final analysis sample is an unbalanced panel of 357
counties in 15 years. About three quarters of the 1,476 county-year observa-
tions are from the years 2003–2007. Panel A of Table 5 presents the summary
statistics of the analysis sample. The average annual real house price appreci-
ation rate is 4% in the sample, but the standard deviation is 9%, suggesting a
large degree of time-series and cross-sectional variation in house price move-
ments. The average county-level HECM loan origination growth rate is 39%,
reflecting the rapid growth in the reverse mortgage market in the mid-2000s.
I use the following panel regression model to estimate the effect of house price
appreciation on HECM loan originations:

Origc,t −Origc,t−1
Origc,t−1
= α + βhc,t + γ1 incc,t + γ2 unempc,t + ηc
+θt + c,t
where hc,t is the county-level house price appreciation rate from t − 1 to
t, and ηc and θ t are the county fixed effects and year fixed effects, re-
spectively. Note that because the dependent variable is the change in loan
originations, the county fixed effects allow for county-specific linear trends
in house price appreciation. To control for local economic conditions that
may affect both house price appreciation and reverse mortgage origina-
tion, I also include the county-level percentage change in real per-capita
income from t − 1 to t (incc,t ) and the county-level change in unem-
ployment rates from t − 1 to t (unempc,t ) in the regression model.27 If
higher house prices lead to a higher demand for reverse mortgages, we
expect β > 0.
Column (1) of Table 5 displays the baseline estimation results. The numbers
in brackets indicate the effect of a one standard deviation increase in the re-
gressor on the dependent variable. The estimated coefficient on the house price
26
About 85% of county-year combinations are lost because of this restriction. However,
the restricted sample covers about 80% of all HECM loans originated. In robustness
checks not shown, I also try changing the cutoff point to 20 or 50. The results are
essentially the same.
27
The county-level data on per-capita income are from the Bureau of Economic Analysis,
and the county-level unemployment rate data are from the Bureau of Labor Statistics.
Table 5 House price appreciation and the growth of reverse mortgage market.
Mean Median Std. Dev.

A. Summary Statistics of Regression Variables
County HECM loan origination growth rate (%) 38.6 26.7 55.9
County annual real house price appreciation rate (%) 3.9 2.9 8.6
County real per capita income (in 2007 dollars) 39,676 37,565 9,952
County unemployment rate (%) 5.0 4.7 1.5
County non-prime loan origination growth rate (%) 0.7 5.9 59.1
B. Regression results
(1) (2) (3) (4) (5)
House price appreciate in year t 3.39∗∗ 3.27∗∗ 3.46∗∗ 2.97∗ 2.67∗∗
(10.9) (7.9) (10.2) (2.3) (6.5)
[29.2] [28.1] [29.8] [25.5] [23.0]
House price appreciate in year t − 1 1.47∗∗
(3.1)
−0.16
House price appreciate in year t − 2 (0.3)
(0.52)
[−1.2]
Change in per capita income −1.14 −0.96 −1.17 −0.12 −0.70
(1.4) (0.8) (1.2) (0.1) (0.8)
[−2.6] [−2.2] [−2.6] [−0.3] [−1.6]
Change in unemployment rate 9.12∗∗ 8.87∗ 10.89∗ 4.84 6.62
(2.6) (2.1) (2.4) (0.6) (1.8)
[5.7] [5.7] [6.9] [3.2] [4.4]
Non-prime loan growth rate −0.12
(0.9)
[−7.1]
County FE Y Y Y Y Y
Year FE Y Y Y Y
MSA-Year FE Y
Adjusted R2 0.355 0.308 0.326 0.666 0.365
N 1,476 1,476 1,234 1,476 1,476
Note: Observations are at the county-year level. The dependent variable is the county
HECM loan origination growth rate. A county must have had at least 30 reverse mortgage
loans in the previous year to be included in the sample. A total of 357 unique counties are
included in the sample. Numbers in parentheses are t-statistics based on standard errors
clustered at the county level. Numbers in brackets are the effects of a one standard
deviation increase in the explanatory variable on the dependent variable. ∗ indicates
significance at the 0.05 level. ∗∗ indicates significance at the 0.01 level.
appreciation rate is positive and statistically significant, suggesting that places

with rising house values tend to have more growth in reverse mortgages. The
magnitude of the estimate suggests that a one-percentage-point increase in the
annual real house price appreciation rate is correlated with a 3.4-percentage-
point increase in the HECM loan origination growth rate. This magnitude is
equivalent to a 9% increase from the average HECM loan origination growth
764 Shan
rate of 38.6%. The estimated coefficient on the percentage change in real per
capita income is negative but statistically insignificant. The estimated coef-
ficient on the change in unemployment rate is positive and statistically sig-
nificant. The magnitude suggests that a one-percentage-point increase in the
county unemployment rate is correlated with a nine-percentage-point increase
in the HECM loan origination growth rate.28
The OLS estimate of the house price effect may be biased due to measurement
error or omitted variables.29 For example, if the county-level house price in-
dexes are measured with classical measurement error, my estimate of β would
suffer from attenuation bias. In the extreme case where the indexes contain
no meaningful information about house prices and are purely white noise, my
estimate of β would equal zero. To address this concern, I weight each ob-
servation by the number of housing units in the county. Counties with fewer
housing units have lower weights because repeat sales price indexes may be
measured less accurately in these counties. Column (2) shows the estimation
results when the weights are applied. The coefficient on the annual real house
price appreciation rate is almost identical to the estimate shown in column (1).
Even though I control for local economic conditions such as changes in per-
capita income and unemployment rate in addition to county fixed effects and
year fixed effects, one may be concerned that the estimate of β shown in
column (1) is biased because some omitted factors correlate with both house
price appreciation and the growth rate of reverse mortgage originations. For
example, predatory lending and substandard underwriting in some areas may
have fuelled the housing boom. At the same time, predatory lenders may have
targeted less sophisticated borrowers such as elderly homeowners and pushed
them to buy reverse mortgages. In this case, I may have overestimated the effect
of house price appreciation on reverse mortgage originations.
To address this concern, I estimate two alternative models in columns (3)

and (4). First, I include in the regression model an additional control—the
county-level growth rate of non-prime loan originations.30 This variable is
28
In results not shown, dropping the income and unemployment controls has no effect
on the key estimate.
29
Note that there is unlikely a simultaneity problem in estimating β given the small size
of the reverse mortgage market relative to the rest of the housing market.
30
Non-prime mortgages are any mortgages securitized in subprime or Alt-A pools.
Subprime loans are generally characterized as loans to borrowers with low credit scores
and/or little or no down payment, whereas Alt-A securities typically involve mortgages
with reduced or no documentation of the borrower’s income and assets and have a higher
proportion of interest-only mortgages and option adjustable rate mortgages.
constructed using the non-prime mortgage data from CoreLogic. The data
contain detailed information on private-label securitized non-prime mortgages
from 1999 to 2009. The number of non-prime loans originated in the United
States increased sharply in the early 2000s before evaporating in 2007 and 2008.
If some omitted factors affected housing prices through a loosening of lending
standards, this variable may allow me to control for such factors. Column (3) of
Table 5 displays the estimation results when the county-level non-prime loan
origination growth rate is included. The estimates are essentially unchanged
and the coefficient on the non-prime mortgage measure is small and statistically
insignificant.
Second, I control for MSA-year fixed effects in addition to county fixed effects
in the regression model. This specification identifies the effect of house price
appreciation by comparing counties within the same MSA in the same year.
To the extent that the omitted factors that correlate with both house price
appreciation and reverse mortgage originations are MSA-year specific, these
MSA-year fixed effects would control for them flexibly. A total of 927 MSA-
year fixed effects are included in the regression model, and column (4) of
Table 5 displays the estimation results. Because of the large number of fixed
effects, standard errors of the estimated coefficients increase substantially.
Nevertheless, the estimated coefficient on the house price appreciation rate
remains positive and statistically significant. Its magnitude is about one-eighth
smaller than the baseline estimate shown in column (1).
In column (5), I add lags of the real house price appreciation rate to the regres-
sion model. This specification allows elderly homeowners to respond to rising
house prices with a lag. The coefficient on the house price appreciation rate in
the current year becomes smaller, but it remains large, positive and statistically
significant. The coefficient on the previous year’s house price appreciation rate
is also positive and statistically significant, although the magnitude is about
one-half of the contemporaneous house price effect. The house price appreci-
ation rate from 2 years ago has no effect on current HECM loan origination.
The evidence shown above suggests that areas with rising house prices are more
likely to have a greater number of reverse mortgage originations. From 2003
to 2007, median house values in the United States increased by approximately
5% per year in real terms. During the same time period, the number of HECM
loan originations grew by 350%, or 50% per year. The estimates shown in
Table 5 imply that a house price appreciation rate of 5% raises HECM loan
originations by 17%. Therefore, these regression results suggest that house
price appreciation accounts for about one-third of the overall growth in the
reverse mortgage market between 2003 and 2007.
766 Shan
Conclusion
Combining 1989–2007 loan-level administrative HECM data with several other
data sources, I conduct three empirical analyses in this paper and show three
sets of findings. First, the take up rate of reverse mortgages varies significantly
across ZIP codes, and a great deal of this variation is correlated with the
characteristics of the ZIP codes. Second, reverse mortgage borrowers are a
very select group of elderly homeowners, and borrowers who took out loans
in recent years have significantly different characteristics than those who took
out loans in earlier years. This finding suggests that conclusions about loan
termination outcomes that are drawn by previous studies using earlier data may
not apply to loans originated in recent years. Finally, I show that house price
appreciation may have contributed to the rapid growth in the reverse mortgage
market in the mid-2000s.
Reverse mortgages, especially the tenure payment plan that gives borrower
equal monthly payments for as long as they are alive and continue living in
their homes, have often been regarded as similar to life annuities. Both re-
verse mortgages and annuities are complex financial products designed for the
elderly and both markets are much smaller than economic theory would pre-
dict (see Davidoff, Brown and Diamond 2005, Brown 2007). However, there
exists fundamental differences between the two. For immediate life annuities,
insurance against outliving one’s assets is provided by pooling mortality risks
across a group of people. In contrast, the tenure plan of HECM loans involves
little risk-pooling: if a borrower dies shortly after her HECM loan is origi-
nated, she pays back only the loan balance, which presumably is small. HUD
does not inherit this borrower’s entire housing equity to pay another borrower
who lives to be over 100 years old. Thus, the longevity insurance aspect of a
tenure HECM loan is very limited. Moreover, only 10% of HECM borrowers
choose the tenure payment plan or the modified tenure payment plan, which
suggests that the annuity aspect of reverse mortgages is irrelevant to most
borrowers.
From 2007 to 2009, housing prices plummeted in the United States after a
remarkable boom earlier in the decade. The sharp declines in house prices have
important implications to the reverse mortgage market. Lower house prices in
the incoming year imply higher cost for HUD to provide the insurance program,
especially on loans originated when house prices were near the peak. Also, this
article’s finding that house prices play an important role in the demand for
reverse mortgage suggests that reverse mortgages are less attractive to elderly
homeowners after prices fell significantly. In the face of these challenges,
whether the reverse mortgage market will continue to grow remains an open
question.
I thank Edward Szymanoski for providing the data. I also thank Neil Bhutta,
Daniel Cooper, Raven Molloy, Kevin Moore, Shane Sherlund, Edward Szy-
manoski, Bill Wheaton, and the two anonymous referees for helpful suggestions
and comments. This research was partially supported by the U.S. Social Secu-
rity Administration through grant No. 10-P-98363-1-05 to the National Bureau
of Economic Research as part of the SSA Retirement Research Consortium. The
findings and conclusions expressed in this article are solely my own and do not
represent the views of the Federal Reserve System or its staff, SSA, any agency
of the Federal Government, or the NBER.
References
Bayer, A.-H. and L. Harper. 2000. Fixing to Stay: A National Survey on Housing and
Home Modification Issues. AARP Research Report.
Bernheim, D.B., L. Forni, J. Gokhale and L.J. Kotlikoff. 2000. How Much Should
Americans Be Saving for Retirement? American Economic Review 90(2): 288–292.
Brown, J.R. 2007. Rational and Behavioral Perspectives on the Role of Annuities in
Retirement Planning. National Bureau of Economics Research Working Paper 13537.
Case, B. and A.B. Schnare. 1994. Preliminary Evaluation of the HECM Reverse Mort-
gage Program. Real Estate Economics 22(2): 301–346.
Davidoff, T. 2006. Maintenance and the Home Equity of the Elderly. Working Paper.
University of British Columbia, Vancouver, Canada.
—— 2010. Home Equity Commitment and Long-Term Care Insurance Demand. Journal
of Public Economics 94(1): 44–49.
Davidoff, T., J.R. Brown and P.A. Diamond. 2005. Annuities and Individual Welfare.
American Economic Review 95(5): 1573–1590.
Davidoff, T. and G. Welke. 2007. Selection and Moral Hazard in the Reverse Mortgage
Market. Working Paper. University of British Columbia, Vancouver, Canada.
Kutty, N.K. 1998. The Scope for Poverty Alleviation among Elderly Home-owners in
the United States through Reverse Mortgages. Urban Studies 35(1): 113–129.
Mayer, C.J. and K.V. Simons. 1994. Reverse Mortgages and the Liquidity of Housing
Wealth. Real Estate Economics 22(2): 235–255.
Merrill, S.R., M. Finkel and N.K. Kutty. 1994. Potential Beneficiaries from Reverse
Mortgage Products for Elderly Homeowners: An Analysis of American Housing Survey
Data. Real Estate Economics 22(2): 257–299.
Michelangeli, V. 2008. Does it Pay to Get a Reverse Mortgage? Working Paper. Con-
gressional Budget Office, Washington, DC.
Mitchell, O.S. and J.F. Moore. 1998. Can Americans Afford to Retire? New Evidence
on Retirement Saving Adequacy. Journal of Risk and Insurance 65(3): 371–400.
Munnell, A.H., M. Soto and J.-P. Aubry. 2007. Do People Plan to Tap Their Home
Equity in Retirement? Center for Retirement Research Brief Number 7-7.
Rasmussen, D.W., I.F. Megbolugbe and B.A. Morgan. 1995. Using the 1990 Public
Use Microdata Sample to Estimate Potential Demand for Reverse Mortgage Products.
Journal of Housing Research 6(1): 1–23.
Rodda, D.T., C. Herbert and H.-K. Lam. 2000. Evaluation Report of FHA’s Home Equity
Conversion Mortgage Insurance Demonstration. Abt Associates Inc., Cambridge, MA.
768 Shan
Sinai, T. and N.S. Souleles. 2008. Net Worth and Housing Equity in Retirement. J.
Ameriks and O.S. Mitchell, editors. Recalibrating Retirement Spending and Saving.
Oxford University Press: Oxford, U.K.
Skinner, J. 2007. Are You Sure You’re Saving Enough for Retirement? Journal of
Economic Perspectives 21(3): 59–80.
Szymanoski, E.J. 1994. Risk and the Home Equity Conversion Mortgage. Real Estate
Economics 22(2): 347–366.
Szymanoski, E.J., J.C. Enriquez and T.R. DiVenti. 2007. Home Equity Conversion
Mortgage Terminations: Information to Enhance the Developing Secondary Market.
Cityscape: A Journal of Policy Development and Research 9(1): 5–45.
Venti, S.F. and D.A. Wise. 1989. Aging, Moving and Housing Wealth. D.A. Wise, editor.
Economics of Aging. University of Chicago Press: Chicago.
—— ——. 1990. But They Don’t Want to Reduce Housing Equity. D.A. Wise, editor.
Issues in the Economics of Aging. University of Chicago Press: Chicago.
—— ——. 1991. Aging and the Income Value of Housing Wealth. Journal of Public
Economics 44(3): 371–397.
—— ——. 2004. Aging and Housing Equity: Another Look. D.A. Wise, editor. Per-
spectives on the Economics of Aging. University of Chicago Press: Chicago.

Shan

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Shan

Uploaded by

Copyright:

Available Formats

2011 V39 4: pp.

Reversing the Trend: The Recent

In this article, I examine loan-level data on reverse mortgages originated from

I then compare reverse mortgages originated in earlier years (1989–2002) to

Overview of Existing Studies

housing equity should be included as consumable wealth. Because housing

The degree to which we should consider housing equity as retirement savings

Background on the Home Equity Conversion Mortgage (HECM) Program

Table 1 HECM proceeds to a hypothetical borrower.

Expected Rate Age = 65 Age = 75 Age = 85

Figure 2 Growth in HECM loans, 1989–2007.

observe income or financial wealth of these borrowers, nor do I observe their

ZIP Code Characteristics and HECM Loan Origination

The regressors are a set of demographic and social-economic measures in-

Table 2 Summary statistics of ZIP code characteristics.

Mean Median Std. Dev.

value and median owner costs as a percentage of household income.20 In ad-

Table 3 ZIP-code-level characteristics and HECM loan originations from 1995 to

(1) (2) (3)

Table 4 Compare HECM borrowers to elderly homeowners in the SCF.

Single Male Single Median

All of these estimated coefficients are statistically significant, although eco-

Changing Characteristics of Reverse Mortgage Borrowers

Figure 3 Distribution of HECM borrower age.

by year.22 It also tabulates the characteristics of elderly homeowners in the

Figure 4 Distribution of the initial principal limit as a percentage of house value.

ratio of HECM loans originated to housing units owned by householders aged

Figure 5 Reverse mortgage originations as a percentage of owner-occupied units

Early Years (1989-2002)

Recent Years (2003-07)

Figure 6 Fraction of indebted homeowners aged 62 or above, 1989–2007.

Credit Card Home Equity Any Debt

Note: Data are from the survey of consumer finances.

Explaining the Recent Expansion of the Reverse Mortgage Market

example, Table 1 shows that a 75-year-old homeowner with a $200,000 house

In addition, house prices experienced unprecedented increases during this pe-

To construct a regression analysis sample, I collapse the loan-level HECM data

county-level house price appreciation rate, is constructed using monthly county-

Mean Median Std. Dev.

appreciation rate is positive and statistically significant, suggesting that places

To address this concern, I estimate two alternative models in columns (3)

You might also like