Work and Family at Midlife¹

Judith Fischer, Cheryl Juergens, Rachel Engler, Anisa Zvonkovic

Dept. Of Human Development & Family Studies

Texas Tech University

Lubbock, TX 79409

judith.fischer@ttu.edu

In 1982, Sally (an alias, edited to protect identity) was 24, a newlywed, and a 1980 college graduate. On the Locke Wallace Marital Adjustment (MAT, 1959) she had a 1982 score nicely above the cutoff for a well-adjusted marriage. Over time, her scores declined considerably so that by 1990 she had a MAT score well below the cutoff. And in 1990 she wrote that she was going through a rough time:

Although Sally’s experiences are not completely typical of her fellow college graduates of the classes of 1980 and 1982, the intertwining of work and family is a strong theme in the narratives and in the quantitative data gathered in this research project that has spanned 25 years.

This paper briefly reviews conceptualizations of work and family and approaches to the topic from a person-centered approach (cf., Voydandoff, 2002). Work and family conflict measures are used to support a person-centered conceptualization that is operationalized through cluster analysis. Following a short overview of theory about work and family is a description of the 25-year longitudinal study used for analysis. The discussion of the results concludes with a return to Sally’s narrative.

A traditional approach to work and family involves the idea of "conflict" or "spillover" predicated on the basis that time and energy are finite (Marks, 1977). Marks proposed that time and energy are also capable of expansion; thus, work and family activities may enhance or facilitate each other. Research from the conflict perspective has considered correlates and predictors of work to family and family to work spillover (Netemeyer, Boles, & McMurrian, 1996). The enhancement perspective has seen similar efforts to identify work to family and family to work enhancements (Grzywacz & Marks, 2000). Both approaches, conflict and enhancement, have found support in the literature. When the two perspectives are both measured in one study, the results indicate mixed findings such that there are higher scores on conflicts than enhancements (Boles, Howard, & Donofrio, 2001). But there are also different directions of association depending on the domain studied (work to family or family to work) (Grzywacz & Marks, 2000).

In keeping with MacDermid and Harvey’s (2006) argument that the total role system may be more important than the specific, this study considers that the intersection of the two dimensions is important. The two dimensions are (1) the domain of work and family, and (2) the valence of conflict and enhancement. Rather than conceptualizing these as dimensions where peoples’ scores on each are treated separately, they can be considered together. A person-centered approach would suggest that for some people conflict is evident regardless of domain, and that for others enhancement would better fit their experiences. As well, the domain may be important for some with either work or family holding center stage. And for some, neither or both dimensions could be salient. With two domains, work and family, in two orders (work to family; family to work), and two valences, conflict and enhancement that vary in strength, there are many configurations possible.

A statistical approach, such as cluster analysis, can reduce the theoretically possible combinations into those that fit the data or that support a theory. In this study, the clusters identified are based on measures of work to family conflict and family to work conflict. Unique combinations of these two domains would provide a starting point for considering broader applications of a person-centered approach. Advancing theory about these dimensions through adoption of a person-centered approach is a major goal of this study. Although the larger conceptual model includes enhancement, the study does not have an adequate measure of this valence. Nonetheless, information from the conflict valence with respect to family to work and work to family should indicate the extent to which a person-centered approach provides valuable information. There should be differences across clusters in job satisfaction and marital satisfaction. As well, gender is an important consideration in the work and family literature (e.g., Graham, Fischer, Crawford, & Fitzpatrick, 2000; Maume, 2006; Sweet & Moen, 2006) and will be examined in this study.

In 1980 Ronald Reagan was elected U. S. President. The U.S. Census Bureau reported that the median age at marriage in 1980 was 24.7 for men and 22.0 for women. By way of comparison, the age at marriage rose to 27.1 and 25.3, respectively, in 2003. (U.S. Census Bureau, 2005). VCR’s were just becoming affordable, personal computers were without hard drives, the internet and email were absent from households, and mobile technology was largely nonexistent. Connections to mainframe computers were by very slow (300 baud) telephone modems. In popular culture, the 1982 Academy Award went to Ghandi. The top four TV programs in 1982 were Dallas, M*A*S*H, Magnum P.I. and 60 Minutes. The graduating classes of 1980, 1981, and 1982 met recession in the US economy. "Then and now" mortgage interest rates were about 14% for 1982 compared to the current rate of about 6.25%. Median household income in Texas has risen from $27,016 in the late 1980’s to $39,967 in 2005.

The Longitudinal Study

The study of adult transitions began in 1982 (cf., Fischer, Sollie, Sorell, & Green, 1989) with randomly selected participants from the 1980 and 1982 graduating classes of a large Texas university. Equal numbers of men and women from each cohort were recruited so that there were approximately 100 men from the 1980 graduating class, 100 women from 1980, 100 men from the 1982 graduating class, and 100 women from 1982. The research was designed to follow each participant at roughly equal intervals from 1982 to 1984. A follow-up was conducted in 1990, primarily to consider patterns of child bearing. In 2007, 25 years after the initial study began, participants were followed up again. Table 1 provides information about participation at each wave.

At wave 1, participants were predominately Protestant or Christian (80%), white (96%), from nondivorced families (93%), single or engaged (54%), and childfree (88%). Average age was 23.

Analyses are based on wave 5 data. Of those who returned to the study for wave 5, 82% were Protestant or Christian, 97% were white, 90% were married, number of children averaged 2, length of marriage averaged 22 years and ranged from 1 to 43 years, and 17% had experienced divorce. Seventy-four percent were in professional or managerial occupations, 8% were in occupations such as ranching, religious, clerical, sales, and blue collar, 7% had retired, and 10% were homemakers. Approximately 57% resided in Texas. Average age was 48 years. Only the 153 married individuals with outside-the-home employment were analyzed in this study. Missing data lowered the N for some analyses. More women were retired or homemakers than were men in 2007.

This study examines the wave 5 scores on work to family and family to work conflict, marital satisfaction, and job satisfaction. There were 5 items assessing each of the work/family scales. All conflict items were answered on a 7 point scale from 1 = Strongly Disagree to 7 = Strongly Agree (Netemeyer et al., 1996).

Work to family conflict. A sample item is "The demands of my work interfere with my home and family life". The scale values ranged from 1 to 7 with a mean of 3.90 and SD of 1.51. Cronbach’s alpha was .90.

Family to work conflict. A sample item is "The demands of my family or spouse/partner interfere with work-related activities". The scale values ranged from 1 to 6.6 with a mean of 2.86 and SD of 1.28. Cronbach’s alpha was .86.

The work to family and family to work conflict measures formed the basis for the cluster analyses. Job and marital satisfaction measures were used to examine how participants varied across clusters.

Job satisfaction. There were 4 items answered on a scale from 1 = A lot to 4 = Not at all (Bond, Galinsky, & Swanberg, 1998). These items were reverse scored before the scale was computed so that higher scores indicated greater job satisfaction. A sample item is "I feel a sense of accomplishment from my work". Scores ranged from 1.75 to 4 with a mean of 3.54 and SD of .55. Cronbach’s alpha was .83.

Marital satisfaction. The Locke-Wallace Marital Adjustment Test (Locke & Wallace, 1959) was used to assess marital satisfaction. Items were scaled according to the original criteria for each item. Scores ranged from 29 to 151 with a mean of 110 and SD of 27.66. The cutoff for an adjusted marriage is 110. Cronbach’s alpha was .80.

None of these scales showed evidence of significant skew or the presence of outliers. In order to use them in cluster analyses the work family measures were standardized with a mean of 0 and SD of 1.0. A participant’s positive score meant the person scored above the group average for the scale; similarly, a negative score meant the person scored below the group average. A standardized score of 0 on the work to family conflict scale corresponded to 3.90 (where the scale average could range from 1 to 7). A standardized score of 0 on the family to work conflict scale corresponded to 2.86 on the same 1 to 7 possible range.

First, a hierarchical cluster analysis established the number of clusters through the perusal of a dendrogram. A k-means cluster analysis followed with the number of clusters set equal to three. This produced a set of clusters with reasonably equal Ns and convergence within 6 iterations. Follow-up multiple means comparisons (Tukey, p < .05) established the differences across clusters for each of the two variables that went into the clustering. As expected when clustering and anovas are based on the same data, all comparisons across clusters were highly significant. As well, within-cluster comparisons were made using paired t-tests. In each cluster the two variables were significantly different from each other.

Figure 1 graphs the clusters and identifies each with a label that most closely reflects the makeup of the cluster. Cluster 1 (n=50) is termed low work to family conflict (Lo WF C) in that work to family conflict was quite low and lower than family to work conflict. The second cluster (n=69) is termed moderate work to family conflict (Mod WF C) to indicate that although work to family conflict was greater than family to work conflict, this elevated conflict was moderate in scale. The third cluster (n=33) is labeled high family to work conflict (Hi FW C) based on the high score on this variable, a score that was higher than the moderately elevated work to family score.

Chi-square analyses were calculated to determine if (a) gender and (b) occupation (professional, manager, and other) of the participant were differentially associated with cluster membership. Neither chi-square was significant.

A gender by cluster manova with job satisfaction and marital satisfaction as dependent variables was calculated to determine if gender was a main effect and/or interacted with cluster. The manova was nonsignificant for gender and gender by cluster. The Wilks’ lambda for cluster was significant (Wilks = .91, F(4, 284) = 3.28, p < .05) as was the univariate analysis for marital satisfaction (F(2, 143) = 6.74, p < .01). There were no significant differences on job satisfaction by cluster.

Note. – Matching superscripts in a row are significantly different means, Tukey, p < .05.

As may be seen, marital satisfaction was lowest when family to work conflict was highest. Cluster membership explained 8.6% of the variance in marital satisfaction.

Discussion

The kinds of clusters that reflected the data in this study of highly educated, mostly professional and mostly long-term married people demonstrated that (a) that ratings of work to family conflict and family to work conflict somewhat, but not completely, covaried. Thus, conflict in one of the domains was accompanied by conflict in the other, and lack of conflict in a domain was accompanied by lack of conflict in the other. However, (b) this covariation of conflict across domains was also distinct. Within each cluster, work to family and family to work conflict scores differed significantly. In cluster 1 (Lo WF C), where conflict was generally low, work to family conflict was significantly lower than family to work conflict. But in cluster 3 (Hi FW C), where conflict was generally high, family to work conflict was significantly higher than work to family conflict. The second cluster (Mod WF C) was identified by moderately high work to family conflict but negligible family to work conflict. Thus, each cluster reflected a unique combination of family to work and work to family conflict, variations that would have gone undetected without a more person-centered approach.

These patterns need to be replicated on other samples to determine their generalizability across education levels, job categories, and ages. As well, parallel analyses of the enhancement valence would demonstrate how work-to-family and family-to-work enhancement covary. Including both conflict and enhancement in the same person-centered study would provide considerable expansion of the understanding of how domain and valence are salient in people’s work and family lives. Although membership in a cluster was unrelated to occupation, the job categories in this study were heavily weighted toward the professional level. The lack of variation in the sample may have restricted ability of the chi-square test to detect cluster differences in occupation.

Of great interest is the finding that these clusters were independent of gender. Moreover, gender did not interact with cluster to predict scores on job or marital satisfaction. A follow-up analysis found that number of children did not vary by cluster.

Obtaining clusters that were gender free was not an objective of the study, but the results suggest some gender parity with respect to how family and work are associated with satisfaction with job and marriage. It may be that these findings are an artifact of the select nature of the sample: men and women well advanced in their careers and marriages report similar experiences with conflict and enhancement.

Although job satisfaction was similar across clusters, marital satisfaction was not. The average scores for those in the high family to work conflict cluster 3 were well below the cutoff for a satisfied marriage. This suggests that reducing conflict, especially in this family to work domain could be important to marital satisfaction. In future research it would be valuable to see how enhancement in the family to work domain is related to marital satisfaction. Including other wave 5 variables and measures taken in earlier waves would provide additional information regarding characteristics of members of wave 5 clusters.

Returning to Sally, how did she fare with respect to this quantitative approach? She reported low scores on both family-to-work and work-to-family conflict measures. The cluster analysis placed her in the first cluster, the one characterized by low scores across both measures. But her marital and job satisfaction scores were inconsistent with others in these clusters. On average, her peers were among the higher scorers on marital satisfaction (M = 117) whereas her own marital adjustment was very, very low. At the same time, her job satisfaction was somewhat higher than that of her fellow members of cluster 1. Perhaps her words offer some insight.

In 2007, Sally wrote:

Clearly, Sally’s scores on the conflict scales placed her in the cluster that closely resembled her self-reports. Yet her job and marital satisfaction scores were at odds with others’ scores in her cluster. Perhaps her working part time and going to school meant that her work and family were separate enough not to intrude on each other. She may have compartmentalized her job from her marriage so that she did not experience spillover from one arena to the other. Her seemingly anomalous data encourages looking for moderators such as coping skills, work to family enhancement, extended family and social network supports to gain increased understanding of her circumstances. Sally’s descriptions of her struggles in 1990 and her husband’s situation in 2007 remind researchers of the importance of understanding people in their contexts. Her husband battled addition issues and his job had been fairly recently outsourced. What happens in the global arena affects families, even well educated families in the US.

References

Bond, J. T., Galinsky, E., & Swanberg, J. E. (1998). The 1997 national study of the changing workforce. NY: Families and Work Institute.

Fischer, J. L., Sollie, D. L., Sorell, G. T. & Green, S. (1989). Marital status and career stage influences on social networks of young adults. Journal of Marriage and the Family, 51, 521-534.

Sweet, S., & Moen, P. (2006). Advancing a career focus on work and family: Insights from the life course perspective. In M. Pitt-Catsouphes, E. E. Kossek, & S. Sweet (Eds.). The work and family handbook: Multi-disciplinary perspectives, Methods and approaches (pp. 189- 208). Mahwah, NJ: Lawrence Erlbaum Assoc.