Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

DENISE PREVITI AND PAUL R.

AMATO
Pennsylvania State University

Why Stay Married? Rewards, Barriers, and


Marital Stability

This paper describes people’s open-ended, per- ton & Albrecht, 1991; Levinger, 1965, 1976;
sonal accounts of why they stay married. Most White & Booth, 1991).
people perceived the cohesiveness of their mar- To assess the extent to which these factors op-
riages in terms of rewards and barriers, and few erate in marriages, researchers usually define cer-
people referred to a lack of good alternatives. tain variables as rewards (such as satisfaction with
People who reported barriers only tended to be the spouse as a companion or with the sexual re-
relatively unhappy with their marriages and were lationship) and other variables as barriers (such as
more likely than other individuals to be thinking being religious or having children). Researchers
about divorce. People who reported barriers only, then use these variables to predict whether peo-
compared with people who reported rewards only, ple’s marriages end in dissolution. Because these
were more likely to divorce during the next 14 rewards and barriers are defined in an a priori
years. This association was significant even after fashion, however, it is not clear whether the re-
controlling for marital happiness and divorce wards and barriers studied by researchers are the
proneness. most salient ones for individual husbands and
wives. Although rarely used, another approach is
to ask open-ended questions about why people re-
Over the past few decades, numerous researchers
have used the constructs of rewards, barriers, and main in their marriages. This method has the ad-
alternatives to study marital cohesion. Rewards in- vantage of allowing individuals to describe, in
volve the positive outcomes associated with being their own words, the most important factors that
in a relationship, barriers encompass psychologi- maintain the cohesiveness of their unions.
cal forces that restrain people from leaving rela- The present study used national, longitudinal
tionships, and alternatives reflect the attractiveness data to investigate people’s open-ended responses
of potential partners other than one’s spouse (as to the question, ‘‘What are the most important fac-
well as the attractiveness of having no partner at tors keeping your marriage together?’’ After cod-
all). Many researchers assume that people’s per- ing these responses into a set of categories reflect-
ceptions of rewards, barriers, and alternatives de- ing rewards, barriers, and alternatives, we
termine whether a marriage ends in divorce (Hea- examined the links between these accounts and
the odds of divorcing in subsequent years. Our
Department of Sociology, Pennsylvania State University, goal was to see if people’s open-ended reports in-
211 Oswald Tower, University Park, PA 16802-6207 crease our ability to predict divorce, net of the
(pxa6@psu.edu).
marital quality scales typically used by researchers
Key Words: barriers, divorce, marital cohesion, marital (e.g., scales of marital happiness and divorce
happiness, rewards. proneness). A related goal was to compare peo-

Journal of Marriage and Family 65 (August 2003): 561–573 561


562 Journal of Marriage and Family

ple’s open-ended reports of barriers with people’s leaving their marriages. Spouses in troubled mar-
responses to a series of fixed-choice questions riages, for example, may decide that it is in their
about the importance of specific (and largely ob- children’s best interest for the marriage to contin-
jective) barriers in keeping their marriages togeth- ue. Under these circumstances, spouses explicitly
er. As we show below, people who reported that recognize the role of their children in stabilizing
specific barriers were important tended to be in the marriage.
relatively stable marriages. In contrast, our open- Levinger (1965, 1976) also assumed that the
ended question revealed that individuals who presence (or absence) of alternatives to the mar-
viewed their relationships as being maintained pri- riage can undermine (or strengthen) marital co-
marily by barriers (as opposed to rewards) tended hesion, irrespective of people’s level of marital
to be in marriages that were relatively unstable. satisfaction. According to this perspective, even
We argue that barriers can reflect either greater happily married individuals may be tempted to
marital stability or greater marital instability, de- leave their spouses if they believe that other re-
pending on how they are conceptualized and op- lationships would bring greater rewards. Corre-
erationalized. spondingly, even unhappily married individuals
may remain with their spouses if they believe that
no viable alternatives to the marriage exist. Of
BACKGROUND
course, some spouses may view being alone as a
Rewards, barriers, and alternatives are central desirable alternative to being in an unrewarding
constructs in marital cohesion frameworks (John- marriage.
son, Caughlin, & Huston, 1999; Levinger, 1965, Marital cohesion frameworks have proven to
1976; Rusbult, 1983). Levinger, for example, be useful in understanding and predicting marital
drew on exchange theory to explain why some instability. A large number of studies have shown
marriages remain intact and other marriages end that the rewards of marriage, as reflected in peo-
in divorce. According to Levinger, attraction to a ple’s reports of marital happiness and satisfaction,
spouse is proportional to the rewards received are negatively associated with thoughts of divorce,
from the relationship minus the costs involved in as well as subsequent marital dissolution (Brad-
the relationship. Rewards include pleasant aspects bury, Fincham, & Beach, 2000; Glenn, 1991;
of the relationship, such as companionship, emo- Lewis & Spanier, 1979). Other research supports
tional support, affection, sex, and everyday assis- the assumption that barriers inhibit divorce. For
tance. Costs reflect unpleasant aspects of the re- example, studies show that couples are less likely
lationship, such as having to make compromises, to divorce if they believe in marriage as a lifelong
dealing with verbal or physical aggression, and commitment, have children, own a home together,
putting up with the spouse’s bad habits. In general, have strong religious beliefs, or are economically
people are motivated to remain in marriages when dependent on one another (Heaton & Albrecht,
relationship rewards are high and relationship 1991; Larson & Goltz, 1989; Martin & Bumpass,
costs are low. 1989; South & Spitze, 1986; White & Booth,
Although a low level of rewards may lead to 1991). Finally, research shows that the availability
thoughts of divorce, spouses who wish to end their of attractive alternatives increases the likelihood
marriages must first overcome a variety of barri- of divorce (South & Lloyd, 1995; White &
ers. Barriers include feelings of obligation (to the Booth).
spouse, children, and extended family members); Scales of marital happiness or satisfaction are
moral or religious values; a concern about social often used to represent rewards. For example,
stigma; legal restrictions; and financial depen- Johnson et al. (1999) determined rewards on the
dence on one’s spouse. Levinger (1965, 1976) not- basis of (a) how much the respondent loves the
ed that the strength of barriers matters little when partner, (b) how much the respondent needs the
people are strongly attracted to one another. For partner, (c) the respondent’s marital satisfaction,
example, happily married spouses are unlikely to and (d) the respondent’s ratings of the marriage
think of their children as barriers to leaving the on several dimensions, including miserable-enjoy-
marriage. Instead, mutual feelings of affection, able and interesting-boring. Most studies have op-
along with other rewards, are likely to be the most erationalized barriers with reference to objective
salient cohesive forces. Barriers, according to indicators, such as the presence of children or
Levinger, primarily affect marital cohesion among owning a home. Knoester and Booth (2000) mea-
unhappily married spouses who are thinking about sured barriers by asking respondents to rate the
Why Stay Married? 563

importance of several factors in keeping their mar- wards are satisfied with their relationships. For
riages together, including financial security, reli- this reason, we hypothesized that accounts of re-
gious beliefs, the children’s distress, or giving up lationship rewards are positively associated with
their residences. Although these measurement marital happiness, negatively associated with di-
strategies have proven to be useful, they are based vorce proneness, and negatively associated with
on a priori decisions about the aspects of relation- subsequent divorce.
ships that are rewarding or serve as barriers. Con- Our predictions with respect to barriers were
sequently, it is not clear whether the rewards and more complex. Levinger (1965, 1976) and other
barriers examined by most researchers are the authors have viewed barriers as forces that keep
most salient ones for the individuals being stud- marriages together. Consequently, the presence of
ied. barriers should be associated with greater marital
A rare example of a study based on a subjec- stability. Knoester and Booth (2000) tested this
tive methodology was conducted by Albrecht, idea directly by asking survey respondents to rate
Bahr, and Goodman (1983), who asked previously the importance of eight barriers in keeping their
married individuals to recall the main barriers to marriages together. Although these authors found
their divorces. The most commonly mentioned that ratings of some barriers (religious beliefs, de-
factors included concerns about financial security, pendence on the spouse) were negatively associ-
children, religious beliefs, legal restrictions, and ated with subsequent divorce, ratings of other bar-
the disapproval of friends and family. A limitation riers were not. Furthermore, a summary score
of their study, however, is that post hoc rational- based on all eight barriers did not predict divorce
izations and justifications can bias the recall of significantly. Knoester and Booth suggested that
social events and interpersonal processes. Because the construct of barriers may have less predictive
people’s memories are, to a certain extent, recon- utility than prior researchers have assumed.
structed, a retrospective approach cannot deter- Our study builds directly on the Knoester and
mine whether these barriers affected the happiness Booth (2002) study by using the same data set and
or stability of the marriage while it was intact, or by incorporating the same eight-item measure of
whether overcoming these specific barriers played barriers. We were particularly interested to see
a role in precipitating the divorce. whether people’s ratings of specific, largely ob-
jective barriers (as studied by Knoester and
Booth) and people’s open-ended reports of barri-
GOALS OF THE PRESENT STUDY
ers (which were not included in the Knoester and
Our analysis focused on the main components of Booth study) were related differently to divorce.
Levinger’s (1965, 1976) model: rewards, barriers, As Levinger (1965, 1976) noted, barriers do not
and alternatives. In contrast to prior studies, we become salient until people begin to contemplate
explored people’s responses to an open-ended the costs of ending their marriages. Presumably,
question: ‘‘What are the most important factors happily married people, when reflecting on what
keeping your marriage together?’’ Given that re- keeps them together, focus more on the rewarding
wards, barriers, and alternatives may be defined aspects of the relationship and less on the barriers
idiosyncratically, an open-ended approach gener- to leaving the relationship. In contrast, people who
ates the most salient cohesive forces from the per- are unhappily married are likely to view barriers
spective of the respondents—the prominent fac- as the most important forces holding their mar-
tors that come to mind when people reflect on riages together.
their marriages. The first goal of our study was These considerations suggest that barriers can
descriptive—that is, to report the frequency of either represent cohesive forces or serve as indi-
various rewards, barriers, and alternatives (or cators of marital instability, depending on how
more precisely, lack of alternatives) that spouses they are conceptualized and operationalized.
mention when describing why they stay together. When asked to report whether specific, objective
Our second goal was to see how people’s per- barriers exist in their marriage, we assume that
sonal accounts of marital cohesion are related to people who report a large number of barriers have
marital happiness; divorce proneness (e.g., think- relatively stable marriages, all things being equal.
ing that one’s marriage is in trouble, thinking Consequently, we hypothesized that reports of
about divorce, talking with one’s spouse about di- barriers (when elicited through fixed-choice, ob-
vorce); and divorce. Presumably, people who jective questions) are negatively associated with
view marital cohesion primarily in terms of re- subsequent divorce. But when people are asked to
564 Journal of Marriage and Family

describe, using an open-ended format, what keeps conjunction with barriers may weaken the asso-
their marriages together, reports of barriers should ciation between barriers and divorce. As we de-
be more common among individuals who perceive scribe later, few people in our sample referred to
relatively few intrinsic rewards. Although barriers the absence of alternatives as reasons for remain-
might keep couples together for a period of time, ing married. For this reason, we examined this
unhappy spouses are likely to find ways, eventu- category of responses separately from accounts of
ally, to overcome obstacles to marital dissolution. rewards and barriers.
Because barriers are unlikely to maintain marital Our presentation of results proceeds in three
cohesiveness indefinitely, we hypothesized that steps. First, we present descriptive data on the fre-
open-ended reports of perceived barriers are pos- quency with which people report various reasons
itively associated with subsequent divorce. for staying married. Second, we combine these
The presence of attractive alternatives also can reasons into broad categories reflecting rewards,
contribute to marital instability. We assumed, barriers, and alternatives, and we examine asso-
however, that people who remain in marriages pri- ciations between these categories and marital hap-
marily because of a lack of viable alternatives are piness, divorce proneness, and subsequent divor-
dissatisfied with the intrinsic rewards from their ce. Third, we use event-history methods to see if
relationships. Moreover, although the unavailabil- people’s subjective perceptions of rewards and
ity of alternatives might keep couples together for barriers predict divorce with other dimensions of
awhile, we assume that most people eventually are marital quality controlled, including scales of mar-
capable of locating suitable alternatives to being ital happiness, perceived barriers, and divorce
unhappily married, even if this means being sin- proneness. Our statistical model also controlled
gle. Consequently, we hypothesized that people for demographic variables that may be correlated
who report remaining in their marriages mainly with sources of marital cohesion (rewards, barri-
because of a lack of good alternatives tend to be ers, and alternatives) as well as divorce. These
unhappy with their marriages, are high in divorce variables include gender, age, years married, ed-
proneness, and are particularly likely to divorce. ucation, family income, and race (see Booth,
Our third goal was to see if personal accounts Johnson, White, & Edwards, 1985; Bradbury et
of rewards, barriers, and alternatives predict di- al., 2000; Johnson et al., 1999; and White, 1991,
vorce, net of marital happiness and divorce prone- for relevant studies).
ness. Open-ended responses may capture subtle
relationship characteristics not reflected in stan-
METHOD
dard marital quality scales. We also assumed that
barriers are the most proximal predictors of di-
Sample
vorce, with marital unhappiness leading to
thoughts of divorce, and thoughts of divorce, in This analysis was based on the 17-year longitu-
turn, leading to evaluations of barriers. Although dinal study of Marital Instability Over the Life
we could not test the time ordering of these var- Course (Booth, Amato, & Johnson, 1998). The
iables, we were able to test the hypothesis that original target population consisted of all married
perceiving the cohesiveness of one’s union pri- individuals in households with a telephone in the
marily in terms of barriers is associated with an United States in 1980. In order to qualify for the
elevated level of divorce, even with marital hap- study, spouses had to be living together and both
piness and divorce proneness controlled. spouses had to be age 55 or less. Telephone in-
With respect to hypothesis testing, we distin- terviewers used random digit dialing to select a
guished between people who reported rewards sample of households and a second random pro-
only, people who reported barriers only, and peo- cedure to determine whether to interview the wife
ple who reported rewards as well as barriers. We or husband. Seventeen percent of selected individ-
assumed that people who reported rewards only uals could not be reached after 20 telephone calls.
would have the lowest level of divorce, that peo- Of those individuals contacted, 78% provided
ple who reported barriers only would have the complete interviews. The final sample consisted
highest level of divorce, and that people who re- of 2,034 married persons (not couples). When
ported both rewards and barriers would have an compared with 1980 U.S. Census data, the sample
intermediate level of divorce. The third category, was representative of married individuals with re-
in which rewards and barriers were combined, spect to age, race, urban residence, household
was of interest because thinking of rewards in size, home ownership, and presence of children.
Why Stay Married? 565

The sample was contacted again in 1983, 1988, responses, compared the codes, and obtained a k
1992, and 1997, resulting in subsequent inter- value of 0.89 (p , .001). k is a measure of in-
views with 78%, 66%, 59%, and 53% of the orig- tercoder agreement for categorical data and shows
inal sample, respectively. The interview question how closely the decisions of two judges corre-
on which our analysis was based was included in spond after correcting for chance agreement. This
the 1983 interview. Our analysis included all mar- statistic ranges from 1 (perfect agreement) to 21
ried individuals interviewed in 1983 and 1988, (perfect disagreement), with 0 indicating no as-
along with 94 individuals who declined to be in- sociation (Norusis, 1998). Given the high level of
terviewed in 1988 but provided information on agreement, we coded the remaining cases, allow-
their marital status in that year. The total sample ing multiple responses per individual. The final
size was 1,424. Note also that our sample under- scheme consisted of 18 categories, with 11 cate-
represented recently married individuals, because gories reflecting rewards, six categories reflecting
the shortest duration of marriage in 1983 was 3 barriers, and one category reflecting alternatives.
years. To the extent that our sample overrepre- The categories obtained from the open-ended
sented people in longer duration (and hence, rel- responses are listed in Table 1, along with cate-
atively stable) marriages, our findings probably err gory definitions and examples of responses. Re-
on the conservative side. wards consisted of love, respect, trust, communi-
We relied on Heckman’s widely used method cation, shared past, friendship, happiness,
to adjust for potential attrition bias. We first con- compatibility, emotional security, commitment to
ducted a probit analysis predicting whether re- the spouse (as opposed to commitment to the in-
spondents dropped out of the study using the five stitution of marriage), and sex. Barriers consisted
variables that best predicted attrition: being male, of children, religion, finances, commitment to the
being relatively young, not having a college de- institution of marriage (vs. commitment to the
gree, being non-White, and living in rented ac- particular spouse), interdependent marital roles
commodation. (Note that measures of marital (having a traditional breadwinner-homemaker di-
quality did not predict attrition.) Based on the vision of labor), and other barriers. A single cat-
probit regression equation, we calculated l—the egory turned out to be adequate for the small
probability of attrition for each case in the sam- number of references to alternatives.
ple—and used this as a control variable in all anal-
yses. Controlling for attrition had only trivial ef- Divorce. Our main dependent variable, divorce,
fects on sample estimates and significance tests, was based on whether respondents divorced or
so for the sake of parsimony, we excluded l from separated permanently between 1983 and 1997.
subsequent analyses. During this period, 201 marriages ended in di-
vorce. The year in which the divorce occurred
also was recorded. Because we relied on event
Variables
history methods to analyze these data, we were
Rewards, barriers, and alternatives. The 1983 able to use all respondents who provided infor-
wave of the interviews included the question, mation on marital status in 1988, irrespective of
‘‘What are the most important factors keeping whether they dropped out of the panel in 1992 or
your marriage together?’’ This question appeared 1997 (Allison, 1984).
in the interview schedule after a series of ques-
tions on marital happiness, divorce proneness, and Marital happiness. We relied on an 11-item scale
ratings of specific barriers to divorce. (These to assess marital happiness in 1983. Sample items
questions are described later.) The question was included, ‘‘How happy are you with the amount
open-ended, and interviewers transcribed as many of understanding you receive from your spouse?
factors as the respondent mentioned, with the With the amount of love and affection you receive
number of factors ranging from 1 to 6. The au- from your spouse? With the extent to which you
thors read all of the responses and created a pre- and your spouse agree about things?’’ Responses
liminary set of categories reflecting rewards, bar- were scored in the direction of greater happiness
riers, and alternatives. Based on these preliminary (1 5 not too happy, 2 5 pretty happy, 3 5 very
categories, the two authors independently coded happy), and the mean response served as the scale
100 responses. We then compared the codes, dis- score. This scale had an a reliability of .87. Prior
cussed the discrepancies, and refined the catego- to analysis, we standardized the scale to have a
ries. We then independently coded another 100 mean of 0 and a standard deviation of 1.
566 Journal of Marriage and Family

TABLE 1. PEOPLE’S REASONS FOR STAYING MARRIED: DEFINITIONS AND EXAMPLES OF THE 18 CATEGORIES

Category Definition Example

Rewards
Love Having strong feelings of love and af- ‘‘I guess we love each other. That’s it.’’
fection.
Respect Respecting the spouse and the spouse’s ‘‘We respect each other’s feelings and
needs. privacy.’’
Trust Being honest with each other, trusting ‘‘We trust each other.’’
the spouse.
Communication Understanding, listening, talking about ‘‘We talk to each other before anything
concerns. becomes a problem.’’
Shared past Having a shared past, joined lives, be- ‘‘Our love for each other for 26 years
ing married many years. keeps us together. You just don’t
give that up.’’
Friendship Feeling comfortable together, getting ‘‘We try to spend as much time togeth-
along well, spending time with one er as we can and do things togeth-
another. er.’’
Happiness Feeling happy with one’s spouse or ‘‘I get a charge out of her.’’
one’s marriage.
Compatible Sharing similar goals, interests, or be- ‘‘We have the same goals, principles,
liefs. and beliefs.’’
Emotional security Gaining emotional security from the ‘‘We’re very close and dependent on
marriage, depending on the spouse to one another.’’
meet one’s emotional needs.
Commitment to the partner Being dedicated, faithful, or loyal to ‘‘I want to stay with him. I’m here be-
the spouse. cause I want to be. The same is true
for him.’’
Sex Being physically attracted to the ‘‘Our physical relationship keeps us to-
spouse, enjoying the sexual relation- gether.’’
ship
Barriers
Children Feeling concerned about the children ‘‘The kids are keeping us together.’’
Religion Believing that marriage is sacred, hold- ‘‘We have a personal relationship with
ing religious beliefs that bind the God through Jesus Christ.’’
couple together.
Financial need Being concerned about financial inde- ‘‘I can’t get a job anyplace and depend
pendence or losing one’s home. on his income.’’
Interdependent roles Dividing responsibilities for housework, ‘‘I do the housework and he makes the
child care, and breadwinning through money to pay the bills.’’
a traditional division of labor.
Commitment to marriage Upholding the norm of marital perma- ‘‘Mutual dedication to the institution of
nence, being committed to marriage marriage.’’
(as opposed to being committed to
the particular spouse).
Other barriers Referring to other barriers. ‘‘He’ll kill me if I try to leave.’’
Alternatives Seeing no good alternatives to the mar- ‘‘The fact that it could be worse alone
riage. is what keeps me in this marriage.’’

Divorce proneness. The second scale, divorce scale had an a reliability of .91, with higher scores
proneness, measured the extent to which respon- indicating greater perceived instability. To facili-
dents viewed their marriages as being unstable in tate later interpretation, this scale was standard-
1983. Items dealt with divorce-related thoughts ized to have a mean of 0 and a standard deviation
(such as thinking that one’s marriage is in trouble of 1.
or thinking about divorce) as well as behaviors
(such as discussing divorce with the spouse or dis- Specific barriers. To measure the importance of
cussing divorce with friends and family mem- specific barriers in 1983, we relied on an eight-
bers). Items were summed, but because the re- item scale. People were asked to rate the extent
sulting distribution was positively skewed, the log to which eight factors were ‘‘important in keeping
of the sum served as the scale score. This 13-item your marriage together.’’ These items referred to
Why Stay Married? 567

TABLE 2. PERCENTAGE OF RESPONSES AND RESPONDENTS RESULTS


REFERRING TO 18 CATEGORIES OF MARITAL COHESION

% % Descriptive Results
Category Responses Respondents

Rewards
Table 2 shows the percentage of respondents who
Love 28 60 cited various reasons for staying married. The ta-
Friendship 11 24 ble indicates that people generally thought in
Communication 8 18 terms of rewards when describing their marriages.
Commitment to spouse 6 12 Overall, 74% of responses fell into the rewards
Respect 6 12
Compatibility 5 10 category, 25% of responses fell into the barriers
Trust 4 9 category, and 1% of responses fell into the alter-
Shared past 2 4 natives category. Specifically, love (28%), friend-
Happiness 2 4 ship (11%), and communication (8%) were the
Emotional security 2 5
Sex life 1 1
most frequently mentioned rewards, and children
Barriers (14%) and religion (6%) were the most frequently
Children 14 31 mentioned barriers. Although Levinger’s (1965,
Religion 6 13 1976) framework also includes alternatives, only
Financial need 3 6 a small number of respondents (32 people) re-
Interdependent roles 1 1
Commitment to marriage 0.5 2 ported staying in their marriages because they saw
Other 0.5 1 no good alternatives to living with the current
Alternatives 1 2 spouse. The low representation of responses in
Totals 100 214 this category suggests that few people view the
Note: Responses in column 2 can sum to more than 100 lack of alternatives as a prominent reason for re-
because most people listed more than one reason for stay- maining in their current marriages. Alternatively,
ing married. N 5 3,087 responses and 1,424 respondents. this finding may reflect a degree of social desir-
ability, with people being reluctant to cite a lack
not wanting to leave the residence, financial se- of alternatives as a reason for staying married.
curity, religious beliefs, the spouse’s dependence Overall, 47% of respondents mentioned re-
on the respondent, the respondent’s dependence wards only, 14% of respondents mentioned bar-
on the spouse, the disapproval of friends and fam- riers only, and 38% of respondents mentioned a
ily, concern about whether the children will suffer, combination of rewards and barriers. To simplify
and concern about losing the children (1 5 not later analyses, we created three dichotomous var-
very important, 2 5 somewhat important, 3 5 iables to represent these responses, one for re-
very important). Individuals who did not have wards only (1 5 rewards only, 0 5 other); one
children were given mean scores based on the first for barriers only (1 5 barriers only, 0 5 other);
six items. Responses to these questions were pos- and one for a combination of rewards and barriers
itively correlated, and treating the items as a scale (1 5 rewards and barriers, 0 5 other).
produced an a coefficient of .72. We standardized Table 3 contains correlations between all var-
the scale to have a mean of 0 and a standard de- iables. The three dichotomous variables (rewards
viation of 1. (This instrument was the same as the only, barriers only, and a mix of rewards and bar-
one used in the Knoester and Booth, 2000, study.) riers) were negatively correlated, but because re-
spondents could be in only one category, these
Control variables. The control variables in this correlations reflect logical, rather than substantive,
analysis were gender (0 5 husband, 1 5 wife); associations. More importantly, reporting rewards
age; years married; years of education; family in- only was positively correlated with marital hap-
come in thousands of dollars; and race (0 5 piness and negatively correlated with divorce
White, 1 5 non-White). Wives made up 60% of proneness. Correspondingly, reporting barriers
the sample. The average respondent was about 35 only was negatively correlated with marital hap-
years old; the sample was predominantly White piness and positively correlated with divorce
(12% non-White); the typical respondent had be- proneness. These correlations are consistent with
tween 13 and 14 years of education; mean family our reasoning that people who view the cohesive-
income was about $27,000 (in 1983 dollars); and ness of their unions exclusively in terms of bar-
the average duration of marriage was about 12 riers are relatively unhappy with their marriages
years. and are relatively high in divorce proneness. Peo-
568 Journal of Marriage and Family

ple who reported rewards only scored compara-

12.42
9.25

14
tively low on the eight-item barriers scale, where-
as people who reported barriers only (as well as
a mix of rewards and barriers) scored compara-

.18
27.04
12.95

13

tively high on the eight-item barriers scale. Re-


porting rewards only was negatively associated
with divorce, whereas reporting barriers only was

.36
2.06
13.50
2.53

12

positively associated with divorce. These results


are consistent with the assumption that percep-
tions of rewards and barriers are linked with

2.04
2.08
2.05
0.12
0.32

11

whether couples subsequently dissolve their mar-


riages.
Given the small number of people who men-
2.04
2.01
.25
.84
35.34
9.28

tioned alternatives, we created a single variable


10

for these responses (1 5 reported alternatives, 0


ALL VARIABLES

5 did not report alternatives). People who men-


2.08
.01
2.13
2.06
.03
0.60
0.49
tioned alternatives reported lower levels of marital

9

happiness, higher levels of divorce proneness, and


were more likely to end their marriages in divorce.
FOR

Table 3 also indicates that no correlations between


.02
2.01
2.03
.02
.01
.01
0.14
0.35

8

the demographic control variables and reports of


CORRELATIONS

rewards, barriers, and alternatives were signifi-


cant. This finding suggests that people’s reasons
2.06
2.02
.05
.01
2.01
2.04
.06
0.00
1.00

for staying together are specific to the particular


7

relationship and have relatively little to do with


broad social structural factors such as gender, so-
AND

2.06
.31
.02
.00
.01
.01
.01
.00
0.00
1.00

cial class, race, and age. Of course, people’s rea-


6
TABLE 3. MEANS, STANDARD DEVIATIONS,

sons for staying together may be related to other


Note: Correlations $.06 or #2.06 are significant at p , .05 (two-tailed). N 5 1,424.

structural factors not included in our study.


2.57
2.03
2.25
.02
.02
2.03
2.09
2.01
.05
0.00
1.00

To assess the predictive utility of people’s re-



5

ports of rewards and barriers, we relied on dis-


crete-time, event history methods (Allison, 1984).
We constructed a person-year file, with each year
2.19
.15
2.05
.07
.01
2.03
2.02
.00
.04
2.02
0.02
0.14

that respondents remained in the sample repre-


4

senting a separate observation. Cases were cen-


sored from the data file if they divorced in a par-
2.06
2.01
.01
.13
2.03
2.01
.02
.02
2.01
.03
.03
0.38
0.49

ticular year or dropped out of the panel. Cases that


3

remained in the panel until 1997 were censored in


that year. The total number of person-years was
2.31
.08
2.30
.22
.10
.17
.02
2.04
.01
.03
.00
2.02
0.14
0.35

15,518. We used this file to predict divorce be-


2

tween 1983 and 1997 from people’s reports of


barriers and rewards, the marital quality scales de-
scribed earlier, and the six control variables. We
2.37
2.73
2.04
.19
2.14
2.19
2.09
2.01
2.02
2.04
2.01
2.04
2.02
0.47
0.50

1

also included dummy variables to capture the pas-


sage of time, with the years immediately follow-
ing the interview (1983–1985) serving as the ref-
Non-White respondent

erence category.
Rewards and barriers

Divorce 1983–1997
Female respondent
Age of respondent
Divorce proneness

The results of a logistic regression analysis are


Marital happiness

Family income

shown in Table 4. The control variables were not


Years married
Barriers scale
Alternatives

significant predictors of divorce. The dummy var-


Education
Rewards

iables for time indicated that divorce was more


Barriers

common in the years following the interview, with


a significant peak occurring in the early 1990s.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.

SD

Marital happiness was negatively associated with


M
Why Stay Married? 569

TABLE 4. PREDICTORS OF DIVORCE BETWEEN 1983 AND 1997

Predictor B SE Exp (B)

Female 2.028 .168 .972


Age 2.023 .019 .978
Non-White 2.106 .141 .899
Education .015 .034 1.015
Family income .000 .000 1.000
Years married .024 .018 1.025
Years of observation
1983–1985 — — —
1986–1989 .286 .234 1.331
1990–1993 .748 .227 2.110***
1994–1997 .285 .265 1.330
Marital happiness 2.166 .089 .847
Divorce proneness .605 0.92 1.831***
Barriers scale 2.197 .090 .821*
Rewards only — — —
Barriers only .529 .245 1.70*
Barriers and rewards .148 .190 1.160
Constant 24.786
Note: N 5 15,518 person years. Coefficients are based on logistic regression. Significance tests are two-tailed. Model x2
5 144.985, df 5 14, p , .001.
*p , .05. **p , .01. ***p , .001.

divorce, with a 1 standard deviation increase in representing these data. In one analysis, for ex-
marital happiness predicting a 15% decline in the ample, we used a variable based on the percentage
odds of divorce, although this association only ap- of open-ended responses that reflected barriers.
proached statistical significance (p 5 .062). Di- This analysis revealed that the percentage of bar-
vorce proneness was significantly linked with di- riers was a significant predictor of divorce, with
vorce (p , .001), with a 1 standard deviation the difference between people who reported 100%
increase in divorce proneness predicting an 83% barriers and 0% barriers representing a doubling
increase in the odds of divorce. In addition, the in the odds of divorce (odds ratio 5 1.97, p ,
eight-item barriers scale was negatively associated .01). Similarly, the proportion of responses that
with divorce (p , .05), with a 1 standard devia- represented rewards was a significant predictor of
tion increase in barriers predicting an 18% reduc- divorce, with the odds of divorce being lower for
tion in the odds of divorce. individuals who reported 100% rewards than for
Table 4 also indicates that people’s open-ended individuals who reported 0% rewards (odds ratio
reports in 1983 were associated with divorce be- 5 .46, p , .01). These results indicated that our
tween 1983 and 1997. Compared with people who findings did not depend on the manner in which
reported rewards only (the omitted comparison barriers and rewards were represented in the anal-
category), people who reported barriers only were ysis.
significantly more likely to divorce (odds ratio 5 Because the number of respondents who re-
1.7, p , .05). People who reported a mix of re- ferred to a lack of good alternatives was small, it
wards and barriers, however, were no more likely was difficult to incorporate these responses into
to divorce than were people who reported rewards the model described in Table 4. Consequently, we
only. A subsequent analysis (not shown) indicated conducted a supplementary analysis in which
that people who reported barriers only were some- mentioning alternatives was examined separately.
what more likely to divorce than people who re- At the bivariate level, mentioning alternatives was
ported a mix of rewards and barriers (odds ratio associated with a 156% increase in the odds of
5 1.40), although this difference was not statis- subsequent divorce. With marital happiness, di-
tically significant (p . .1). vorce proneness, the eight-item barriers scale, and
To assess whether these results depended on the control variables in the model, however, men-
the particular way in which rewards and barriers tioning alternatives was not significantly related to
were categorized, we explored alternative ways of divorce.
570 Journal of Marriage and Family

DISCUSSION cilitate divorce among unhappily married people,


this factor does not seem to have a prominent
Using national data, this study examined people’s place in people’s thoughts about marital cohesion.
personal accounts of the factors that hold their Consistent with our hypothesis, people who
marriages together. In general, people perceived viewed the cohesiveness of their marriages pri-
the cohesiveness of their marriages in terms of marily in terms of rewards tended to be happy
rewards, rather than barriers or alternatives, with with their marriages and were unlikely to be
love, respect, friendship, and good communication thinking (or acting) in ways that might lead to
being cited often. The fact that love was men- divorce. Correspondingly, people who attributed
tioned by the majority of respondents (60%) is not the cohesiveness of their marriages primarily to
surprising. A study by Buss, Shackelford, Kirk- barriers or to the absence of alternatives tended to
patrick, and Larsen (2001) documented the ascen- be relatively unhappy with their marriages and
dance of love as the defining criterion of a good were likely to be thinking (or acting) in ways that
marriage. These researchers administered ques- might lead to divorce. This finding is consistent
tionnaires to samples of college students, begin- with Levinger (1976), who pointed out that per-
ning in 1939 and continuing through the 1990s. ceptions of barriers become important psycholog-
In the early surveys, out of 18 characteristics, fe- ical forces only when people are unhappy and be-
male students ranked love in fifth place and male gin to think about leaving their marriages.
students ranked love in fourth place as the most Moreover, consistent with our hypothesis, think-
important reason for marrying. But by the 1970s ing about marital cohesion exclusively in terms of
(and continuing through the 1990s), students of barriers predicted divorce up to 14 years later,
both genders ranked love in first place. Moreover, even after controlling for marital happiness and
the emphasis on friendship as a reward is consis- divorce proneness. We assume that the perception
tent with the work of Gottman and Silver (1999), of barriers is significantly associated with divorce
which stresses a strong friendship as the founda- because it is the most proximal predictor of di-
tion of a satisfying and stable marriage. Finally, vorce (that is, marital unhappiness is followed by
the importance of good communication is consis- thoughts of divorce, and thoughts of divorce, in
tent with research by marriage therapists on the turn, are followed by a consideration of barriers).
centrality of communication skills in facilitating We also hypothesized that referring to the lack of
positive marital relationships (Bradbury, et al., good alternatives would predict divorce, but this
2000). association was not significant with other vari-
In addition to rewards, some individuals cited ables in the model. Given the small number of
barriers as factors that keep their marriages to- people who mentioned alternatives, however, this
gether. Staying married for the sake of the chil- null result may reflect nothing more than a lack
dren was the most commonly mentioned barrier, of statistical power.
being cited by 31% of all respondents. Other bar- The finding that barriers are positively associ-
riers included a lack of financial resources, reli- ated with divorce appears to contradict previous
gious beliefs, commitment to the norm of lifelong theory and research that treats barriers as cohesive
marriage, and the maintenance of traditional forces (Heaton & Albrecht, 1991; Martin & Bum-
breadwinner-homemaker roles. Almost no respon- pass, 1989; White & Booth, 1991). When exam-
dents referred to stigma or the disapproval of oth- ined closely, however, our findings are not con-
ers as barriers, which suggests that the social ac- trary to previous work. Barriers such as religious
ceptance of divorce was widespread by 1983. In beliefs, concerns about children, or worries about
general, the types of barriers that emerged from financial independence may act as cohesive forces
the open-ended question were comparable to the to the extent that they keep unhappily married
types of barriers that other researchers have stud- couples together for a period of time. But barriers
ied using objective indicators (Heaton & Albrecht, do not appear to be as powerful as rewards in
1991; Larson & Goltz, 1989; Martin & Bumpass, maintaining cohesion. Without a strong attraction
1989; South & Spitze, 1986; White & Booth, between spouses (as reflected in love, friendship,
1991). In this sense, our data support the validity or positive communication), many people even-
of previous efforts to study barriers. tually find ways to overcome existing barriers and
Few people (2%) referred to the absence of leave their marriages. For example, wives who are
good alternatives as a reason for staying married. economically dependent on husbands may wait
Although the availability of alternatives may fa- until they have developed a sufficient degree of
Why Stay Married? 571

human capital or saved enough earnings to be- Why did our study yield different results? Our
come independent. Similarly, couples that are con- analyses differed in three respects. First, Knoester
cerned about the effects of divorce on children and Booth limited their analysis to individuals in
may wait until their children are older or have left first marriages, whereas we used all married in-
home before divorcing. Because barriers do not dividuals, irrespective of marriage order. This dif-
promote marital cohesion as effectively as re- ference did not explain the divergent finding,
wards, the more barriers that people cite as rea- however, because subsequent analyses (not
sons for staying together (and correspondingly, shown) indicated that marriage order did not in-
the fewer rewards that people cite), the more like- teract with the barriers scale in predicting divorce.
ly these marriages are to end in divorce. There- Second, the two studies used slightly different sets
fore, although barriers represent short-term cohe- of control variables. But even when we used the
sive forces, they also serve as indicators of later same controls as Knoester and Booth, we contin-
marital instability. The fact that perceptions of ued to find a significant effect for the barriers
barriers predict divorce up to 14 years into the scale. Third, the Knoester and Booth study used
future indicates that the process of overcoming data through 1992, whereas we used data through
barriers can take a good deal of time. 1997. The longer time span of our study made it
Our study suggests that barriers operate differ- possible to include an additional 42 cases of di-
ently when assessed in open-ended and forced- vorce in the analysis (a 26% increase in the num-
choice formats. Table 4 shows that when barriers ber of divorces). Because the power of an event
are assessed with a standardized survey format history analysis is a function of the number of
and are treated as a scale, as in the Knoester and events (rather than the number of cases), our anal-
Booth (2000) study, they are associated with less ysis had greater statistical power to detect asso-
divorce. But when barriers are assessed with an ciations in the population (Allison, 1984). Consis-
open-ended format, they are positively associated tent with this explanation, when we restricted our
with divorce. These two methods of assessing bar- analysis to divorces that occurred prior to 1992
riers reflect different ways of conceptualizing and (and omitted divorces between 1992 and 1997),
operationalizing this construct. When people are the association between the barriers scale and di-
asked to rate whether a particular barrier exists in vorce no longer was statistically significant. We
their marriage (e.g., children, owning a house), conclude, therefore, that the discrepancy in find-
their responses refer to relatively objective cir- ings is due to the fact that our analysis had greater
cumstances, irrespective of marital happiness. statistical power.
Contrary to the usual survey questions, however, Like most studies, our investigation has several
when people are asked open-ended questions limitations. First, our classification of responses
about marital cohesion, their responses reflect the may not have reflected the respondents’ intended
thoughts that are most salient to them at the time, meanings in some cases. Although relying on sub-
and these thoughts are bound up with the quality jective accounts allows respondents to express
of the marriage. Our study indicates that the their views in their own words, it is necessary to
thoughts that come to mind when people answer categorize and combine these idiosyncratic re-
this question are good predictors of divorce, net sponses for data analysis. This process may have
of scales of marital happiness and divorce prone- distorted some respondents’ intended meanings.
ness. Because the inclusion of data on people’s Similarly, some rewards and barriers are difficult
subjective perceptions of rewards and barriers ap- to classify. For example, we classified a shared
pears to increase the predictive power of statistical past as a reward because it was our impression
models, survey researchers may wish to explore that these responses primarily reflected satisfied,
the utility of other open-ended questions in future nostalgic memories. Some people may have
research on marital cohesion and stability. viewed a shared past as a barrier, however, in the
Our findings appear to clash with the study by sense that one does not have the time or energy
Knoester and Booth (2000). Their study was to start a new life with a different partner (or by
based on the same data set, and they used the oneself). Although we were able to establish a
same eight-item barriers scale. But contrary to our high degree of intercoder reliability in the present
findings in Table 4, they found that the barrier study, some degree of ambiguity inevitably re-
scale was not a significant predictor of divorce. mains when classifying open-ended responses.
They concluded, on this basis, that barriers may A second limitation is that people’s perceptions
not be useful predictors of marital dissolution. of rewards and barriers may change over time.
572 Journal of Marriage and Family

Because the open-ended question was asked in NOTE


1983 only, we could not determine the stability of
This research was supported by grant 5 R01 AG04146
people’s perceptions or how changes in percep- from the National Institute on Aging and the Pennsyl-
tions might accompany changes in marital hap- vania State University Population Research Institute,
piness or the odds of divorcing. Another limitation with core support from the National Institute of Child
is that we could not assess the strength of each Health and Human Development grant 1 HD282663.
reward or barrier that respondents mentioned. Be-
cause there was no way to assign weights to peo- REFERENCES
ple’s responses, we had to assume that all respons-
Albrecht, S., Bahr, H., & Goodman, K. (1983). Divorce
es carried equal weight—an unlikely assumption. and remarriage: Problems, adaptations, and adjust-
It also is possible that people’s responses were in- ments. Westport, CT: Greenwood.
fluenced by the placement of the open-ended Allison, P. D. (1984). Event history analysis: Regression
questions in the interview schedule. The question for longitudinal data. Beverly Hills, CA: Sage.
Booth, A., Amato, P. R., & Johnson, D. R. (1998). Mar-
on reasons for staying married was located at the ital instability over the life course: Methodology re-
end of a long series of questions on marital qual- port for fifth wave. Lincoln, NE: Bureau of Sociolog-
ity. Because we cannot determine the extent to ical Research.
which people’s responses may have been influ- Booth, A., Johnson, D., White, L., & Edwards, J.
(1985). Predicting divorce and permanent separation.
enced by the preceding questions, we leave this Journal of Family Issues, 6, 331–346.
issue for future researchers to address. Finally, we Bradbury, T. N., Fincham, F. D., & Beach, S. (2000).
had information from only one partner, so we Research on the nature and determinants of marital
could not incorporate the views of both spouses. satisfaction: A decade in review. Journal of Marriage
and the Family, 62, 964–980.
Research on marital cohesiveness and divorce that Buss, D. M., Shackelford, T. K., Kirkpatrick, L. A., &
does not incorporate input from both spouses is Larsen, R. J. (2001). A half century of mate prefer-
necessarily incomplete. ences: The cultural evolution of values. Journal of
Despite these limitations, our study makes sev- Marriage and Family, 63, 491–503.
Glenn, N. D. (1991). Quantitative research on marital
eral contributions to the research literature on quality in the 1980s: A critical review. In A. Booth
marital cohesion. First, using national data, we de- (Ed.), Contemporary families: Looking forward, look-
scribed the most common factors that people see ing back (pp. 28–41). Minneapolis, MN: National
as keeping their marriages together. Rewards (es- Council on Family Relations.
Gottman, J., & Silver, N. (1999). The seven principles
pecially love, respect, friendship, and communi- for making marriage work. New York: Three Rivers
cation) were more common than barriers, and al- Press.
ternatives were mentioned rarely. Second, we Heaton, T. B., & Albrecht, S. L. (1991). Stable unhappy
found that happily married people tended to view marriages. Journal of Marriage and the Family, 53,
747–758.
their marriages mainly in terms of rewards, Johnson, M. P., Caughlin, J. P., & Huston, T. L. (1999).
whereas unhappily married people tended to view The triparate nature of marital commitment: Personal,
their marriages mainly in terms of barriers. And moral, and structural reasons to stay married. Journal
third, we found that people who focused mainly of Marriage and the Family, 61, 160–177.
on barriers in response to an open-ended question Knoester, C., & Booth, A. (2000). Barriers to divorce:
When are they effective? When are they not? Journal
were more likely to see their marriages end in of Family Issues, 21, 78–99.
divorce. In contrast, people who reported a large Larson, L. E., & Goltz, W. (1989). Religious participa-
number of objective barriers in response to a se- tion and marital commitment. Review of Religious
ries of fixed-choice questions were less likely to Research, 30, 387–400.
Levinger, G. (1965). Marital cohesiveness and dissolu-
see their marriages end in divorce. Our research tion: An integrative review. Journal of Marriage and
suggests that reports of barriers based on fixed- the Family, 27, 19–28.
choice questions and reports of barriers based on Levinger, G. (1976). A socio-psychological perspective
open-ended questions are related differently to di- on marital dissolution. Journal of Social Issues, 52,
21–47.
vorce. Finally, the articulation of barriers in re- Lewis, R. A., & Spanier, G. B. (1979). Theorizing about
sponse to an open-ended question about marital the quality and stability of marriage. In W. R. Burr,
cohesion is a good indicator of future marital in- R. Hill, F. I. Nye, & I. L. Reiss (Eds.), Contemporary
stability, even with traditional marital quality theories about the family. Research based theories
(Vol. I, pp. 268–294). New York: Free Press.
scales in the analysis. Given these results, we rec- Martin, T., & Bumpass, L. L. (1989). Recent trends and
ommend the use of more open-ended items in sur- differentials in marital disruption. Demography, 26,
veys on marriage and family life. 37–51.
Why Stay Married? 573

Norusis, M. J. (1998). SPSS 8.0 Guide to Data Analysis. South, S. J., & Spitze, G. (1986). Determinants of di-
Upper Saddle River, NJ: Prentice-Hall. vorce over the life course. American Sociological Re-
Rusbult, C. E. (1983). A longitudinal test of the invest- view, 51, 583–590.
ment model: The development (and deterioration) of White, L. K. (1991). Determinants of divorce: A review
satisfaction and commitment in heterosexual involve- of research in the eighties. In A. Booth (Ed.), Con-
temporary families: Looking forward, looking back
ments. Journal of Personality and Social Psychology, (pp. 141–149). Minneapolis, MN: National Council
45, 101–117. on Family Relations.
South, S. J., & Lloyd, K. M. (1995). Spousal alterna- White, L. K., & Booth, A. (1991). Divorce over the life
tives and marital dissolution. American Sociological course: The role of marital happiness. Journal of
Review, 60, 21–35. Family Issues, 12, 5–12.

You might also like