marriage and divorce

Published on June 2016 | Categories: Types, School Work | Downloads: 38 | Comments: 0 | Views: 437
of 19
Download PDF   Embed   Report

researchmarriage and divorce

Comments

Content

The Impact of Homeownership on Marriage and
Divorce: Evidence from Propensity Score
Matching
Michal Grinstein-Weiss, Kim R. Manturuk, Shenyang Guo, Pajarita Charles, and Clinton Key
This research examined the relationship between homeownership and the likelihood of marriage or divorce. Drawing on exchange theory and an economic understanding of marriage,
the authors hypothesized that single homeowners are less likely to marry than single renters,
whereas married homeowners are less likely to divorce than married renters. These hypotheses were tested using longitudinal data collected from a group of lower income homeowners
and a comparison group of renters. Propensity score models were used to account for selection bias. Results indicate that single homeowners are, in fact, less likely to marry than their
renting counterparts, whereas married homeowners are less likely to divorce than married
renters. These findings suggest that assets, such as a home, can play a significant role in the
likelihood of both marriage and divorce.
KEY WORDS:

divorce; homeownership; marriage; propensity score analysis

T

his article reports on a study investigating
the relationship between homeownership
and decisions to marry or divorce. Traditionally, choices about homeownership and marriage have been tightly linked; however, changing
demographic patterns suggest a decoupling of these
major life decisions. For example, trends show a
notable increase in the number of nonmarried
homeowners (Allen, 2009; Sykes, 2005). Likewise,
trends point to women marrying later (Qian &
Preston, 1993; Stevenson & Wolfers, 2007), increased prevalence of cohabitation outside of marriage (Kennedy & Bumpass, 2008; Thornton,
1988), and rising divorce rates (Cherlin, 2005).
These demographic shifts underscore the importance of understanding the relationship between
the two institutions of marriage and homeownership, both of which have important economic
and cultural impacts. As such, this study focused
on two research questions: Are single homeowners
less likely than single renters to marry? What role
does homeownership play in strengthening marriage and reducing the risk of divorce?
The relationship between homeownership and
marriage is important yet difficult to examine because characteristics that improve odds of marriage
(or decrease the risk of divorce) are also characteristics that are closely associated with homeownership.

doi: 10.1093/swr/svu016

© 2014 National Association of Social Workers

For example, people with higher incomes are less
likely to divorce and are more likely to be homeowners than people with lower incomes. Past research has
been unable to establish whether homeownership
has a direct effect on marriage and divorce or whether
the two are simply correlated with homeownership.
The present study applied propensity score analysis,
an innovative and rigorous methodology, to isolate
homeownership’s relationships with marriage and
divorce.
The methodological contribution is only one of
several contributions made by this study. Although
previous research examined the association between tenure (that is, owning versus renting) and
divorce, little, if any, research has considered the
owning-versus-renting relationship among a sample
of low- to moderate-income participants. Exploring
this relationship among a low socioeconomic status
(SES) population is warranted because evidence has
suggested that homeownership confers social benefits, such as increased community involvement and
greater psychological well-being to householders, if
“done right” (Grinstein-Weiss, Key, Guo, Yeo, &
Holub, 2013). If the benefits of homeownership
were found to include a reduction in the risk of
divorce among economically disadvantaged households, such a finding could have important implications for policy and practice.

73

In addition, the current study has implications for
social welfare policy and research. Although homeownership used to be predominantly the domain of
married couples, policymakers have promoted it as
an asset-building strategy to help lower income
individuals, including unmarried individuals, achieve
financial stability. These policy changes brought
about a meaningful rise in homeownership among
single adults; however, a second area of focus for federal policymakers has been the promotion of marriage
and the stabilization of couple relationships among
economically disadvantaged individuals (Andersson,
2002; Bumpass & Lu, 2000). The premise behind
these strategies is that stable, two-parent partnerships
are beneficial for both child and adult well-being.
This premise is supported by research that examined
household composition effects on child well-being
and showed that, as compared with children raised
by their married biological parents, children raised
by unmarried parents had significantly higher risks
of living in poverty and of experiencing a range of
poor outcomes (that is, mental, physical, social, and
educational) (Amato, 2005; McLanahan, 2011;
McLanahan & Sandefur, 1994). Therefore, understanding the factors that influence the decision to
marry (or divorce) has important implications for
policymakers and researchers seeking to improve
the welfare of families. One such factor is tenure.
THEORETICAL FRAMEWORK

Several theories have attempted to explain why
people elect to marry and how they choose a partner. One such effort was Becker’s (1974) economic
theory of marriage, which argued that the decision
to marry is fundamentally an economic decision.
Becker outlined two principles of marriage. First,
people marry to gain resources and to establish a
division of household labor. He observed: “Persons
marrying (or their parents) can be assumed to
expect to raise their utility level above what it
would be were they to remain single” ( p. 300). Traditionally, men sought out marriage as a way to
form a partnership with someone who would manage the household and raise children, whereas
women traded their domestic labor for financial
security.
Becker’s second principle holds that marriage
will occur only when the benefits outweigh the
costs. Therefore, it follows that single people with
more resources will be less likely to marry because
they will marry only when the union increases their

74

already-high utility. Following Becker’s logic, single homeowners are likely seen as desirable partners
because they possess significant resources. However, it is more challenging for these singles to
find a suitable marriage partner—that is, someone
who will improve their financial situation—because
they already have financial security. Becker’s argument also held that if a marriage occurred, both
partners would strive to improve their utility; this
principle is closely related to the second theory
that informed this research—the social exchange
theory of marriage.
Proposed by Edwards (1969), the social exchange theory of marriage is similar to Becker’s in
many respects. Edwards perceived marriage as an
exchange relationship in which the commodities
of exchange included economic and noneconomic
goods. In Becker’s view, as in Edwards’s, people
enter into relationships that are mutually beneficial
or that have the potential for reciprocity. Likewise,
Edwards’s second principle is that each party within
an exchange relationship seeks to maximize gains
and minimize costs.
The third principle of the social exchange theory, and the point at which it departs from Becker’s
economic theory, is that people will terminate an
exclusive exchange relationship (that is, marriage)
when the benefits of terminating the relationship
outweigh the costs of ending the relationship.
Exchange theorists go beyond simply understanding marriage decisions by seeking to explain why
and when people terminate marriages. These theorists have posited that relationships are more stable
when the costs of terminating the relationship are
high, when the rewards for sustaining the relationship are desirable, and when the available alternatives are less attractive (Floyd & Wasner, 1994).
Although we chose to view marriage and divorce
through a rational choice lens (that is, to presume
that people are driven by rational behavior), we recognize that other models of marriage and divorce
exist. For example, the institutional view of marriage emphasizes that couples marry because of societal norms, values, and pressures (Goode, 1982).
However, research examining factors associated
with marriage has suggested that patterns of economic independence, particularly independence
achieved through educational gains, play an important role in marriage decisions (Goldstein &
Kenney, 2001). Given that individuals who have
attained higher levels of education are expected to

Social Work Research Volume 38, Number 2

June 2014

marry at higher rates than those with less education
(Goldstein & Kenney, 2001; Payne & Gibbs, 2011),
we argue that it is important to examine the effect
of other proxies of SES, such as homeownership,
on marriage. Thus, the economic and exchange
theories together help to explain the relationship
between homeownership and marriage. Using
this framework, we examined the ways in which
homeownership can play an influential role, affecting a person’s likelihood of both marriage and
divorce.
Homeownership and Marriage

According to the economic theories of marriage,
the marriage market is filled with people seeking
to maximize their resources and minimize their
costs. We posit that homeownership functions as a
resource on the marriage market through three
mechanisms: (1) wealth as a direct result of homeownership, (2) other financial resources correlated
with homeownership, and (3) social resources associated with homeownership. First, studies have
demonstrated that homeowners, even lower income
homeowners, accumulate greater wealth than renters
with comparable income. Examining wealth accumulation during an eight-year period (1984 to
1992), Boehm and Schlottmann (2008) found
that nonhousing wealth was almost nonexistent
for lower income households and that minority
families at all income levels had, on average, insignificant or even negative growth in nonhousing
wealth. In contrast, these researchers found that
even the lowest income homeowners realized significant equity gains and accumulated housing
wealth over the same time.
Second, homeowners have resources that are
correlated with, rather than caused by, homeownership. Purchasing a home implies that a person has
a stable income, some degree of financial literacy,
and other financial assets such as savings accounts
or investments. Homeownership generates wealth,
and homeowners have more wealth and assets
before purchasing a home than do renters who continue renting (Di & Liu 2007).
Third, homeowners are more likely than renters
to have social resources that are valuable or advantageous. A recent study by Manturuk, Lindblad,
and Quercia (2010) reported that, as compared
with renters, homeowners have more social ties
to others in their neighborhoods. Given these
social ties, homeowners have a greater network of

resources, including contacts who can help with
tasks such as finding a job or fixing a computer.
Other research has found that homeowners have
more social capital and are more connected with
the public infrastructure of their communities
(Glaeser, Laibson, & Sacerdote, 2002; Glaeser &
Sacerdote, 2000). According to economic theories
of marriage, a key function of marriage is to provide a division of labor, which increases personal
efficiency; however, the evidence that homeowners have higher levels of social capital suggests
that single homeowners may be able to achieve
this type of division of labor through their social
ties.
In summary, homeownership functions as a financial and social resource, meaning that homeowners
“bring more to the table” when forming a marriage
partnership and have more to lose. Although homeowners may be desirable marriage partners, they
might find it difficult to find someone who brings
enough resources to the partnership to outweigh
the costs. As such, the study’s first hypothesis was
that single homeowners would be less likely to marry
than single renters.
Homeownership and Divorce

As discussed earlier, exchange theory asserts that the
way in which people value the costs and rewards of
a relationship affects the stability of the relationship.
Thibaut and Kelley (1959) proposed that people
evaluate relationships using two primary criteria.
The first is the comparison level (CL), which is
the attractiveness of the relationship, or how satisfying the person finds the relationship. The second
criterion is the comparison level for alternatives
(CLalt), which is the perception of barriers to alternative relationships and potential rewards those
relationships might bring (Karney & Bradbury,
1995). Relationships with many rewards and few
costs tend to be highly satisfactory and are more
likely to be stable.
For married homeowners who find their relationship unsatisfying, one primary cost of divorce
is often the loss of a major asset: the house. The
house represents the single largest asset and source
of wealth for many homeowners, especially lowincome homeowners (Boehm & Schlottmann,
2008). Generally, ending a marriage means parting
with at least a portion of that wealth. A recent study
found that couples with no assets were 70% more
likely to divorce than couples with at least $10,000

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

75

in shared assets (Dew, 2007). Prior research has also
found that homeownership is an asset that decreases
a couple’s likelihood of marital dissolution (Bracher,
Santow, Morgan, & Trussell, 1993; South & Lloyd,
1995), although these studies failed to account for
selection bias. Drawing on these theories and past
research, the current study’s second hypothesis
was that the likelihood of divorce would be lower
among married homeowners than among married
renters.
METHOD

Data

This study used longitudinal survey data from the
Community Advantage Program Study (CAPS),
which had followed a sample of homeowners
whose mortgages were funded through a secondarymarket mortgage loan program called the Community Advantage Program (CAP). CAP funded
prime-rate mortgages for low- and moderateincome homebuyers who otherwise would not
have qualified for a mortgage. To qualify for the
program, the borrower must have had income
that was less than 80% of the area median income
or 115% of the area median income in a predominantly minority neighborhood. The initial CAPS
survey sampled 3,743 of these loan recipients.
To evaluate the impact of homeownership, a
sample of comparable renters was added to the
panel. Renters were selected by randomly calling
potential respondents who lived within a threemile radius of a participating CAPS homeowner.
To qualify for the study, the renters had to meet
the same income criteria as the CAPS homeowners:
total income under 80% of area median income or
115% in a predominantly minority census tract. At
baseline, the renters group comprised 1,530 respondents. Riley and Ru (2009) compared the CAPS
renter panel with renters in the Current Population
Survey (CPS) who met the same income criteria as
those used for CAP. Overall, Riley and Ru found
that the CAPS renters were similar to a nationally
representative sample of low-income renters, but
with notable differences: On average, CAPS renters
are older and more likely to be Hispanic or black
than is a typical low-income renter. Compared
with the CAPS homeowners, the renters were
older, less likely to be married, and less likely to
have children. The renters also had lower levels of
education.

76

This analysis draws information from five waves
of CAPS data collected from 2004 to 2008. Over
the successive waves of the survey, attrition has
occurred in the panels. At wave 5, the survey was
completed by 3,357 of the original 5,273 respondents. Data for these respondents were examined
for missing values on important study covariates.
Multiple imputations were performed on subsamples of respondents who were unmarried at baseline
and married at baseline. The unmarried analytic
sample includes 1,580 respondents. To test the
relationship between homeownership and divorce,
1,115 married homeowners and renters were
included.
Measures

Dependent Variables. The key outcome measures in
this study are time to marriage and time to divorce.
For the models predicting marriage, all respondents
were single at the beginning of the study, and we
measured years to outcome in discrete units (1 to
4). In the analyses on marriage, we included respondents who were unmarried, divorced, widowed, or
cohabiting as single. To make full use of the available data, we included cohabiting couples in the
models predicting marriage, because these individuals are eligible for marriage. For the models
predicting divorce, we included only married
respondents.
Independent Variables. The key independent
variable in this study was homeownership. In
CAPS, homeownership is a dichotomous variable,
measured with owners (coded 1) and renters (coded
0). This study treated homeownership as time varying, and the variable was coded corresponding to
the time of marriage or divorce. For instance, if a
respondent married in study year 3, then we
checked the respondent’s homeownership status
at year 2.
Control Variables. The models also controlled
for important sociodemographic measures. For
the propensity score matching, we used respondent
characteristics measured at year 1 of CAPS participation. We controlled for race and ethnicity
because research has shown significant racial differences in both marriage rates and likelihood of
homeownership (Boehm & Schlottmann, 2004).
In addition, existing evidence identifies significant differences in homeownership by race, ethnicity, and gender. Single men are more likely than
women to be homeowners, as are white singles

Social Work Research Volume 38, Number 2

June 2014

and couples. It is interesting that the financial benefits associated with homeownership are fairly equitable across racial, ethnic, and gender groups once
property is purchased (Sedo & Kossoudji, 2004).
Therefore, we controlled for race, ethnicity, and
gender in our models. We measured gender using
a 1/0 indicator variable: female was coded as 1,
and male coded as 0. Race/ethnicity was measured
using indicator variables for white, black, Hispanic,
and other race/ethnicity.
Researchers have found interesting interactions
among age, homeownership, and marriage. A recent
study found that, among 18- to 24-year-olds,
married couples had the highest rate of transitioning
to homeownership. However, among 30- to 34year-olds, single people were more likely than couples to become first-time homeowners (Feijten,
Mulder, & Baizán, 2003). In all of our analytic
models, we used a continuous variable to control
for age.
A correlation between level of education and
homeownership has been established by previous
research. On average, levels of education are slightly
higher among low-income homeowners than
among low-income renters (Rohe & Stegman,
1994). In our analysis, we used a series of indicator
variables for education. The education categories
ranged from less than a high school degree to
advanced or graduate degree.
We included two financial predictors of homeownership: income and employment status. Given
that employment and income are documented during the mortgage application process, people who
are unemployed or have very low incomes are less
likely to be approved to buy a home. We measured
income in units of $1,000 and included an indicator
variable for whether a respondent was employed.
Finally, we included an indicator variable for
whether a respondent had children living in the
home. Homeowners are more likely than renters
to have children at home because buying a home
is often considered a precursor to having children
(Townsend, 2002). In addition, couples with children are less likely to divorce than couples without
children (Cleek & Pearson, 1985).
The analysis also controlled for the characteristics
of the respondent’s census tract at baseline, including median tract house value, median tract rent, and
tract disadvantage score. The tract neighborhood
disadvantage score was constructed from the following four tract-level items in the 2000 census:

percentage unemployed, percentage in poverty,
percentage on public assistance, and percentage of
single-headed households with children (Caughy,
Hayslett-McCall, & O’Campo, 2007). We included
tract-level characteristics to distinguish the effect of
homeownership on marriage or divorce from the
effect of neighborhood context. Although most of
the variables discussed earlier were fixed characteristics, the number of children and employment status were likely to change over time. Thus, for the
logistic models used to create the propensity scores,
we measured all characteristics at baseline; for the
subsequent survival analysis, we used time-varying
measures of number of children and employment.
Missing Data Imputation

The data were imputed using multiple imputation
through chained equations (Royston, 2005; Rubin,
1987; van Buuren, Boshuizen, & Knook, 1999). To
implement this procedure, we used the /ice/ package for Stata. As compared with the more common
practice of listwise deletion, imputation helps to
preserve sample size and reduce bias caused by nonresponse to survey questions (Raghunathan, 2004).
We chose to use multiple imputation rather than
single imputation because multiple imputation
allows researchers to include variability from the
imputation in the analyses. The imputation model
included all variables in the analytic model. Each
variable with missing information was modeled
using all of the other variables in the data set. Missing values were selected from the resultant conditional density and applied in a new iteration of
the data set. Because of the uncertainty inherent
in the selection of a value, this process was repeated
iteratively. About a third of cases had item-missing
data on at least one measure in the analytic model.
About 3% of information was missing before
imputation. The data were found to be missing at
random. Given the fraction of missing data, 15
imputed data sets were created.
Analytic Methods

This research aimed to evaluate causality and
answer this question: Does homeownership cause
an increase or decrease in the likelihood of marriage
or divorce? Over the past 30 years, statisticians (for
example, Rosenbaum & Rubin, 1983) and econometricians (for example, Heckman, 1978, 1979)
have made substantial advances in developing and
refining new approaches for the estimation of causal

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

77

effects from observational data. Collectively these
approaches are known as propensity score analysis
(Guo & Fraser, 2010). We elected to use propensity
score analysis for this study because the approach
enabled us to address the selection bias that is inherent in observational studies. Study respondents
self-select whether to be homeowners or renters;
therefore, it is difficult to use traditional regression
analysis to isolate the independent effect that homeownership might have on marital status. However,
in propensity score models, selection is explicitly
modeled and selection effects are considered in
the calculation of causal impacts (Heckman, 1978,
1979).
Traditional approaches to regression analysis
attempt to control for variables that are correlated
with both the independent and dependent variables. For example, in the current study, we anticipated that the level of education could be
correlated with both homeownership and marital
status. However, simply controlling for level of
education by adding the variable to a regression
model is inadequate because respondents’ levels of
education likely differ by their homeownership status; that is, level of education is not evenly distributed within each group. Propensity score analysis
enables us to identify homeowners and renters
who appear highly similar when compared using
a set of variables that have been shown in earlier
research to predict homeownership status. If this
propensity score analysis identifies significant effects
associated with homeownership, we then will have
confidence that such effects represent the independent effects of homeownership rather than other
correlated variables.
To draw valid causal inference, this study applied
the Neyman-Rubin counterfactual framework
(Morgan & Winship, 2007; Neyman, 1923/1990;
Rubin, 1974, 1990, 2006) as a conceptual model
to guide the data analysis. Under this framework,
a counterfactual is a potential outcome or state of
affairs that would have happened had a person
been of the opposite status on the explanatory variable (Shadish, Cook, & Campbell, 2002). For
example, in the context of the current study, we
needed to consider what a homeowner’s likelihood
of divorce would have been if the person had not
been a homeowner, but we also recognize that
this alternative outcome was not the same as a renter’s likelihood of divorce. The Neyman-Rubin
framework offers a practical way to evaluate the

78

counterfactuals. In working with data from a sample that represents the population of interest, the
standard estimator for the average treatment effect
is seen as the difference between two estimated
median times from the sample data. In other words,
^1 jw ¼ 1Þ  Median ðT
^0 jw ¼ 0Þ;
t^ ¼ Median ðT
^1 is the event time under the treated conwhere T
^0 is the event time under the control condition, T
dition, and w is a binary variable indicating
treatment receipt (that is, w = 1, treatment; and w =
0, comparison). The counterfactual framework
underscores the importance of balancing data, carefully seeking the potential outcome (equivalently
choosing the most comparable comparison group)
for a valid causal inference, and estimating appropriate treatment effects by using methods that are
appropriate and suitable for the research questions.
Recognizing that each method has strengths and
limitations, Sosin (2002) underscored the importance of using multiple methods to account for
selection bias. Using different methods (that is, sample selection, conventional control variables, instrumental variables, and propensity score matching)
to examine a common data set, Sosin found that
the methods provided widely divergent estimates.
Given his finding, Sosin suggested that researchers
should regularly compare estimates across multiple
methods.
Following Sosin’s recommendation, we ran models using the following five approaches: (1) a discretetime survival analysis applied to the original sample
without matching; (2) a propensity score greedy
matching (that is, the nearest neighbor within caliper matching) followed by a discrete-time survival
analysis; (3) a propensity score optimal-pair matching
that uses the generalized boosted regression to
estimate the propensity score and a follow-up
discrete-time survival analysis; (4) a propensity score
optimal full matching that uses generalized boosted
regression to estimate the propensity score and a
follow-up Hodges-Lehmann aligned rank test;
and (5) 15 files that multiply imputed missing data
of independent and matching variables (on which
all analyses were conducted). For each model,
results from each of the imputed data sets were
aggregated using either the Rubin rule or procedures developed for multivariate models. To warrant the robustness of the final findings about the
impact of homeownership on timing of transition

Social Work Research Volume 38, Number 2

June 2014

into marriage and to ensure that those findings hold
in a differently defined population, we conducted a
sensitivity analysis using a sample consisting of only
those respondents who indicated a baseline marital
status of single-never married.
RESULTS

Impact of Homeownership on Timing of
Transition to Marriage

The original sample after multiple imputation of
missing data contained 1,580 participants; of these,
926 (58.6%) were homeowners who either owned
a house at the baseline or purchased a house during
the four-year study period and 654 (41.4%) were
renters. The sample’s descriptive statistics and results
from imbalance checks before and after matching
are presented in Table 1. A majority of respondents
were female, white, and employed. Both the homeowner and renter groups commanded modest
incomes, with homeowners reporting significantly
higher earnings than renters. The average study
respondent was approximately 40 years old. As
shown, before the matching procedure, the overall
sample was not balanced on several covariates. In
the table, levels of statistical significance for categorical covariates come from a chi-square test, and
those for continuous covariates come from an independent sample t test.
The table also presents sample sizes after the
greedy matching and optimal matching procedures.
After greedy matching, the matched sample contained 347 homeowners and 347 renters. After
optimal pair matching, the matched sample contained 654 homeowners and 654 renters. Optimal
full matching retained all 926 homeowners and
654 renters of the original sample but grouped
them in matched strata. The ratio of homeowners
to renters varied by stratum; however, within each
stratum, the owners and renters had similar propensity scores.
Also indicated in Table 1 is that both greedy
matching and optimal matching improved sample
balances. After greedy matching, none of the 10
covariates showed a significant difference. The
absolute standardized difference in covariate means
before optimal matching (that is, dX) usually has a
higher value than the index after optimal matching
(that is, dXm). As shown in Table 1, most covariates
decreased in values from dX to dXm. This suggests
that optimal matching indeed improved balances.

Another important consideration is whether the
propensity scores for treated participants overlap
sufficiently those for the controls. The consequence
of the overlap varies by the model used. A lack of
overlap between the propensity score group suggests that the sample of comparable cases was small.
Propensity scores estimated by binary logistic
regression (greedy matching) showed that the two
groups did not share a sizable common support
region. Therefore, the number of treated participants was reduced from 926 in the original sample
to 347 in the matched sample. This significant loss
of sample size in greedy matching is a limitation. In
the optimal matching approach, the two groups differed from each other on the distribution of estimated propensity scores. However, this difference
was not a problem for optimal matching. Pair
matching created 654 matched pairs that contained
all controls, even though the pair matching lost 272
treated participants (29.4%); full matching created a
matched sample containing all 926 treated and 654
comparison participants.
The four methods used to estimate difference on
timing of transition into marriage between homeowners and renters all showed consistent findings,
regardless of whether the models controlled for
selection bias (see Table 2). Analysis of data for
this study population showed that homeowners
transitioned into marriage at a slower speed than
renters. The estimated odds ratio of getting married
is 0.503 ( p < .001) for the original unmatched sample, indicating that the odds of getting married were
49.7% lower for the homeowners than for the renters. For the matched sample created by greedy
matching, the odds ratio of getting married was
0.447 ( p < .001). For the matched sample created
by the optimal pair matching, the odds ratio of transition to marriage was 0.494 ( p < .001); both findings indicate that the homeowners transitioned into
marriage at a slower speed than the renters. Using
the optimal full matched sample to compare length
of time to transition to marriage, we found that the
transition to marriage took 0.360 of a year (approximately 4.32 months) longer for homeowners than
for renters. The Hodges-Lehmann aligned rank test
showed that this difference was statistically significant ( p < .001).
The sensitivity analysis used a sample of 707
respondents who were single and never married at
the baseline. Of these participants, 402 owned
homes and 305 never owned homes during the

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

79

Table 1: The Study of Transition to Marriage: Sample Description and Imbalance Check
before and after Matching
% of Homeowners on Mean (SD)
of the Covariate by Groupa

Covariate

Number of
homeowners
Number of
renters
Number (%) of
homeowners
lost after
matching
Gender (%)
Male
Female
Race (%)
White
Africa
American
Hispanic
Other
Age at baseline M
(SD)
Homeowners
Renters
Education at
baseline (%)
11th grade or
less
High school
diploma
Some two-year
college, no
degree
Associate’s
degree
Some four-year
college, no
degree
Bachelor’s
degree
Some graduate
school, no
degree
Graduate/
professional
degree
Number of
children at
baseline M
(SD)
Homeowners
Renters

Overall Sample
before
Matching
(N = 1,580)

Sample after
Greedy
Matching
(N = 694)

Absolute
Standardized
Difference in
Covariate Means
before Matching
(dx)

Absolute Standardized
Difference in Covariate Means
after Matching (dxm)
Sample after
Optimal Full
Matching
(N = 1,580)

Sample after
Optimal Pair
Matching
(N = 1,308)

926

347

654

926

654

347

654

654

.638

272 (29.4)
.618

0 (0)
.105

.176b
.015b
.031b

.123b
.022b
.018b

.146b
.059b
.060b

.193

.085

.054

579 (62.5)
8.9***
63.7***

38.7
50.5

62.3**

47.7

52.3**
57.3**
55.1**

51.9
56.5
47.3

38.72 (11.25)**
41.11 (13.34)**

40.48 (12.59)
39.54 (12.61)

37.1***

52.2

48.1***

56.0

.269c

.212c

.114c

60.8***

50.7

.042c

.085c

.061c

70.9***

37.5

.192c

.168c

.061c

64.9***

42.6

.074c

.062c

.032c

63.6***

48.0

.090c

.037c

.023c

90.2***

57.0

.227c

.166c

.132c

71.7***

41.5

.154c

.081c

.041c

.027

.057

.050

.49 (.83)
.46 (.88)

.50 (.81)
.48 (.87)
(continued)

80

Social Work Research Volume 38, Number 2

June 2014

Table 1: Continued
% of Homeowners on Mean (SD)
of the Covariate by Groupa

Covariate

Employment
status at
baseline (%)
Working
Not working
Income at
baseline (in
$1,000) M
(SD)
Homeowners
Renters
Census tract’s
median house
value M (SD)
Homeowners
Renters
Census tract’s
median rent
value
Homeowners
Renters
Census tract’s
disadvantage
score
Homeowners
Renters

Overall Sample
before
Matching
(N = 1,580)

Sample after
Greedy
Matching
(N = 694)

67.6***
23.1***

48.6
56.5

33.75 (16.04)***
19.61 (12.47)***

23.45 (10.47)
25.06 (12.40)

93577.8 (39617.4)
91673.6 (37737.0)

88966.0 (39158.5)
91224.8 (36264.5)

467.8 (164.90)
461.7 (136.58)

461.9 (184.97)
470.2 (142.96)

.02 (.57)***
.19 (.64)***

.18 (.66)
.14 (.62)

Absolute
Standardized
Difference in
Covariate Means
before Matching
(dx)

Absolute Standardized
Difference in Covariate Means
after Matching (dxm)
Sample after
Optimal Full
Matching
(N = 1,580)

Sample after
Optimal Pair
Matching
(N = 1,308)

.754

.670

.121

.978

.625

.013

.049

.040

.024

.041

.061

.024

.275

.171

.119

a

Each entry is % of homeowners in the categorical covariate, or mean (SD) of the continuous covariate by group.
Race is recoded as three dummy variables using White as a reference.
Education is recoded as seven dummy variables using “11th grade or less” as a reference.
*p < .05. **p < .01. ***p < .001. Chi-square test or independent-sample test two-tailed.
b
c

four-year study period. Results of the sensitivity
analysis were consistent with the reported results
that showed a pattern of slower transition to marriage among homeowners. For the original
unmatched sample, the estimated odds ratio of getting married was 0.556 ( p < .01), indicating that the
odds of getting married were 44.4% lower for the
homeowners than for the renters. Similarly, the
odds ratio for the matched sample created by the
optimal pair matching was 0.609 ( p < .05), also
indicating that, during the study period, homeowners transitioned to marriage more slowly than renters. Using the optimal full matched sample, we
found that the transition to marriage took 0.249
of a year (approximately 2.99 months) longer for
the homeowners than for the renters.

Impact of Homeownership on Timing of
Transition to Divorce

For the study of transitioning from marriage to
divorce, the original sample after multiple imputation of missing data contained 1,115 participants; of
these, 938 (84.1%) were homeowners who either
owned a house at baseline or purchased a house
during the four-year study period and 177 (15.9%)
participants who were renters.
Descriptive statistics of the sample for the study of
transition into divorce, and results from imbalance
checks conducted before and after matching, are
presented in Table 3. Similar to the sample used
to investigate the transition to marriage, the overall
transition-to-divorce sample was not balanced on
various covariates before matching. Significant

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

81

82
Table 2: The Study of Transition to Marriage: Results of the Discrete Time Models and the Hodges–Lehman Aligned Rank Test

Covariate

Social Work Research Volume 38, Number 2

Homeownership (Renter)
Owner
Gender (Female)
Male
Race (White)
African American
Hispanic
Other
Age at baseline
Education at baseline (11th grade or less)
High school diploma
Some two-year college, no degree
Associate’s degree
Some four-year college, no degree
Bachelor’s degree
Some graduate school, no degreea
Graduate/professional degree
Number of children – time varying covariate
Employment status – time varying covariate (not
working)
Working
Income at baseline (in $1,000)
Year indicator variable (year 4)
Year 1
Year 2
Year 3

Estimated Odds Ratio
from the Discrete-Time
Model for the Overall
Sample before Matching
(N = 1,580)

Estimated Odds Ratio
from the Discrete-Time
Model for the Sample
after Greedy Matching
(N = 694)

Estimated Odds Ratio
from the Discrete-Time
Model for the Sample
after Pair Matching
(N = 1,308)

Mean Difference of Time-to-Event
with the Hodges-Lehmann
Aligned Rank Test for the Sample
after Optimal Full Matching
(N = 1,580)

.503***

.447***

.494***

.360***

1.670**

1.033**

1.657**

.445***
1.034
1.021
.962***

.431***
1.059
.887
.965***

.485***
1.061
1.221
.966***

.597*
.668†
.780
.961
.897
.731
.868
.790**

.661
.695
.899
.904
1.863

.768†
1.005

.914
.999

5.522***
.963
1.628*

June 2014

Note: Reference groups are in parentheses.
a
Variable was excluded from the discrete-time model after greedy matching due to small size of event.
†p < .1. *p < .05. **p < .01. ***p < .001. Ttwo-tailed test.

1.090
.862

5.785***
.442†
1.660

.606*
.658
.901
.901
.848
.642
.970
.756**

.807
1.003
7.536***
.880
2.096**

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

Table 3: The Study of Transition to Divorce: Sample Description and Imbalance Check before and after Matching
% of Homeowners on Mean (SD) of the
Covariate by Groupa

Covariate

Number of homeowners
Number of renters
Number (%) of homeowners lost after
matching
Gender (%)
Male
Female
Race (%)
White
Africa American
Hispanic
Other
Age at baseline M (SD)
Homeowners
Renters
Education at baseline (%)
11th grade or less
HS graduated
Some 2yr college no degree
Associate degree
Some 4yr college no degree
Bachelor degree
Some graduate school no degree
Graduate/professional degree
Number of children at baseline M (SD)
Homeowners
Renters
Employment status at baseline (%)
Working
Not working

Overall Sample before
Matching (N = 1,115)

938
177

Sample after Greedy
Matching (N = 256)

Absolute Standardized Difference in
Covariate Means after Matching (dxm)
Absolute Standardized
Difference in Covariate
Means before Matching
(dx)

128
128

49.8
51.7

88.4***
73.6***
79.1***
80.4***

47.0
51.6
54.0
50.0

34.14 (10.01)***
37.82 (11.21)***

38.38 (13.3)
36.96 (11.2)

72.6***
81.6***
86.5***
84.0***
89.0***
90.0***
89.3***
92.2***

54.0
49.4
51.1
46.4
41.7
46.4
57.1
60.0

1.04 (1.10)
.97 (1.18)

.97 (1.17)
.92 (1.16)

88.0***
58.3***

46.7
56.2

Sample after
Optimal Full
Matching
(N = 1,115)

177
177

938
177

.854

761 (81.1)
.653

0 (0)
.021

.296b
.176b
.054b
.345

.176b
.137b
.103b
.033

.124b
.297b
.105b
.064

.115c
.084c
.009c
.116c
.194c
.066c
.167c
.062

.066c
.034c
.032c
.027c
.040c
.023c
.084c
.137

.103c
.030c
.025c
.023c
.114c
.070c
.123c
.141

.636

.030

.192

810 (86.4)
21.7***
88.2***

Sample after
Optimal Pair
Matching (N = 354)

83

(continued)

84
Table 3: Continued
% of Homeowners on Mean (SD) of the
Covariate by Groupa

Covariate

Social Work Research Volume 38, Number 2

Income at baseline (in $1,000) M (SD)
Homeowners
Renters
Census tract’s median house value
Homeowners
Renters
Census tract’s median rent value
Homeowners
Renters
Census tract’s disadvantage score M (SD)
Homeowners
Renters
a

Overall Sample before
Matching (N = 1,115)

Sample after Greedy
Matching (N = 256)

42.65 (19.62)***
27.01 (14.61)***

28.36 (12.02)
29.49 (14.90)

92,254.3 (35,343.9)
90,455.4 (35,104.9)

82,843.8 (29,342.4)
88,765.6 (32,980.2)

458.81 (158.09)
462.79 (127.40)

445.33 (146.40)
455.45 (127.18)

–.05 (0.49)***
.15 (.59)***

.32 (.69)*
.12 (.58)*

Each entry is % of homeowners in the categorical covariate, or mean (SD) of the continuous covariate by group.
Race is recoded as three dummy variables using White as a reference.
Education is recoded as seven dummy variables using "11th grade or less" as a reference.
*p < .05. **p < .01. ***p < .001. Chi-square test or independent-sample t test two-tailed.

b
c

Absolute Standardized Difference in
Covariate Means after Matching (dxm)
Absolute Standardized
Difference in Covariate
Means before Matching
(dx)

Sample after
Optimal Pair
Matching (N = 354)

Sample after
Optimal Full
Matching
(N = 1,115)

.897

.070

.025

.051

.178

.073

.028

.075

.031

.379

.127

.154

June 2014

covariates predicting differences on homeownership include gender, race/ethnicity, age at baseline,
education at baseline, employment status at baseline, income at baseline, and the census tract’s disadvantage score. The table also presents levels of
statistical significance. For categorical covariates,
these levels come from a chi-square test; for continuous variables, they come from an independent
sample t test. As before, both greedy matching
and optimal matching improved sample balances.
Examining the distribution of propensity scores
estimated by the binary logistic regression (that
is, the scores used by the greedy matching), we
observed that the two groups generally had a sizable
region common support. This shared area indicated
that a large proportion of the sample was available
to be used. The nearest neighbor within caliper
matching, by design of this study, used 1-to-1
matching. Therefore, only 128 treated participants
were included in the matched sample.
As in the earlier analysis of the transition to marriage, the analysis of the transition to divorce indicated that the homeowners and the renters had very
different distributions on the estimated propensity
scores generated with generalized boosted regression. Optimal pair matching created 177 matched
pairs that contained all comparison participants,
although this matching procedure lost 761 treated
participants (81.1%); full matching created a
matched sample containing all 938 treated and
177 comparison participants.
All four methods for estimating the timing of
homeowners’ and renters’ transition into divorce
showed consistent findings, regardless of whether
the model controlled for selection bias (see Table 4).
For the original unmatched sample, the estimated
odds ratio of getting divorced was 0.446 ( p < .01),
indicating that the odds of getting divorced were
55.4% lower for homeowners than for renters. Similarly, the odds ratio was 0.306 ( p < .01) for the
matched sample created by greedy matching and
0.404 ( p < .05) for the one created by the optimal
pair matching; both of these findings indicate that,
during the study period, the homeowners transitioned
to divorce more slowly than the renters. Using the
optimal full matched sample, we found that homeowners took 0.068 of a year (approximately 0.816
month longer) longer to get divorces than the renters. The Hodges-Lehmann aligned rank test shows
that this difference was not statistically significant,
although the direction of the mean difference was

consistent with that from other propensity score
models.
The findings showed a high level of consistency
in the impact of treatment on both the transition to
marriage and the transition to divorce. The convergence of findings between the unmatched analysis
and the matched analysis is not uncommon in
observational studies in which the cause has a strong
influence on effect. The matched studies likely
missed important covariates that not only affected
sample selection, but also left the findings prone
to hidden selection bias; nevertheless, our analysis
produces revealing results concerning the impact
of homeownership on the timing of this low- to
moderate-income sample’s transitions into marriage
and divorce. We found that single homeowners
were less likely to marry than single renters and
that married homeowners were less likely to divorce
than married renters. Taken together, these findings
support the economic theory of marriage and
exchange theory explanations for patterns in divorce.
DISCUSSION

This study presents evidence of homeownership’s
relationships with marriage and divorce decisions
among low- to moderate-income households.
The article addresses two important questions: (1)
Are single homeowners less likely to marry than
single renters? and (2) Are married homeowners
less likely to divorce than married renters? Drawing
on Becker’s (1974) economic theory of marriage
and Edwards’s (1969) social exchange theory, we
expected that single homeowners would be less
likely to marry than renters and that alreadymarried homeowners would be less likely to
divorce. Our results confirmed both hypotheses,
indicating that single homeowners have less to
gain financially from marriage than do renters and
that the financial costs of divorce are greater for
married homeowners than for their counterparts
who rent.
The findings suggest that, among our sample,
owning a home reduced the odds of marriage in a
balanced comparison between renters and owners.
According to the economic theory of marriage and
exchange, marriage occurs when both partners perceive the marriage contract as beneficial and likely
to result in new resources. For single homeowners,
marriage entails providing a new partner with the
financial and nonfinancial resources associated with
homeownership. The other partner would need to

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

85

86
Table 4: Final Results of Discrete-time Models and Hodges-Lehmann Test

Covariate

Social Work Research Volume 38, Number 2

Homeownership (Renter)
Owner
Gender (Female)
Malea
Race (White)
African American
Hispanic
Othera
Age at baseline
Education at baseline (11th grade or less)
High school diploma
Some two-year college, no degree
Associate’s degree
Some four-year college, no degree
Bachelor’s degree
Some graduate school, no degree
Graduate/professional degree
Number of children – time varying covariate
Employment status – time varying covariate (not working)
Working
Income at baseline (in $1,000)
Year indicator variable (year 4)
Year 1
Year 2
Year 3

Estimated Odds Ratio
from the Discrete-Time
Model for the Overall
Sample before Matching
(N = 1,115)

Estimated Odds Ratio
from the Discrete-Time
Model for the Sample
after Greedy Matching
(N = 256)

.446*

.306**

.439†
1.289
1.357
.952
1.003

Mean Difference of
Estimated Odds Ratio
Time-to-Event with the
from the Discrete-Time Hodges-Lehmann Aligned Rank
Model for the Sample
Test for the Sample after
after Pair Matching
Optimal Full Matching
(N = 354)
(N = 1,115)

.404*
.425†

.760
.612
1.009

1.039
1.061
.685
.923
.771
.990
.955
.764**

.159
.709
1.355
.769

1.171
.499
.839
.494
.238
.897
.600
.871

1.041
.997

1.300
.995

1.229
1.002

2.780***
.309**
1.186

1.659
.220†
1.163

1.817
.187*
1.583

June 2014

Note: Reference groups are in parentheses.
a
Variable was excluded from the discrete-time model after greedy matching due to small size of event.
†p < .1. *p < .05. **p < .01. ***p < .001. Two-tailed test.

.850
.453
.602

1.082
1.134
1.563
1.008

.068

possess significant resources to offset the costs to the
homeowner of sharing his or her already-acquired
assets. It is simply more difficult for homeowners to
maximize the cost–benefit ratio of marriage than it is
for renters, and single homeowners are therefore less
likely to marry than are single renters.
Our findings also show that homeownership
might provide some protection against the risk of
divorce. Generally, once people commit to homeownership, they prefer to remain in owneroccupied quarters unless there is some unplanned,
typically negative circumstance under which they
are forced to move (Clark & Huang, 2003). Typically, these negative circumstances include divorce
and unemployment. The finding that odds of
divorce are lower among married homeowners
than among married renters confirms a position
held by previous research and theory: Residential
moves occur only if the benefits of moving exceed
the costs (Feijten, 2005). As suggested by social
exchange theory, romantic relationships with
many rewards and few costs tend to be satisfactory
and stable. As such, the combination of homeownership and marriage may act as a barrier against what
is perceived as the less-appealing alternative: The
partners divorce and one or both likely move into
a rented dwelling.
Limitations

Although this research used rigorous analytic methods and longitudinal data, three limitations must be
understood to adequately interpret the findings.
First, although propensity score matching corrects
for selection bias based on observed characteristics,
this analytic method cannot adjust for selection bias
that occurs as a result of unobserved variables.
The second limitation relates to the generalizability of our findings. It is important to acknowledge that the sample used in this study was not a
nationally representative sample; it was representative of the low-income people who received mortgages with favorable terms through the CAP
program and of renters who lived near the sampled
owners. Even though comparisons with the sample
used in the 2004 CPS suggested that our study
sample was very similar to a nationally representative sample of low-income people (Riley & Ru,
2009), further research is needed to determine
whether the findings we report are generalizable
to all lower-income owners and renters and to
other socioeconomic groups.

The third limitation pertains to our assumption
about the respondents used in the first analysis that
modeled marriage. Single respondents are defined as
those who are unmarried, divorced, widowed, or
cohabiting. Although any of these respondents
could, in theory, be in a romantic partnership (and
presumably this is the case for the cohabiting
respondents), each respondent with single status
remains eligible for marriage. Our interest was in
modeling marriage; thus, by definition, all respondents who indicated a single marital status were
appropriate study candidates. From our previous
research, we know that most cohabiters (and we
would extend this claim to noncohabiting romantic
partners) do not typically pool sources of income,
do not accumulate levels of wealth similar to those
of married couples, and do not have legally binding
agreements about their joint economic resources
(Wilmoth & Koso, 2002). These differences suggest
that the effect of being in an unmarried (but possibly
noncohabiting) romantic partnership would be
unlikely to influence the marriage decision.
Implications for Policy and Practice

This study has several implications for social work
policy and practice. First, the study speaks to policy
debates about the costs and benefits of homeownership for lower-income individuals. We found that
homeownership may be linked to the odds of marriage: The odds of marrying were lower among
single owners than among their counterparts who
rented, and the odds of divorce were lower among
married owners ( particularly low-income owners)
than among married renters. These findings suggest
that marriage and homeownership are closely coupled. As practitioners and policymakers contemplate
the roles that the social institutions of marriage and
homeownership may play in antipoverty efforts, it
could be helpful to consider how the “combined
effect” of these institutions could be used to improve
the overall welfare of disadvantaged families
(Grinstein-Weiss, Charles, Guo, Manturuk, &
Key, 2011).
We conclude that a benefit of homeownership is
its function as a stimulus for family stability among
low-income families. Other research has provided
evidence that stable marriages and homeownership
can result in important economic, behavioral, educational, and social gains for both adults and children (Cherlin, 2005). However, such research has
generally conducted separate examinations of

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

87

homeownership and marriage decisions. Given the
valuable gains that may be associated with homeownership and, in turn, with marriage, policymakers may want to devise policies that
simultaneously increase opportunities for marriage
and homeownership. Such social welfare policies
could be especially beneficial to low-income families struggling to maintain family and housing
stability.
The finding that homeownership is associated
with the odds of marriage (the odds are lower for
single owners than for single renters) has several
implications. The findings are aligned with the
rational choice model used to frame this study, suggesting that individuals will choose marriage if sufficient positive economic incentives are present and
if the benefits outweigh the costs. In the context of
social policy, it could be important to increase practitioners’ understanding that single, low-income
homeowners may be less motivated to marry than
their renting counterparts. This understanding
could be especially important for practitioners helping lower-income families achieve stability across a
number of social and economic dimensions. This
finding suggests that other approaches to familial
and economic stability (aside from marriage) could
be warranted.
Implications for Research and
Methodological Advantages

This study used a rigorous analytic approach with
propensity score analysis to address the problem of
self-selection into homeownership. Many studies
examining the effect of homeownership fall short
because the researchers fail to model both the
choice to own a home (over renting) and also
how that choice or selection affects the outcome
of interest. The conventional covariance control
approach commonly used in homeownership studies cannot adjust for selection bias. As such, we used
a methodologically advanced, innovative method
to draw causal inference between homeownership
and the decisions to marry and divorce.
CONCLUSION

The socioeconomic context in which people decide
to marry and divorce remains complex. Competing
motivations, desires, and incentives make the decision highly idiosyncratic. Still, this study suggests
that assets, such as an owned home, may influence
marriage and divorce decisions. In today’s highly

88

fluid and uncertain socioeconomic environment,
homeownership may act as a stabilizing anchor
that makes a person reluctant to engage in major,
disruptive life changes, including the initiation of
divorce or the entrance into marriage. In addition
to offering important new evidence on classic theories of marriage, this article demonstrates rigorous
analytic techniques and offers important insights
into the role that homeownership might play in
the lives of low- and moderate-income individuals.
The institutions of marriage and homeownership
are socially enduring and have significant economic
impact on most American families. Understanding
what, if any, relationship exists between these institutions could be informative to key social welfare
policies aimed at supporting and strengthening the
lives of disadvantaged families.
REFERENCES
Allen, B. L. (2009). Race and gender inequality in homeownership: Does place make a difference? Rural Sociology, 67, 603–617.
Amato, P. R. (2005). The impact of family formation
change on the cognitive, social, and emotional
well-being of the next generation. Future of Children,
15(2), 75–96.
Andersson, G. (2002). Children’s experience of family disruption and family formation: Evidence from 16 FFS
countries. Demographic Research, 7, 343–364.
Becker, G. S. (1974). A theory of marriage. In T. W. Schultz
(Ed.), Economics of the family: Marriage, children and
human capital ( pp. 299–344). Chicago: University of
Chicago Press.
Boehm, T., & Schlottmann, A. (2008). Wealth accumulation
and homeownership: Evidence for low-income
households. Cityscape, 10(2), 225–231.
Bracher, M., Santow, G., Morgan, S. P., & Trussell, J.
(1993). Marriage dissolution in Australia: Models and
explanations. Population Studies, 47, 403–425.
Bumpass, L., & Lu, H. (2000). Trends in cohabitation and
implications for children’s family contexts in the
United States. Population Studies, 54, 29–41.
Caughy, M. O., Hayslett-McCall, K. L., & O’Campo, P. J.
(2007). No neighborhood is an island: Incorporating
distal neighborhood effects into multilevel studies of
child developmental competence. Health and Place,
13(4), 788–798.
Cherlin, A. J. (2005). American marriage in the early
twenty-first century. Future of Children, 15(2),
33–55.
Clark, W. A. V., & Huang, Y. (2003). The life course and
residential mobility in British housing markets. Environment and Planning, 35, 323–339.
Cleek, M. G., & Pearson, T. A. (1985). Perceived causes of
divorce: An analysis of interrelationships. Journal of
Marriage and Family, 147(1), 179–183.
Dew, J. (2007). Two sides of the same coin? The differing
roles of assets and consumer debt in marriage. Journal of
Family and Economic Issues, 28(1), 89–104.
Di, Z. X., & Liu, X. (2007). The importance of wealth and
income in the transition to homeownership. Cityscape,
9(2), 137–152.
Edwards, J. N. (1969). Familial behavior as social exchange.
Journal of Marriage and Family, 31, 518–526.

Social Work Research Volume 38, Number 2

June 2014

Feijten, P. (2005). Union dissolution, unemployment and
moving out of homeownership. European Sociological
Review, 21(1), 59–71.
Feijten, P., Mulder, C. H., & Baizán, P. (2003). Age differentiation in the effect of household situation on firsttime homeownership. Journal of Housing and the Built
Environment, 18, 233–255.
Floyd, F. J., & Wasner, G. H. (1994). Social exchange,
equity, and commitment: Structural equation modeling of dating relationships. Journal of Family Psychology,
8(1), 55–73.
Glaeser, E. L., Laibson, D., & Sacerdote, B. (2002). An
economic approach to social capital. Economic Journal,
112(483), 437–458.
Glaeser, E. L., & Sacerdote, B. (2000). The social consequences of housing. Journal of Housing Economics, 9,
1–23.
Goldstein, J. R., & Kenney, C. T. (2001). Marriage delayed
or marriage forgone? New cohort forecasts of first
marriage for U.S. women. American Sociological Review,
66, 506–519.
Goode, W. J. (1982). The family. Englewood Cliffs, NJ:
Prentice Hall.
Grinstein-Weiss, M., Charles, P., Guo, S., Manturuk, K., &
Key, C. (2011). The effect of marital status on homeownership among low-income households. Social Service Review, 85, 475–503.
Grinstein-Weiss, M., Key, C., Guo, S., Yeo, Y. H., &
Holub, K. (2013). Homeownership and wealth among
low- and moderate-income households. Housing Policy
Debate, 23(2), 259–279.
Guo, S., & Fraser, M. (2010). Propensity score analysis: Statistical methods and applications. Thousand Oaks, CA: Sage
Publications.
Heckman, J. J. (1978). Dummy endogenous variables in a
simultaneous equation system. Econometrica, 46,
931–959.
Heckman, J. J. (1979). Sample selection bias as a specification
error. Econometrica, 47, 153–161.
Hodges, J. L. Jr., & Lehmann, E. L. (1962). Rank methods
for combination of independent experiments in analysis of variance. Annals of Mathematical Statistics, 33(2),
482–497.
Karney, B. R., & Bradbury, T. N. (1995). Assessing longitudinal change in marriage: An introduction to the
analysis of growth curves. Journal of Marriage and Family,
57, 1091–1108.
Kennedy, S., & Bumpass, L. (2008). Cohabitation and
children’s living arrangements. Demographic Research,
19, 1663–1692.
Manturuk, K., Lindblad, M., & Quercia, R. (2010). Friends
and neighbors: Homeownership and social capital
among low- to moderate-income families. Journal of
Urban Affairs, 32, 471–488.
McLanahan, S. (2011). Family instability and complexity
after a nonmarital birth: Outcomes for children in
fragile families. In M. J. Carlson & P. England (Eds.),
Social class and changing families in an unequal America
( pp. 108–133). Palo Alto, CA: Stanford University
Press.
McLanahan, S., & Sandefur, G. (1994). Growing up with a
single parent: What hurts, what helps. Cambridge, MA:
Harvard University Press.
Morgan, S. L., & Winship, C. (2007). Counterfactuals and
causal inference: Methods and principles for social research.
New York: Cambridge University Press.
Neyman, J. S. (1990). On the application of probability
theory to agricultural experiments: Essay on principles,
Section 9 (Dorota M. Dabrowska &Terry P. Speed,
Trans.). Statistical Science, 5, 465–472. (Original work
published 1923)

Qian, Z., & Preston, S. H. (1993). Changes in American
marriage, 1972 to 1987: Availability and forces of
attraction by age and education. American Sociological
Review, 58, 482–495.
Payne, K. K., & Gibbs, L. (2011). First marriage rate in the U.S.,
2010 (F.P.-11-12). Retrieved from http://ncfmr.bgsu.
edu/pdf/family_profiles/file104173.pdf
Raghunathan, T. E. (2004). What do we do with
missing data? Some options for analysis of
incomplete data. Annual Review of Public Health,
25, 99–117.
Riley, S. F., & Ru, H. Y. (2009). Community Advantage Panel
Survey technical sampling report: Owners, 2003–2008.
Chapel Hill: University of North Carolina at Chapel
Hill, Center for Community Capital.
Rohe, W. M., & Stegman, M. A. (1994). The effects of
homeownership: On the self-esteem, perceived
control and life satisfaction of low-income people.
Journal of the American Planning Association, 60(2),
173–184.
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role
of the propensity score in observational studies for
causal effects. Biometrika, 70(1), 41–55.
Royston, P. (2005). Multiple imputation of missing values:
Update of ice. Stata Journal, 5(4), 527–536.
Rubin, D. B. (1974). Estimating causal effects of treatments
in randomized and nonrandomized studies. Journal of
Educational Psychology, 66, 688–701.
Rubin, D. B. (1987). Multiple imputation for survey nonresponse.
New York: John Wiley & Sons.
Rubin, D. B. (1990). Comment: Neyman (1923) and causal
inference in experiments and observational studies.
Statistical Science, 5(4), 472–480.
Rubin, D. B. (2006). Matched sampling for causal effects.
New York: Cambridge University Press.
Sedo, S. A., & Kossoudji, S. A. (2004). Rooms of one’s own:
Gender, race and home ownership as wealth accumulation in the United States (IZA Discussion Paper No.
1397). Retrieved from http://ssrn.com/abstract=
621061
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002).
Experimental and quasi-experimental designs for generalized
causal inference. Boston: Houghton Mifflin.
Sosin, M. R. (2002). Outcomes and sample selection: The
case of a homelessness and substance abuse intervention. British Journal of Mathematical & Statistical Psychology, 55(1), 63–91.
South, S. J., & Lloyd, K. M. (1995). Spousal alternatives and
marital dissolution. American Sociological Review, 60(1),
21–35.
Stevenson, B., & Wolfers, J. (2007). Marriage and divorce:
Changes and their driving forces. Journal of Economic
Perspectives, 21(2), 27–52.
Sykes, L. L. (2005). A home of her own: An analysis of asset
ownership for non-married black and white women.
Social Science Journal, 42(2), 273–284.
Thibaut, J. W., & Kelley, H. H. (1959). The social psychology of
groups. New York: John Wiley & Sons.
Thornton, A. (1988). Cohabitation and marriage in the
1980s. Demography, 25, 497–508.
Townsend, N. (2002). The package deal: Marriage, work, and
fatherhood in men’s lives. Philadelphia: Temple University Press.
van Buuren, S., Boshuizen, H. C., & Knook, D. L. (1999).
Multiple imputation of missing blood pressure covariates in survival analysis. Statistics in Medicine, 18,
681–694.
Wilmoth, J., & Koso, G. (2002). Does marital history matter?
Marital status and wealth outcomes among preretirement adults. Journal of Marriage and Family, 64(1),
254–268.

Grinstein-Weiss et al. / The Impact of Homeownership on Marriage and Divorce

89

Michal Grinstein-Weiss, PhD, is associate professor and
associate director, Center for Social Development, George Warren Brown School of Social Work, Washington University in
St. Louis. Kim R. Manturuk, PhD, is research assistant,
Center for Community Capital, University of North Carolina
at Chapel Hill. Shenyang Guo, PhD, is professor, University
of North Carolina at Chapel Hill. Pajarita Charles, PhD, is
researcher, School of Social Service Administration, University of
Chicago. Clinton Key, MA, is research manager, University of
North Carolina at Chapel Hill. Address correspondence to
Michal Grinstein-Weiss, George Warren Brown School of
Social Work, Washington University in St. Louis, One Brookings Drive, Campus Box 1196, St. Louis, MO 63130;
e-mail: [email protected].
Original manuscript received February 10, 2012
Final revision received August 8, 2012
Accepted October 15, 2012

90

Social Work Research Volume 38, Number 2

June 2014

Copyright of Social Work Research is the property of Oxford University Press / USA and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close