Wives ’ Work Hours and Marital Dissolution : Differential Effects across Marital Duration

In this article, I ask: Does the effect of wives’ work hours on marital dissolution change across marital duration? Using the first two waves of the National Survey of Families and Households (NSFH), I find only weak evidence that wives’ work hours are associated with higher marital dissolution. The effect, however, is more positive and significant among long-term marriages. In addition, this study also tests whether couples’ gender ideology and marital interaction explain this differential effect of wives’ work hours. The results suggest that couples’ gender ideology does not account for this differential effect of wives’ work hours. The more positive effect among long-term marriages, however, is reduced to insignificance as soon as a marital interaction measure is introduced into the model. This study contributes to broader research in two ways. Despite the weak effect of wives’ work hours on marital dissolution, the buffering effect of marital duration challenges the prior assumption that the effect of wives’ work hours is invariant across marital duration. Second, this study suggests that the more positive effect of wives’ work hours on marital dissolution among long-term marriages can be attributed to couples’ marital interaction in these marriages becoming more important in mediating the effect of wives’ work hours. Given these results, this study suggests that future research should consider the buffering effect of marital duration in understanding the determinants of marital dissolution.


Introduction
Wives' employment has long been considered one of the most important determinants of marital instability.Previous studies have identified three major elements of wives' employment that have been connected to marital dissolution: wives' income, wives' income relative to husbands' income, and wives' work hours (Greenstein, 1990).This study focuses on the effect of the time aspect of wives' employment (i.e., their work hours) on marital dissolution.Researchers have proposed several mechanisms by which wives' work hours may lead to an increase in marital dissolution, such as through an increase in family conflict (Voydanoff, 1988), decreased marital happiness (Booth, Johnson, White, & Edwards, 1985, 1986), or decreased marital interaction (Poortman, 2005).The direction of the relationship, however, is not consistent.For instance, Schoen, Rogers, and Amato (2006) found that wives' full-time employment is associated with greater marital instability, and that changes in wives' employment have no significant effect on how marital quality changes between two waves of data collection.Some other studies, on the other hand, have focused on the possibility of a reverse causal relationship between wives' work hours and marital dissolution (Austen, 2004;Greene & Quester, 1982;Gray, 1995;Johnson & Skinner, 1986;Montalto & Gerner, 1998;Sen, 2000).For instance, using the Panel Study of Income Dynamics, Montalto and Gerner (1998) concluded that expectation of divorce is positively associated with labor force participation among married women, whereas among men, the probability of divorce was found to reduce given labor force participation.
Regardless of these mechanisms and inconsistent conclusions about the direction of the relationship between wives' work hours and marital dissolution, most of these previous studies assume implicitly that the relationship between wives' employment and marital dissolution is invariant across the marital life course.Surprisingly, few studies have challenged this assumption by testing whether the effect of wives' work hours changes across marital duration (Booth et al., 1986;South, 2001;South & Spitze, 1986).Overall, no consistent findings exist.Notably, the study by Bumpass, Martin, and Sweet (1991), which explored the determinants of marital dissolution among the early years of marriage, argued that the effects of work, financial stress, and marital interaction time during the first years of marriage may change in longer durations.
Using life-course perspective, this study asks four main questions: 1) Are wives' work hours correlated with marital dissolution?2) Does the relationship between wives' work hours and marital dissolution differ between short-and longterm marriages?3) If marital duration has a buffering effect, does couples' gender ideology account for the differential effect of wives' work hours across marital duration?4) Finally, is the differential effect of wives' work hours attributed to couples' marital interaction?Overall, I explore these four questions using nationally representative couple-level data from the first two waves of the NSFH.

Theoretical Framework-The Life Course Perspective
The life course approach emphasizes the importance of tim-D.YUCEL ing and the sequencing of events in an individual's life trajectory (Esterberg, Moen, & Dempster-McClain, 1994).Previous studies have used the life course approach to theorize the timing of events in an individual's life, including the transition to divorce (Heaton, 1991;South & Spitze, 1986).Using this approach, the main argument in this paper is that a stressor such as wives' work hours might have differential effects on marital dissolution across the marital life course-in other words, there is a moderating (buffering) effect of marital duration.This study also tests whether particular factors might account for these differential effects across marital duration.Specifically, this study tests for a possible mediating effect of couples' marital interaction and gender ideology in explaining the differential effect of wives' work hours on marital dissolution, across marital duration.

Wives' Work Hours
Previous studies measured the time aspect of wives' employment in different ways, investigating whether the wife participated in the labor force for some time, her average number of hours worked per week, and/or the number of weeks she worked per year.Out of the various dimensions of wives' economic situation, wives' work hours seems to have the strongest association with marital dissolution (Spitze & South, 1985).Prior research findings, however, are not consistent.Whereas some studies showed a positive relationship between hours worked and marital instability (Booth, Johnson, White, & Edwards, 1984;Greenstein, 1995;Spitze & South, 1985), particularly for women who work full time (Schoen, Astone, Rothert, Standish, & Kim, 2002;South & Spitze, 1986), among working couples (based on cross-sectional data; Johnson, 2004) some other research concluded that women's full-time employment does not destabilize happy marriages but only increases the risk of disruption in unhappy marriages (Schoen et al., 2002).Other studies have concluded that there is only a weak effect of wives' work hours on marital dissolution, despite not controlling for marital quality or gender ideology measures (South, 2001).
For instance, using a sample of working couples from the Survey of Income and Program Participation, Johnson (2004) found that the incidence of divorce is much greater when both spouses are working than when only one spouse is employed.In addition, the same study also found that wives' work hours are more highly correlated with divorce than are husbands' work hours.On the other hand, some other research has tested the reverse relationship, i.e., the effect of anticipated divorce risk on labor supply (Greene & Quester, 1982;Montalto & Gerner, 1998;Sen, 2000).Sen (2000) constructed a longitudinal dataset and compared two cohorts: the National Longitudinal Survey of Young Women (NLSYW) for 1968-1983and the National Longitudinal Survey of Youth (NLSY) 1979for 1979-1993.Her measure of divorce risk was a dummy variable indicating whether divorce or separation occurred in the next three years.Her results suggested that the risk of divorce significantly increased labor supply, but by less in the more recent cohort.Using data from the National Longitudinal Survey of Youth 1979 and Cox proportional hazard models, Papps (2006) found that married women are found to work more when they face a high probability of divorce.This relationship holds both over an individual's life-cycle and across people with different inherent risks of divorce.Despite the inconsistent findings, I expect to find a positive effect of wives' work hours on marital dissolution.In addition, I expect to find that this positive relationship will exist after taking into account the demographic and socio-economic control variables of the married couples (Hypothesis 1).

Marital Duration
Few previous studies have tested whether the determinants of marital dissolution depend on marital duration (Heaton, Albrecht, & Martin, 1985;Morgan & Rindfuss, 1985;White & Booth, 1991), and similarly few have asked whether the effect of wives' employment on divorce varies by marital duration (Booth et al., 1986;South, 2001;South & Spitze, 1986).Moreover, studies of the moderating effect of marital duration had inconsistent findings.Whereas some studies found no significant moderating effect of marital duration (Booth et al., 1986;South & Spitze, 1986), South (2001), focusing on married couples observed between 1969 and 1993 by the Panel Study of Income Dynamics (PSID), found that the effect of wives' hours worked on the risk for marital dissolution is greater for longer marital durations and in more recent cohorts.Despite the inconsistency of the conclusions, these studies imply that our understanding of marital dissolution would benefit from further examination of the dependence of its determinants across a marital life course (i.e., marital duration).
Several different possible explanations exist as to why the effect of wives' work hours on marital dissolution might differ depending on marital duration.Because long-term marriages are known to be qualitatively different than short-term marriages, the motives to establish a close relationship differ over time.Whereas some studies have argued that emotional intensity and physical attractiveness play an important role at the beginning of romantic relationships (Kenrick, Linsenmeier, Norman, & Bailey, 2002), others have suggested that positive marital interactions become more important for marital satisfaction in long-term marriages (Karney & Bradbury, 1995).Given these arguments, it would be not surprising to find that the determinants of marital dissolution should also differ between long-term and short-term marriages.Specifically, this study tests the effects of couples' gender ideology and marital interaction (see below) to explore possible explanations for the differential effect of wives' work hours on marital dissolution between short-and long-term marriages.
Despite the dearth of studies testing the buffering effect of marital duration, one recent study by Schmitt, Kliegel, and Shapiro (2007) is useful.Using data from 588 married women and men in middle and old age who participated in the Interdisciplinary Longitudinal Study of Adult Development, the authors found that marital interaction is the strongest predictor of marital satisfaction among long-term marriages.This relationship was also found to be stronger for women than men.By contrast, drawing on the attachment hypothesis, Hill (1988) argues that the effect may be greater in the early years of marriage because the amount of time couples spend together will be most effective when spouses have the fewest shared experiences.Thus, increase in wives' work hours might be more detrimental to couples in the early stages of their marriages.Despite these different approaches, the inconsistent findings suggest that the expected direction is not clear.Overall, I expect that, due to changing life circumstances and roles, the effect of wives' work hours on marital dissolution will differ across D. YUCEL short-and long-term marriages (Hypothesis 2).

Gender Ideology
Gender ideology defines expectations regarding the "appropriate" performance of male and female roles (Greenstein, 1995(Greenstein, , 1996)).Traditional gender-role attitudes stress a strong distinction between the husband-breadwinner and the wife-homemaker-mother roles, their interdependence, and the different power relations given to wives and husbands.Nontraditional ideologies emphasize shared roles for both economic productivity and nurturance, and more equal power relations.Due to the conventional belief that one's gender ideology views are constant and can determine the level of wives' employment and/or marital dissolution, previous studies have treated gender ideology as a moderator in explaining the effect of wives' employment on marital dissolution.Spitze and South (1985), for instance, found that the effect of wives' work hours on divorce was stronger for couples in which the husband disapproved of his wife working, and the relationship was significant only for couples in which the husband disapproved of his wife's working.In addition, Greenstein (1995) concluded that the effect of wives' work hours on marital instability is strongest for nontraditional women.
Although some previous studies on the moderating effect of gender ideology have treated it as a static variable, other studies have examined the effect of changes in gender egalitarianism over time.The belief that people change their gender roles and ideologies both between and within generations is gaining acceptance (Wentworth & Chell, 2005), leading researchers to believe that these roles are not created through biology but mainly develop through environmental influences (Cunningham, 2001).Women are expected to expand their roles when they participate in education and the workplace, which might lead them to shift away from traditional gender roles (Wilkerson, Yamawaki, & Downs, 2009).Further, women's exposure to the labor force is expected to foster more egalitarian gender attitudes (Smith-Lovin & Tickamayer, 1978).Along with the increase in their gender egalitarianism, women, especially those in dual-earning marriages, can expect to experience role strain and less happiness in marriage when they bear the burden of the second shift alone (Hochschild & Machung, 1989).Overall, wives' gender egalitarianism may cause them to experience a sense of unfairness when they feel that they do more than their spouse (Frisco & Williams, 2003) and hence take less satisfaction in their marriages (Lye & Biblarz, 1993).The consequent lower marital quality may lead to a higher likelihood of marital dissolution.Thus, an increase in couples' gender egalitarianism, especially for wives, is expected to mediate the effect of wives' work hours on marital dissolution (Greenstein, 1995;Sayer & Bianchi, 2000).
Despite these perspectives, to my knowledge no previous study has tested whether the mediating effect of gender ideology depends on marital duration.Using the life course perspective, various explanations exist as to why the mediating effect of couples' gender ideology might differ across short-and longterm marriages.Some studies argue that couples usually create gender role ideologies at the early stages of marriage, and thus role strain and stress due to the presence of children might be more likely among those in early marriages (Hatch & Bulcroft, 2004).This argument would suggest that the mediating effect of couples' gender ideology would be stronger for partners in short-term marriages.However, according to South (2001), while women may be more committed to traditional gender roles in the early years of marriage when they give birth to a child, as marriages age, the wives' commitment to traditional gender ideologies is likely to disappear, and dissatisfactions due to the dual burdens of work and family might emerge.
Among long-term marriages, wives may look for new roles beyond the traditional maternal role, due to retirement or children leaving the household.During this stage in their lives, they may also experience different expectations from their husbands.These changes could lead wives to be less satisfied with the division of labor in the home and thus to develop more egalitarian gender ideologies during the later stages of their marriages.These different approaches suggest that the direction of the mediating effect (i.e., whether it is stronger or weaker across short-and long-term marriages) is not clear.Overall, I hypothesize that couples' gender ideology will explain some of the differential effect of wives' work hours on marital dissolution across short-and long-term marriages (Hypothesis 3).

Marital Interaction
Another possible mediating mechanism between wives' work hours and marital dissolution is the time pressure due to wives' employment (Poortman, 2005).When wives work longer hours outside the home, they may feel more pressure to balance work and family roles.Thus, working longer hours may cause wives to sacrifice time spent with their respective spouses.According to the attachment hypothesis, a decrease in marital interaction may lead to a higher risk of divorce, because shared time is crucial in the fostering of communication and attachment between spouses (Hill, 1988;Kingston & Nock, 1987).Likewise, marital interaction is a key factor affecting marital instability (Booth et al., 1984;Gager & Sanchez, 2003;Hill, 1988;Kingston & Nock, 1987;Poortman, 2005), and couples who interact fewer hours a week are more likely to dissolve their marriages (Spitze & South, 1985).
To my knowledge, no previous studies have empirically tested whether couples' marital interaction mediates the differential effect of wives' work hours on marital dissolution, across short-and long-term marriages.Taking the life course perspective may be useful to evaluate these effects.One explanation for them is that the mediating effect of marital interaction might be stronger for couples in the early stage of their marriage.The reason is that shared leisure time could be most effective early in marriage when the amount of shared experience between couples is lowest (Hill, 1988).Other possible explanations also exist.Spouses may need to develop a strong bond before one or both spouses start working longer hours, or marriages that last longer might be stronger in general (Presser, 2000).The mediating effect of couples' marital interaction, however, might also be stronger for couples in their later years of marriage, and marital interaction may become more important in maintaining marriages among middle-aged and older individuals.Considering that most of these couples have children who are older or leaving the nest, they may need more spousal support and interaction.
Levinger's social exchange theory (1979) argues that the effect of marital attraction is stronger for couples in longer marriages where there are more barriers and fewer alternatives to marital dissolution.Using the same approach, one can also consider marital interaction as a unique aspect of marital attrac-tion.Thus, we might expect to find that the effect of marital interaction might be stronger for long-term marriages, which was also supported in prior literature (Schmitt et al., 2007) (see also the Marital Duration section).Despite these different approaches, the predicted direction of the mediating effect is not clear.I expect to find that couples' marital interaction will also explain some of the differential effect of wives' work hours on marital dissolution across short-and long-term marriages (Hypothesis 4).

Extending Prior Research
This study extends the prior literature in several ways.First, this study emphasizes a life-course perspective in studying the determinants of marital dissolution.Specifically, it tests whether the effect of wives' work hours on marital dissolution differs across marital duration.Second, this study also tests the possible mediating effects of couples' gender ideology and marital interaction in understanding the differential effect of wives' work hours on marital dissolution in long-term and short-term marriages.Third, rather than using married individuals, this study is based on couple-level data; it uses reports from both husbands and wives of the first two waves of a nationally representative sample.
The family has always been a gendered institution, and research has suggested that the characteristic roles of husbands and wives have different influences on marital disruption (Gager & Sanchez, 2003;Heaton & Blake, 1999;Sanchez & Gager, 2000).Unlike many previous studies, this study focuses on married couples and uses couple-level measures of key variables: gender ideology, marital interaction, and control variables.Using couple-level measures of the key variables enables incorporating perspectives from both spouses.In addition, it enables differentiation of the effects of each spouse's views on marital dissolution, when they differ in their reported marital interaction and gender ideology.Lastly, most prior studies have used more conventional methods to account for missing cases, such as listwise deletion (Presser, 2000) or dummy variable adjustment (Schoen et al., 2002).These might cause biased estimates, especially when data are not missing at random.To address these problems, I use multiple imputation, which permits the maximum number of cases to be retained.

Methods
In this study, I use data from the first two waves of the NSFH, a national sample that includes 13,007 primary respondents aged 19 and older, first interviewed in 1987-1988, with oversamples of blacks, Puerto Ricans, Mexican Americans, single-parent families, families with stepchildren, cohabiting couples, and recently married persons.The second wave, conducted from 1992-1994, provides follow-up interviews of 10,008 primary respondents.The sample used here includes married primary respondents from wave 1 (N = 6877) whose spouses were present and completed the questionnaire (N = 5637).The sample was limited to those married couples whose marital status could be determined at wave 2 because at least one member of the original couple was interviewed in wave 2 (N = 4581).In addition, there were too few individuals who belonged to the American Indian or Asian racial groups, so these 90 individuals (N = 61 husbands, N = 29 wives) were excluded from the sample, leaving a sample of 4491.Finally, because the sample was stratified based on marital duration, cases that lacked information regarding marital duration (N = 4) were deleted, leaving a final sample of 4487 couples.
Using the NSFH dataset has some advantages.First, unlike many other studies, the NSFH collected data from both wives and husbands.Using measures from both husbands and wives allows my analysis to represent joint combinations of partner characteristics, minimizing the multicollinearity bias that emerges from individual measures.Second, the NSFH also contains indicators for many aspects of life, including detailed individual characteristics, marital experiences, employment histories, aspects of employment, and income (Sweet, Bumpass, & Call, 1988).Perhaps most significantly, the NSFH is not only a nationally representative survey with rich indicators of marital quality and gender ideology, but also uses married couples as the unit of analysis.The availability of information from both spouses living in the household is well-suited to the couplelevel analyses of this study.

Handling Missing Data
If the data in a given study were missing completely at random, dropping cases with missing data would not lead to biased estimates (Allison, 2002).However, when the data are not missing completely at random, listwise deletion might lead to potential bias and a loss of statistical power.Because the data do not appear to be missing completely at random in this case, I imputed missing values using the imputation by chained equations (ICE) multiple-imputation scheme available in STATA.This procedure generated five datasets, in which I imputed missing information by regressing each variable with missing data on all observed variables and adding random error to the imputed values to maintain variability.This approach allowed utilization of the entire sample (N = 4487 married couples).The relationship between wives' work hours and marital dissolution is likely reciprocal.Some prior studies suggested that wives might work longer hours as a consequence of unstable marriages and to gain economic independence.This argument has also been empirically tested and confirmed (Greene & Quester, 1982;Johnson & Skinner, 1986;Rogers, 1999).More than two waves of data are needed to establish the causal effect of wives' work hours on marital dissolution.Thus, using the first two waves of data, the results presented in this study should be classified as "correlates" and not true "causes" of marital dissolution.

Variables
I use Wave 1 measures for the key independent variables, as well as for all control variables except the dependent variable, marital dissolution.The dependent variable measures the marital status of the couples at wave 2. This variable distinguishes couples that separated or divorced from those who remained married at wave 2 (1 = those who were separated or divorced at time 2; 0 = those who remained married at time 2).

Independent Variable
The primary independent variable is the wives' work hours.The question asked to both the primary respondent and the spouse was: "How many hours do you usually work per week?"This was treated as a continuous variable.Marital duration was Copyright © 2012 SciRes.D. YUCEL divided into two discrete categories added as a dummy variable: 0 for those in shorter marriages (i.e., marital durations of less than 10 years) and 1 for those in longer marriages (i.e., marriages that last at least 10 years).There are several reasons why I chose 10 years as the cutoff point.First is due to definition by law.Marriages are classified as short-term or long-term marriages based upon the number of years the marriage subsists.The time period differs by state, but generally the cutoff is at or below the ten-year mark.Second, due to the high skewness of marital duration, I used the median of marital duration, which was approximately 10 years in this sample.Third, approximately 75 percent of couples that dissolved their marriages between wave 1 and wave 2 were married for slightly more than 10 years.

Gender Ideology
Both the primary respondents and their spouses were asked how much they agreed with six statements.Each item was coded so that higher scores indicate a more egalitarian gender ideology.The indicators are standardized and summed to create continuous and separate gender ideology indexes for the husbands and wives.The scale ranges from -13.69 to 8.84 for the wives, and from -12.55 to 10.41 for the husbands; the alpha level was 0.67 for wives and 0.65 for husbands.Because the gender distribution is almost normal for both wives' and husbands' gender ideology scales, this index was divided into two equal parts: the lower half indicates traditional gender ideology and the upper half indicates egalitarian gender ideology.A couple-level measure of gender ideology with four dichotomous variables was created (see Table 1).

Marital Interaction
Marital interaction was measured by asking both the primary respondent and the spouse about how often they spend time alone as a couple (1 = never, 6 = almost every day).A dummy variable was coded as 1 for those who answered either "two or three times a week" or "almost every day", and 0 otherwise, for both husbands and wives.Consistent with a strategy in prior studies (Schoen et al., 2002), this is a natural breakpoint since three-fourths of both husbands and wives were in one of these two categories.A couple-level measure of marital interaction was created with four dichotomous variables (see Table 1).

Control Variables
Control variables were selected based on their association with the wives' work hours and marital dissolution in earlier empirical research studies (Booth & Edwards, 1985;Bumpass, 1990;Bumpass et al., 1991;Gershuny, Bittman, & Brice, 2005;Schoen, 1975;Sweet & Bumpass, 1987).These included the spouses' ages at the beginning of their current marriage, wives' education, husbands' education relative to their wives, spouses' race/ethnicity, order of marriage, husbands' work hours, and total household income.In order to capture the similarities and differences between spouses' characteristics, couple-level measures were created for all control variables (see Table 1).This is especially important since marriages between individuals with dissimilar characteristics-age, education, and race, for example-have been found to be less stable (Schoen & Wooldredge, 1989).

Results
As a first step, Table 2 displays the means and standard deviations for the variables included in the analysis for the two subsamples: couples who have been married for less than ten years (i.e., short-term marriages) and couples who have been married for at least 10 years (i.e., long-term marriages).Approximately 12 percent of all marriages dissolved between the two waves of data collection (N = 554).Approximately 75 percent of these divorces (N = 411) occurred in short-term marriages, whereas 25 percent (N = 143) occurred between couples in long-term marriages.On average, most of the demographic and socioeconomic indicators, gender ideology, and marital interaction measures significantly differ between short-and long-term marriages (see Table 2).
As a second step, a series of six logistic regression models were run, to predict whether couples married at wave 1 had separated or divorced by wave 2 from their wave 1 spouses.The unstandardized regression coefficients from the logit analyses of the whole sample (N = 4487) are presented in Table 3.
Hypothesis 1: I expect to find a positive relationship between wives' work hours and marital dissolution.I also expect to find that this positive relationship will exist after taking into account the demographic and socio-economic control variables.
The bivariate correlation (Model 1) between wives' work hours and marital dissolution suggests that there is a positive and significant correlation.The coefficient suggests that likelihood of marital dissolution is 37 percent higher for wives who work 35 hours per week [(e (.009)(35) -1) × 100], compared to those who do not work (p < .001).Model 2 tests the effect of wives' work hours on marital dissolution, after taking the control variables into account.Net of the control variables, marital dissolution is expected to be approximately 28 percent higher [(e (.007)(35) -1) × 100] for wives who work 35 hours per week compared to those who do not work (p < .01).This supports the first hypothesis.
Hypothesis 2: I expect that, due to changing life circumstances and roles, the effect of wives' work hours on marital dissolution will differ across short-and long-term marriages.
Model 3 tests the effect of wives' work hours on marital dissolution, net of the control variables, and also includes the interaction term between wives' work hours and marital duration.The positive and significant interaction effect between wives' employment and marital duration suggests that the effect of wives' work hours on marital dissolution is more significant among long-term marriages (as shown in Model 3).Specifically, the inclusion of this interaction term leads to a significant improvement in the fit of the model (the chi-square change between Model 2 and Model 3 = 3.87 is significant at p < .05 with 1 degrees of freedom (df)).This supports the second hypothesis.With these two findings in mind, the following analyses test some of the mechanisms for explaining the more positive association between wives' work hours and marital dissolution among long-term marriages.
Hypothesis 3: Couples' gender ideology will mediate the differential effect of wives' work hours on marital dissolution across short-and long-term marriages.
When couples' gender ideology measures are added (Model ), the interaction effect between wives' work hours and marital 4 Duration of marriage Dummy variable coded 1 for couples who have been married for 10 years or more (long-term marriages) and 0 for couples who have been married for less than 10 years (short-term marriages).

Gender ideology
Four items indicating how much respondents agree with the first four statements (1 = strongly agree to 5 = strongly disagree) and two items indicating how much they approve of the subsequent two circumstances (1 = strongly approve to 7 = strongly disapprove).a) "It is much better for everyone if the man earns the main living and the woman takes care of the home and family".b) "Preschool children are likely to suffer if their mother is employed".c) "If a husband and a wife both work full-time, they should share household tasks equally".d) "Parents should encourage just as much as independence from their daughters as in their sons".e) "Mothers who work full-time when their youngest child is under age 5".f) "Mothers who work part-time when their youngest child is under age 5".Couple-level gender ideology: Dummy variables (1 = Yes, 0 = No) Both spouses egalitarian Both spouses traditional Wives have more egalitarian views than their husbands.Husbands have more egalitarian views than their wives.Both spouses having traditional views is used as the reference category.

Marital interaction
One item asking, "During the past month, about how often did you and your spouse spend time alone with each other, talking, or sharing an activity?"(1=never to 6=almost every day).
Couple-level marital interaction: Dummy variables (1 = Yes, 0 = No) Both spouses have a high marital interaction Both spouses have a low marital interaction Wives have high and husbands have a low marital interaction Husbands have high and wives have a low marital interaction Both spouses reporting high marital interaction is used as the reference category.Both spouses are in their second or higher marriages .20 *** .40.07*** .25

Age at marriage
Wives are in their first, husbands are in their second or higher marriages .12*** .33 .07*** .25 Husbands are in their first, wives are in their second or higher marriages .12*** .32.05marital duration remains the same (b = .011,p < .05).This result suggests that couples' gender ideology does not mediate the more positive effect of wives' work hours on marital dissolution among long-term marriages.This trend does not support Hypothesis 3. Despite this, couples where both spouses have an egalitarian gender ideology and where wives have a more egalitarian gender ideology than their husbands are approximately 1.4 (p < .05) to 1.5 times (p < .01)more likely to dissolve their marriages than couples where both spouses hold a traditional ideology respectively.Adding couples' gender ideology measures improves the previous model (the difference in chi-square between Models 3 and 4 (11.16) is significant at p < .05 with 3 df).Among couples with dissimilar reports of gender ideology, wives' reports of egalitarian gender ideology predict marital dissolution whereas husbands' reports of egalitarian gender ideology have no effect.Hypothesis 4: Couples' marital interaction will mediate he differential effect of wives' work hours on marital dis-t 18 Copyright © 2012 SciRes.Standard errors are in parentheses.Models 2 -6 include the control variables (husbands' work hours, wives' education, age at marriage, husbands' education relative to wives' education, order of marriage, race-ethnicity, and total household income).Reference groups for the categorical control variables are: short-term marriages (i.e., marriages that have lasted less than 10 years), both spouses report traditional gender ideology, both spouses report high marital interaction, both spouses were 20 years or older when married, wife has a college degree or greater, both spouses are in their first marriage, both spouses are white, and total household income is over $50,000.* p < .05,** p < .01,*** p < .001(two-tailed test).

D. YUCEL
solution across short-and long-term marriages.
Once couples' marital interaction measures are added (Model 5), the positive effect of wives' hours on marital dissolution among long-term marriages is no longer significant (b = .007).Marital dissolution is approximately twice as likely among couples where both spouses report low marital interaction and among couples in which only the wives report low marital in-teraction, compared to couples with both spouses reporting high marital interaction (p < .001).Adding marital interaction measures in Model 5 statistically improves the fit of Model 3, which includes the controls only (chi-square change = 52.19 with 3 df, p < .001).This study supports Hypothesis 4. Couples' marital interaction mediates the stronger, positive effect of wives' work hours on marital dissolution among long-term marriages.
D. YUCEL Among couples with dissimilar reports of marital interaction, wives' reports of low marital interaction predict marital dissolution whereas husbands' reports of low marital interaction have no effect.
After testing the effects of couples' gender ideology and marital interaction separately in Models 4 and 5, respectively, both of these key variables were tested together (Model 6).Controlling for couples' gender ideology and marital interaction, the more positive effect of wives' work hours on marital dissolution among long-term marriages is no longer significant (b = .007).In Model 6, the interaction term between wives' work hours and marital duration remains nonsignificant.In addition, the coefficient size and significance level of couples' gender ideology and marital interaction both remain the same.
Standardized regression coefficients or betas were used to identify the strongest predictors of marital dissolution among the independent variables.Table 4 shows the standardized coefficients of the best fitting model from the prior table (Model 6, Table 3).
Out of all the control variables, the strongest predictors of marital dissolution are wives' education, order of marriage, and age of couples (betas highlighted in Table 4).Of the key independent variables, the marital interaction measure is the best predictor of marital dissolution (beta of .098for both spouses reporting low marital interaction and .085for wives reporting low and husbands reporting high marital interaction).

Discussion
This is the first study that tested the effect of wives' work hours on marital dissolution across marital duration, by using relationship assessments that include both male and female reports.Contrary to many other studies that limited their samples, this study includes both working and non-working wives, the full range available in the NSFH, and both first and higherorder marriages.This is important, because some previous studies' failure to find an effect of wives' employment on marital stability can be explained either by those studies' limited focus on young women (Mott & Moore, 1979) or on short marital duration (South, 2001).My results suggest that the more positive effect of wives' work hours on marital dissolution among long-term marriages is accounted for by couples' reported marital interaction, whereas couples' gender ideology does not have any effect.These results make a theoretical contribution to prior studies by emphasizing the importance of the life course perspective.Specifically, they suggest that future research should consider the changing life circumstances of couples across marital duration, and not assume that the determinants of marital dissolution remain constant.
Despite its contributions, this study has limitations.First, this study faces the problem of incomplete data due to attrition of married and separated or divorced couples, from the NSFH.Limiting the sample to couples with a spouse present who has completed the questionnaire, and for whom the marital status could be ascertained by wave 2, reduced the sample from 6877 potentially married couples to 4581 married couples-almost 67 percent of the total number of married primary respondents at wave 1.A second limitation is the possibility of selection bias.People who did not complete the survey might be more likely to have experienced low marital happiness and/or high marital conflict.Similarly, people who did not participate in the second wave may be more likely to have had lower marital Beta of -.241 for marital duration is not considered in the ranking of beta coefficients, since the marital duration variable is a part of the interaction term between wives' work hours and marital duration.
D. YUCEL quality and to have divorced.Overall, some previous studies concluded that several factors measured in wave 1 of the NSFH significantly predict attrition by wave 2 (Mirowsky & Reynolds, 2000;Simon, 2002).These studies showed that the results differ significantly, depending on whether they ignore attrition or adjust the models to compensate for the hazard of attrition.Lastly, I would need more than two waves of data to rule out the causal role of wives' work hours on marital dissolution.Thus, the results presented in this study should be classified as "correlates" and not true "causes" of marital dissolution.This study also suggests some possible avenues for future research.One would be to use the third wave of NSFH data.The same question could be explored by studying married couples at wave 1 and looking at their outcomes in wave 3, which was conducted from 2001 to 2002.Using all three waves would provide researchers with a dataset spanning fifteen years, and enable them to include changes in employment and marital interaction as well as changes in gender ideology between the first two waves.It would also allow them to analyze the effects of these changes on the marital outcome in wave 3.One could also compare couples married at wave 2 with wave 2 employment patterns, to predict dissolution by wave 3.These analyses would permit examination of whether the adverse effects of wives' work hours may have decreased in recent years (i.e., to test whether the effect of wives' work hours on marital dissolution varies depending on the year of observation, consistent with the findings of South (2001)).Despite the advantages of using all three waves of data, a serious problem would result from the attrition rate.Due to funding constraints, wave 3 does not include any respondents under age 45 as of January 2000 (or their spouses) who did not have a wave 1 focal child eligible to be interviewed at wave 2 (Sweet & Bumpass, 2002).
Overall, the results conclude that socioeconomic indicators such as wives' employment may have differential effects on marital dissolution across marital duration.This more positive relationship between marital quality and stability among longterm marriages (where there is external pressure to remain married due to the existence of greater barriers and fewer alternatives) is consistent with Lewis and Spanier's social exchange theory (1979).The results are also consistent with Schmitt et al. (2007), who also suggest that more research and theoretical development are required to understand better the determinants of marital dissolution across marital duration.Using couplelevel data such as those provided by the NSFH (which collects information from both spouses and measurement approaches) is critical to further progress in this area.Moreover, this study advances the testing of some of the key indicators that explain the differential effect of wives' work hours on marital dissolution across short-and long-term marriages.Future studies should test other indicators that might mediate or moderate the varying effects of wives' work hours on marital dissolution across marital duration, such as the quality of marital interaction (how spouses interact with one another), depression, health status, or social networks.

Table 1 .
Measurement of variables.

Table 2 .
Descriptive statistics of independent variables in the analyses.
* p < .01,*** p < .001.Based on one of the five imputed datasets (N = 4487).I report t tests for the continuous variables and chi square tests for the categorical variables.Letter superscripts represent the reference group for each variable.

Table 3 .
Unstandardized coefficients for the logistic regression of wives' work hours, interaction between wives' work hours and marital duration, gender ideology, and marital interaction on marital dissolution (N = 4487).

Table 4 .
Standardized (beta)coefficients for the logistic regression of wives' work hours, interaction between wives' work hours and marital duration, gender ideology, marital interaction, and control variables on marital dissolution (N = 4487).