Deborah Carr and Jennifer Sheridan

Back to School: A Life Course Approach to Understanding Men's And Women's College Attendance in Adulthood

Report No. 99-437

PSC POPULATION STUDIES CENTER UNIVERSITY OF MICHIGAN

The Population Studies Center at the University of Michigan is one of the oldest population centers in the United States. Established in 1961 with a grant from the Ford Foundation, the Center has a rich history as the main workplace for an interdisciplinary community of scholars in the field of population studies. Today the Center is supported by a Population Research Center Core Grant from the National Institute of Child Health and Human Development (NICHD) as well as by the University of Michigan, the National Institute on Aging, the Hewlett Foundation, and the Mellon Foundation.

PSC Research Reports are prepublication working papers that report on current demographic research conducted by PSC associates and affiliates. The papers are written by the researcher(s) for timely dissemination of their findings and are often later submitted for publication in scholarly journals. The PSC Research Report Series was begun in 1981 and is organized chronologically. Copyrights are held by the authors. Readers may freely quote from, copy, and distribute this work as long as the copyright holder and PSC are properly acknowledged and the original work is not altered.

PSC Publications http://www.psc.lsa.umich.edu/pubs/

Population Studies Center, University of Michigan PO Box 1248, Ann Arbor, MI 48106- 1248 USA

Back to School: A Life Course Approach to Understanding Men's And Women's College Attendance in Adulthood

May 1999

Abstract: The composition of college enrollments has shifted drastically in the past 30 years, as older students account for a rapidly growing share of matriculants. We use data from the Wisconsin Longitudinal Study to examine the factors that influence men's and women's college enrollment and degree receipt at midlife. Using the life course paradigm as a guiding framework, we examine whether family transitions - including divorce, widowhood, caregiving, and empty nest - trigger men's and women's educational transitions. We also argue that family characteristics and macroeconomic influences do not operate in a mechanistic fashion; rather, individual career aspirations must also be incorporated into the study of adult education. Results from event history analyses reveal that divorce and widowhood substantially increase women's risk of midlife college attendance. Women who (at age 35) sought a different, higher status job in the future have an elevated risk of returning to school. For men, few variables predicted returns to college although past military service substantially increased men's risk of returning to school at midlife. The findings suggest that women$ life course is more fluid (and less "orderly7~ than men's, given recent shifts in the opportunities facing current cohorts of midlife women.

Datasets Used: Wisconsin Longitudinal Study (WLS): USA, 1957, 1975, 1993.

Acknowledgements: Authors'names are listed in alphabetical order; both contributed equally. An earlier version of this paper was presented at the 1998 International Symposium on Restructuring Work and the Life Course. The Wisconsin Longitudinal Study has its principal support from the National Institute on Aging (AG 9775), with additional support from the National Science Foundation (SES-9023082), the Spencer Foundation, and the Graduate School of the University of Wisconsin-Madison. A public use version of data from the Wisconsin Longitudinal Study is available from the Data Program and Library Service, University of Wisconsin-Madison or from the Inter-university Consortium for Political and Social Research at the University of Michigan- Ann Arbor. The data may also be obtained through the website http://www.dpls.dacc.wisc.edu.

About the Authors: Deborah Carr, Faculty Associate, Population Studies Center, Assistant Professor, Department of Sociology, University of Michigan, Ann Arbor, MI 48103 (carrds@umich.edu).

Jennifer Sheridan, Doctoral Candidate, Department of Sociology, University of Wisconsin, 1 180 Observatory Drive, Madison WI 53706 (jsherida@ssc.wis.edu).

The composition of college enrollments has shifted drastically in the past 30 years, as older or 'hon-traditional" students account for a rapidly growing share of undergraduate and graduate student bodies. Between 1970 and 1990, the total full-time enrollment for students aged 25 and older grew by 164 percent, compared to just an 18 percent increase for traditional (i.e., under age 25) full-time students. The increase is even more marked for women and for students over age 40; the number of nontraditional female students on college campuses increased by 477 percent between 1970 and 1990, and the enrollment of students ages 40 and older grew by 235 percent during this period. In total, students over age 40 now comprise 11.2 percent of total higher education enrollment (The Institute for Higher Education Policy 1996).

At the aggregate level, adults'increased participation in higher education can be explained largely by demographic factors; the large Baby Boom cohort accounted for the overall growth patterns, particularly for men (Jacobs and Stoner-Eby 1998). Less is known, however, about the individual-level factors which affect men's and women's entry to college at midlife. Although psychologists have examined the influence of personality and social support on adults' return to school (Kasworm 1990; Spannard 1990), and sociologists have studied the marital, family, and job factors that are associated with mothers'returning to school (Bradburn, Moen and Dempster-McClain 1995; Felmlee 1988; Suitor 1988), no study systematically assesses the competing influences of family characteristics and transitions, career aspirations, and macroeconomic factors on men's and women's returns to college at midlife.

We use data from the Wisconsin Longitudinal Study ( W S ) to examine the factors that encourage or constrain two types of educational transitions at midlife (age 35-53): enrollment in a formal degree program, and the receipt of a formal college or university degree. Using the life course paradigm as a guiding framework, our central objective is to examine whether family transitions - including divorce, widowhood, caregiving, young children starting school, and older children's transition to adulthood - impact men's and women's educational transitions. We believe that family transitions which increase the risk of entry to college might be classified as "turning points,"in that they are sparking a redirection in one's life course (Gotlib and Wheaton 1997). We also argue that family characteristics and macroeconomic influences do not operate in a mechanistic fashion; rather, career aspirations and indicators of planful behavior must also be incorporated into the study adult educational transitions. Finally, we contend that adult educational pursuits are an increasingly important life transition, as changes in family, gender, and employment roles create a new, more fluid life course for current and future cohorts of midlife adults. The Importance of a College Education

Educational attainment is a crucial factor in determining one's life chances in the United States. Education is associated with myriad outcomes, including adult occupational status, earnings, and work conditions (Featherman and Hauser 1978); values and attitudes towards childrearing (Baumrind 1968); cognitive development and problem-solving skills (Kohn and Slomczynski 1993; Pascarella and Terenzini 199 1) and both physical and psychological well- being over the life course (Pappas et al. 1993; Ross and Wu 1996; Williams 1990). Although the importance of educational attainment has been documented persuasively, nearly all studies of the consequences of education are based on the assumption that schooling is completed in early

adulthood, typically by one's mid 20s. It is reasonable to assume, however, that at least some of the benefits of education attained in early adulthood may also be reaped through education completed later in the life course. Education achieved later in the life course may improve one's chances for promotion in one's current line of work, may enhance earnings, or might provide psychological benefits such as the fulfillment of a personal goal or interest.

Education completed during adulthood may also provide workers with the credentials or skills to enter new jobs or to engage in new activities at their current jobs. The acquisition of new skills and both general and specific knowledge may be crucial for current cohorts of midlife workers. Demographic projections suggest that within the next decade, the decreasing supply of qualified entry-level workers will not meet the demands of industry (Kutscher and Fullerton 1990). At the same time, if the current rapid changes in technology persist, many older workers may be expected to retool and redevelop their current job skills, or to retrain for new jobs all together (Tolbert 1982).

Occupational opportunities (and the accompanying economic returns) may be greatest for those who receive a degree, compared with those who simply enroll in classes yet who do not obtain a credential. Official credentials are critical to occupational success because employers use formal qualifications to screen and select, especially when job candidates outsupply job openings (Collins 1979). Thus, our analyses will assess the predictors of both college enrollment and degree attainment at midlife. Enhancements and Constraints to Midlife College Attendance

Education achieved during mid-adulthood may provide economic, social, and psychological benefits, yet the processes through which adults pursue higher education is far more complex than it is for traditional college students. Although recent statistics attest that adults comprise a sizeable minority of college students, adult college attendance is still generally considered a non-nonnative behavior (The Institute for Higher Education Policy 1996). Kohli (1986) has reasoned that the 'hormative" life course comprises three sections, based on the traditional timing of work and educational experiences. The early segment (young adulthood) is devoted to education and training for work, the second segment (mid-adulthood) includes continuous work activity, and the final segment (late adulthood) is devoted to leisure and the absence of work activity. If this set of normative prescriptions is widely adhered to, then we might that expect that adults seeking education in the atypical 'Second" or "third stages may find neither the institutional nor interpersonal support for their pursuits (Settersten and Lovegreen 1998). Thus, the enhancements and constraints to educational attainment may be quite different for young adults and older adults.

While family socioeconomic status and related factors such as cultural capital are generally viewed as the primary influences on young adults'college attendance (Sewell and Hauser 1975; Bowles and Gintiss 1976; Bourdieu and Passeron 1977), a diverse set of barriers and supports may exist for adult students. Inflexible work schedules, lack of financial or emotional encouragement from peers and family members, family responsibilities, and rigid pension and Social Security policies which sanction job change are among the possible constraints (Settersten and Lovegreen 1998). Moreover, these constraints may operate differently for men and women. Women traditionally have had primary responsibility for raising children,

thus constraints imposed by family responsibilities may be particularly acute for women. In contrast, men have traditionally held the role of 'breadwinner" and have typically engaged in full-time and continuous employment (Bianchi and Spain 1996). Consequently, men may find that an inflexible work schedule, or rigid work rules pertaining to pension and retirement benefits might hinder their prospects of returning to school (Settersten and Lovegreen 1998).

The 'life cycle7'squeeze might also impede an adult's desire or opportunity to return to school. This is a period, usually in midlife, when adults still have primary financial responsibility for their children (in particular, financing their college education), but also take on responsibility for their aging parents. Earnings may plateau at midlife (Moen and Moorehouse 1983), thus midlife adults are faced with the dual constraints of scarce time and money, both of which are critical to one's decision to return to school.

Yet adults might also have unique economic, social and psychological resources that encourage their return to school. Family members might be a source of financial, instrumental and emotional support. Adults might also possess a unique set of psychological resources that would facilitate their return to school. Clausen (1 99 1) argues that 'blanful competence," or the ability to make informed, rational decisions and set realistic short- and long-term goals, is a commodity that increases with age - though in an asymptotic fashion (Clausen 199 1 : 809). Midlife adults may have a more realistic understanding of both their current occupational opportunities, and the necessary steps to achieve their long- and short-term goals. Previous Studies of Adult Students

The central objective of this paper is to assess how family characteristics and transitions, work conditions, and aspirations impact adult men's and women's college attendance. To date, this inquiry has not been explored systematically. Although several sociological studies explore the correlates of adult college attendance, neither gender contrasts nor personal aspirations are considered in these studies (Elman and ORand 1998; Bradburn, Moen and Dempster-McClain 1995; Felmlee 1988).

Taken together, these past studies offer discrepant conclusions about adult college students. All three (i.e., Bradburn et al. 1995; Felmlee 1988; Elman and ORand 1998) found that discontinuities in one's career and higher levels of baseline education and economic resources increase the risk of returning to school. The effects of family characteristics are not consistent, however: Felmlee (1988) finds that young children are a barrier to returning to college, while Elman and ORand (1 998) find no significant relationship between children and adult college attendance. There are several reasons why discrepant findings are revealed. First, different definitions of adult college attendance are used. Bradburn et al. (1995) examined whether a woman was enrolled in any form of formal schooling (i.e., high school, 2- or 4-year college, graduate school, or job-related training) following one's first birth. Consequently, analyses did not focus on older students per se, but women who enrolled following a first birth. Felmlee (1988) examined women's transitions from full-time work to full-time school only, thus capturing a subgroup of women with the greatest investment in the labor market, and who faced the greatest opportunity costs to college attendance. Felmlee (1988) and Bradburn (1995) focused exclusively on women, while Bradburn further limited her sample to mothers.

Our work will expand upon these studies in several ways. First, our analyses include both

men and women, allowing us to contrast whether the barriers and supports for adult college students differ by gender. Secondly, we will examine two types of college attendance; enrollment and degree attainment. Third, we do not impose rigid life-course constraints on our definition of returns to college; we do not define educational transitions as contingent upon former work activities (e.g., full-time employment) or family events (e.g., a first birth). Finally, we explore the role of career aspirations or 'hlanful behavior9'in guiding one's decision to return to school. Life Course Framework

Our research is guided by the life course paradigm, which provides a unifying framework for understanding influences on and gender differences in adult college attendance. A central theme of the life course paradigm is that human lives comprise a series of interlocking age- graded trajectories, such as work careers and family experiences, that are marked by sequences of events and social transitions (Elder 1994). Its four other themes are that (1) the life course unfolds differently for different birth cohorts, reflecting the intersection of historical, social and individual occurrence; (2) diverse life domains, such as family and work, are interconnected; (3) people are planful and make choices among the set of options presented to them; and (4) the timing and sequencing of role transitions vary for individuals (Elder 1994). InJluence of birth cohort

The life course paradigm emphasizes that human lives unfold differently for distinct birth cohorts. In this study, we look intensively at one cohort by using data from the Wisconsin Longitudinal Study (WLS), a sample of men and women who graduated Wisconsin high schools in 1957. The educational transitions of the WLS sample are guided by the social and economic conditions they faced as they came of age, and as they formed families and careers. Men of this cohort typically were raised to believe that they would find a stable career, and would be the main breadwinner of the family. Women were socialized during childhood to value family and domestic roles; careers would come later, or would be entered into only as a short-term role between finishing school and becoming a wife and mother (Coontz 1992; Baruch, Barnett and Rivers 1983). Between the time the WLS participants graduated high school (in 1957) and the time they reached their 30s (in the 1970s), however, major changes in family, gender and work roles occurred. Just at the time the WLS women had reduced childrearing responsibilities in the mid 1970s, diverse work and educational opportunities opened up to women (Bardwick 1980).' Concurrent demographic trends including rising divorce rates and declining mortality rates mean that the WLS graduates are more likely than earlier cohorts to divorce and to see their parents reach old age, and are less likely to experience the early death of a child or spouse (Cherlin 198 1 : 22; Stone, Cafferata, and Sang1 1987).

The WLS graduates also comprise a birth cohort situated at a unique period in economic history. Most entered the workforce in the late 1950s or early 1960s, a time of economic prosperity in the United States. Yet spells of recession and widespread unemployment in the late 1970s and early 1980s, declining employment in the manufacturing and farming sectors

'Marriage and childbearing are nearly universal experiences for members of the WLS sample; 96 percent had ever married as of age 53, and 94 percent have at least one child.

accompanied by job growth in the service sector, a tripling in the number of corporate mergers and acquisitions between 1976 and 1986 (Mergers and Acquisitions 1987), and the glut of workers created by the Baby Boom (Easterlin 1980) created a context where a substantial proportion of WLS respondents experienced career instability at age 35 and beyond (Carr 1998). Consequently, many may need to obtain further education or to retool for new lines of work. Linked Lives

The concept of 'linked 1ives"refers both to links between adult work and family domains, and between one's early social background and later socioeconomic prospects (Elder 1994). Thus, we examine the extent to which family and spouse characteristics enhance or constrain college attendance in adulthood. Family responsibilities, particularly children, may impact whether women work for pay (Salvo 1984), the number of hours women choose to work (Mason and Kuhlthau 1992; Presser 1988), the types of jobs held (Camarigg and Glass 1992; Desai and Waite 1991), and even one's plans for future work endeavors (Maines and Hardesty 1987; Gerson 1985). Most sociological studies imply that achieving a balance between work and family demand is a challenge faced by adults who are raising young children. However, at midlife family characteristics may also guide employment and education decisions.

We expect women with more children to return to school in midlife at lower rates than other women, although we expect no effect for men. Most marriages among the WLS cohort are still characterized by a sharp inequality in the division of household labor and child care, whereby women perform the larger share of household work and childrearing - even if they also work outside the home for pay ( South and Spitze 1994). Returning to college may also be impacted by spouse characteristics. The financial and social support of a spouse can be instrumental in making the transition to college for midlife adults (Suitor 1988). Returning to college often requires that one either reduces hours of paid employment, or exits employment all together. Having the 'Safety net7'of spousal earnings may enable individuals to return to school. We include indicators of spouse's education, employment and occupational status prior to one's return to school in the analyses. We expect that having a spouse who is employed, has higher education, and a higher status occupation will increase the risk of returning to college.

The concept of 'linked lives" also refers to the assumption that one's childhood characteristics and experiences may have long-term cumulative influences on adult behavior (Elder 1994). Because social background characteristics are widely-documented correlates of early educational attainment and adult labor force behavior, we include a detailed set of background characteristics in the analysis, based on the Wisconsin model of status attainment (Sewell and Hauser 1975). These variables are entered as controls only; the coefficients will not be presented or discussed. Plan ful Activity

The third assumption of the life course paradigm is that men and women are planful and make choices among the set of options presented to them (Elder 1994). We argue that while family and human capital characteristics may impact men's and womenk life transitions, these factors do not operate in a deterministic fashion. One's own plans and goals for the fbture are expected to impact adult college attendance.

We know of no large-scale sociological studies which examine the role of educational

and occupational aspirations in guiding adults're-entry to college. The omission of aspiration indicators is surprising; classic studies of educational attainment among young adults find that both occupational and educational aspirations are a strong predictor of educational outcomes (Sewell and Hauser 1975). For midlife adults, occupational aspirations may be even more meaningful than those held by high school students, given that adults'aspirations often reflect their actual work and life circumstances (Jacobs, Karen and McLelland 1991). Adults may have a more realistic understanding of the necessary steps to achieve their long- and short-term goals (Clausen 1991). Although several studies have attempted to incorporate attitudes or values into the study of midlife college attendance, these studies have measured very broad constructs, such as gender role ideology (Davis and Bumpass 1976), or level of contentment with one's work and family situation (Bradburn et al. 1995). In contrast, the WLS data provide a direct assessment of future occupational goals (at age 35), and most importantly, these aspirations are measured temporally prior to the midlife educational transition.

Career aspiration, education, and work experience variables are included in the analysis. We expect that men and women who, at age 35, desired a career change in the future, will have an increased risk of college attendance at midlife. We expect that women whose goal was to be "keeping house" will be substantially less likely to experience an educational transition, given that advanced education is not necessary for one's pursuit of the goal of homemaker.

Occupational aspirations at age 35 may reflect one's contemporaneous work circumstances and past training, thus education, current (1975) occupational status and self- employment status are controlled. Yet one's occupational goals might not only reflect perceptions about chances for mobility in one's current job (given one's skills and experience). Aspirations might also reflect a greater awareness of broader economic factors. Workers who experienced downturns in the farming and manufacturing industries in the 1970s and 80s may have developed plans to seek new lines of work, and may require further education to pursue these new plans. Thus, we include indicators of whether one worked in farming or manufacturing industries. Transitions and Turning Points

The final theme of the life course paradigm is that transitions are embedded in life histories (Elder and ORand 1995). A conceptual distinction can be drawn between transitions and turning points, however. Transitions are normative events occurring to most people (e.g., marriage, parenthood, death of a parent, etc.), while turning points represent "a change in direction in the life course . . . that has the long-term impact of altering the probability of life destinations" (Gotlib and Wheaton 1997; Haraven and Masaoka 1988). In this paper we reason that family transitions may also be turning points if they "trigger" a return to college, given that returning to college may well have 'the long-term impact of altering the probability of life destinations." Six family transitions and one work-related transition will be considered in this analysis: divorce, widowhood, remarriage, youngest child begins school (age 6), launching adult children into adulthood, onset of caregiving duties, and loss of one's "career" or longest job. Marital Transitions

Divorce and widowhood have important economic and psychological effects on midlife men and women. A woman's economic standard of living may drop from 20 to 70 percent

following divorce, while men's standard of living generally increases slightly, from 10 to 20 percent (Holden 1988; Smock and Holden 199 1). Declines in a woman's standard of living upon marital dissolution may force her to re-invest in her own education and job skills, or to seek well- paying work. Men and women who experience a life stressor such as divorce or widowhood (Holmes and Rahe 1967) may also adapt by channeling their energies into self-fulfillment or personal accomplishments.

We expect that both divorce and widowhood will increase the risk that a woman returns to college. Although the psychological consequences of such events may prove devastating, the economic crises that accompany such transitions may force women to make drastic shifts in their work lives. We expect that remarriage will reduce a woman's risk of returning to college. For men, we also expect that divorce or widowhood will increase the risk of returning to school for a degree, as an increase in standard of living and an increased desire for personal growth might be associated with returning for a college degree. We expect remarriage to have no effect for men. Young Children Starting School

Young children constrain women's labor force options: women with children under age 6 have substantially lower rates of full-time employment than both women with older children and men (Bianchi and Spain 1996). We expect that children's influence on college attendance behaves similarly: we expect that for women, having one's youngest child reach age 6 will increase their own rates of entrance to college. We expect men's rates of college attendance in midlife to be unrelated to children starting school. Launching Children into Adulthood

Launching one's adult children from the family home is considered a "developmental task" of midlife (Havighurst 1948). While early research on midlife portrayed the ''empty nest7' stage as a crisis point when midlife women experienced depression following the loss of their role as mother, recent research describes these years as a time when women seek new opportunities for self-definition (Rubin 1979). The financial pressures facing midlife adults may peak during this time, however. For parents of college students, a 'life cycle squeeze" may occur, where the expenses of college coincide with a plateau in one's earnings (Moen and Moorehouse 1983).

We include two indicators of launching children into adulthood; having one's youngest child turn age 18 and making an economic transfer to any child to defray educational expenses. We expect the indicators to behave in different ways, given that the former is associated with increased time resources while the latter is associated with decreased economic resources. Specifically, we expect that women's risk of an educational transition will increase as a child leaves the family home, yet we expect no effect for men, who (in this cohort) invested less time in the rearing of their children. Having a child leave home would not impact men's daily lives to the extent that it would impact their wives'lives. In contrast, we expect that giving money to a child to defray educational expenses reduces both men's and women's risk of attending school, given the high costs of financing parents'and children's educational pursuits simultaneously. Caregiving

Women are the primary caregivers for the elderly, nonelderly disabled, ill children, spouses and other family members (Lee 1992; Stone, Cafferata and Sang1 1987). Women who

are caring for elderly parents are more likely to leave the work force than men, and are more likely to reduce their work hours (Pavalko and Artis 1997). Past studies have focused on the effects of caregiving on labor supply, yet have not extended their analyses to educational transitions among caregivers. We expect that the onset of caregiving duties will reduce the likelihood that a woman would return to college, given that role overload, or the competing demands between caregiving and attending school may occur. We expect no significant relationship for men, given the gendered nature of caregiving duties. Two indicators of caregiving are considered: providing instrumental help to family members, and financial transfers to parents. We expect financial transfers to parents to also decrease college attendance, given that family economic resources might be scarce during these time periods. Job Loss

Lastly, we examine the impact of one non-family event which we believe may be a significant turning point in the life course of adults: job loss. The involuntary loss of one's job may be accompanied by an emotional response such as depression (Smari et al. 1995), or behavioral adaptations such as seeking a new job, forming one's own business, or returning to school to enhance one's chances of finding a better job in the future (Powell and Driscoll 1979; Burman 1988). We argue that the involuntary loss of a job will be negatively associated with men's return to school. Given that men traditionally have been primary breadwinners, men who lose their jobs may feel heightened pressure to find a new job and earn money, rather than to spend money on education. We expect no effect for women. We expect voluntary job exits to increase men's and women's returns to school; voluntary exits suggest that adults may seek new opportunities in life. DATA AND METHODS Data

Analyses are based on the data from the WLS, a long-term study of a random sample of 10,3 17 men and women who graduated from Wisconsin high schools in 1957. Respondents were interviewed during their senior year in high school (1957), at ages 35-36 years old (1975), and ages 53-54 (1 992-3). Of the 10,3 17 members of the original WLS sample, 9,139 (88.6 percent) were re-interviewed in 1975 and 8,493 (82.3 percent) were re-interviewed in 1992-93. Our analysis is limited to the 4,277 women and 3,741 men who completed the 1957 questionnaire, the 1975 phone interview, and the one-hour long 1992 phone survey. By design, all respondents in our sample graduated from high school; thus all returns to school are returns to college (for either undergraduate or graduate study). Dependent Variables

The transition to college is measured in two ways: enrollment in a degree program, and receipt of a college degree. Each event is recorded in terms of the quarter (i.e., the three-month spell, such as January through March) and year when the transition occurred. Events are recorded by quarter rather than by month because many individuals named the academic semester (e.g., Fall 1980) of the event. We focus on transitions that occurred between ages 35 (1975) and 53 (1993). Although it is difficult to set precise ages when midlife begins or ends, midlife is believed to begin at age 35 (Farrell and Rosenberg 198 1 ; Brooks-Gunn and Kirsh 1984). This 17- year period is long enough to observe a significant amount of change in one's family and work

lives, yet it is not so long that the larger context of surrounding labor force patterns changed dramatically.

First, we examine the predictors of enrolling in college after age 35 (1 975). Any respondent who was already enrolled in 1975 is excluded from this analysis. We are studying the process of making a transition, and thus exclude those who completed their degree after 1975 as the natural course of events given they were in already enrolled school in 1975. The definition is further restricted to those entering college five or more years after they last left school, again, in order to capture returns to school rather than continuation of schooling. Both enrollments and degrees are limited to formal educational programs, in two- and four-year colleges and graduate degree-granting institutions. On-the-job training and certificate programs are not included in our definition.

Second, we examine the predictors of obtaining a college degree after age 35 (1975). Any respondent who was in school during the 1975 interview is excluded from this analysis. To be counted as obtaining a college degree after age 35, a respondent had to have at least ten years between this midlife degree, and any previous degree. This strategy ensures that we are examining college degrees earned in midlife that represent a transition, not a continuation. We excluded respondents who did not report the date they received their degree (20 women and 17 men-see Table 1). Independent Variables Linked Lives

The life course paradigm suggests that adult educational and occupational pursuits are linked to family and social background characteristics (Elder 1994). These factors include: number of children as of 1975, marital status in 1975, and spouse characteristics including educational attainment (attended college, college degree, post-graduate work, and high school graduate as the reference group), (1975) occupational education and income (Hauser and Warren 1997); labor forceparticipation status in 1975 (coded 1 if spouse was NOT in the labor force in 1975), and whether spouse was in the military in 1975 (coded 1 if spouse was in the military in 1975).

We use two conceptually and statistically distinct measures of occupational status: occupational education and occupational income scores (Hauser and Warren 1997). Occupational education is the percentage of persons in a three-digit 1970 Census occupation category who completed one or more year of college Occupational income is the percentage of persons in a 1970 Census three-digit occupation category who earned more than $10,000 a year in 1969, as reported in the 1970 Census. By including two distinct measures of status, rather than an averaged index such as the Duncan SEI scale, we can assess the impact of different dimensions of job status on individual outcomes (Hauser, Sheridan and Warren 1998). The raw percentages are reported in Appendix A; however, in the analysis, these percentages are transformed into a started logit in order to normalize the distribution (Hauser and Warren 1997).

Nine indicators of early family background, based on the Wisconsin model of status attainment (Sewell and Hauser 1975) are controlled in the analysis: father's (or householder's) occupational status (Duncan SEI score) in 1957; whether the respondent's mother worked for pay while the respondent was growing up (coded 1 if yes); family income, measured as the

logged average of family income as reported on Wisconsin tax records for the years 1957-60; father's years of completed education; mother's years of completed education; farm background (coded 1 if respondent's father was a farmer); and family structure in 1957 (coded 1 if the family was not intact); number of siblings (top-coded at 8). Early social background characteristics are important largely because they influence educational and occupational pursuits in young adulthood which, in turn, impact adult educational endeavors. We are not focusing on the direct effects on college returns of these characteristics, and thus will not present the coefficients in our tables. Planful Activity

A central theme of the life course is that individuals make choices given the constraints imposed upon them. The indicators of constraint and choice included here are human capital characteristics, work characteristics and work aspirations. Formal education (as of 1975) is years of schooling completed as of 1975, and is coded as three dummy variables (attended college, college degree, post-graduate work), with high school graduate as the reference category.Menta1 ability is assessed with the Henrnon-Nelson test, administered during respondents'junior year in high school. Employment characteristics include the occupational education and occupational income of the respondent; 19 75 occupation ; and dummy variables signifying in military sewice in 19 75, farmer in 1975, manufacturing in 1975, and self-employed in 1 9 75. For certain occupations, college courses may be required, rather than voluntary. For instance, elementary and secondary school teachers and nurses are required to participate in continuing education. Thus, we include dummy variables indicating whether a respondent has worked as a school teacher or nurse prior to enrolling or receiving a degree.2 These variables are included as controls; coefficients will not be presented.

Career aspirations (1 975) were elicited when respondents were ages 35-36, with the question, 'If you were free to choose, what kind of work would you like to be doing ten years from now." Dichotomous variables indicate whether the respondent hoped to be: working in a different occupation (than one's 1975 or most recent job), keeping houselnot workzng, or did not know their future goals (where the reference group is those who wanted to continue working in their 1975 jobs). Missing values for all independent variables are replaced with the sample mean (or the median, for dichotomous variables), and a dummy variable is constructed to indicate that data are missing for that variable (coefficients for the missing data flags are not presented in our tables, and none are statistically significant). Transitions/Turning Points

The central aim of this paper is to assess whether selected transitions influence one's risk of returning to school in midlife. The transitions considered are: divorce, death of spouse, remarriage, youngest child starts in school, launching adult children; onset of caregiving responsibilities; and loss of longestjob. Each of these variables is a time-varying covariate; that

In the WLS, 15 percent of women and 7 percent of men who obtained a degree after age 35 were employed as teachers at or prior to the time they received their degree, and 13 percent of women who received a degree in midlife were in a nursing occupation in 1975.

is, the variable 'turns on" during the quarter of the year that the event took place. Divorce is measured as the quarter when the WLS respondent stopped living with his or her spouse. Youngest child starting school is measured as the quarter when one's youngest child turns age 6. The onset of the "empty nest "phase is indicated by two different variable; one's youngest child turning age 18, and the dates that the respondent transferred money to any child to defray educational expenses (a time-varying covariate). We use two indicators of empty nest because the WLS does not collect data on the timing of each child's departure from the family home. Although the "empty nest7'stage is generally believed to begin when one's youngest child graduates from high school (Rubin 1979), we wanted to include an additional indicator a child's transition into young adulthood. Moreover, the inclusion of both variables allows us to assess the influence of shifts in both time and financial resources.

Caregiving duties are captured with two variables. First, we control whether the respondent was required to provide instrumental assistance to a sick family member or friend for a spell of one month or longer. The WLS assesses whether a respondent has provided care during the last year, and if yes, obtains the start and stop date of the last caregiving spell. For most respondents who provided care, the 'Start date" began far earlier than 1992-93; the longest spells of caregiving began as early as 1958. Secondly, we include a tirne-varying covariate indicating those points in time when a respondent gave money to hislher parents to defray living expenses. Given that caregivers frequently provide instrumental and financial care simultaneously (Hill et al. 1970; Hogan, Eggebeen and Clogg 1993), we believed that this variable might serve as a proxy for caregiving behavior that occurred earlier than the past year. As in the case of launching adult children, the caregiving and intertransfer variables indicate shifts in the time and economic resources available to midlife adults.

The loss of longestjob is a time-varying covariate that turns on if the respondent ended a period of employment which was their longest spell with an employer that was reported between 1975 and 1992193. Involuntary job losses are those longest job exits that occurred due to a business closing/relocation, lay-offs, lost business or farm, or other involuntary loss. If a respondent answered 'ho" to the question "Could you have worked at this job for another 6 months" then that response is also coded as involuntary loss. All other persons who ended their longest job spell are coded as having a voluntary loss of longestjob. Descriptive statistics for independent variables are presented in Appendix A.

3For transitions that are irreversible, an indicator variable is 'turned on" when and if these event occurs, and stays ''on" for the remainder of the analysis. These transitions include youngest child turns age 6, youngest child turns age 18 and job loss. For the remaining transitions, the covariate turns "on" when a spell begins, and is set "off' when a spell ends. Divorce, widowhood, remarriage, and economic transfers to children transitions may have zero, one, or more spells; there can be only one caregiving spell due to data limitations. Economic transfers to parents are 'turned on"at the time of first transfer, but are not set off because the WLS does not obtain information on the date at which these transfers end.

Analytic Method Event history analysis is used to evaluate the effects of covariates on the rate of malung

an educational transition. The basic hazard model is: h,(t) = a,(t) exp(b,x,, + b2xi2 + .. + bnxin + b,nx,m(t) +...+ bp,(t))

where hi(t) is the hazard for individual i at time t, b, to b, are the slopes or regression coefficients associated with fixed covariates xi, b , to b, are the slopes associated with time-varying covariates x,(t), and a,(t) is an unspecified baseline hazard function. A hazard rate is the probability that an event will occur at a particular time to a particular individual, given that the individual is "at risk" at that time.

To identify the baseline hazard appropriate for use in our analysis, we performed our analysis using several different baselines: Exponential, Weibull, Gompertz, quadratic, and cubic. Chi-square tests were used to determine which baseline best fit the data. We report the results based on the best fitting model, although the coefficients of interest (b, to b, and b , to b,) did not change appreciably across models. Because the hazard rate is different for men and women and for the two types of educational transitions, our final analyses use a variety of baseline functions, based on the best fitting model for men and women, for each dependent variable. For men's receipt of a degree and men's enrollment, the Exponential model is used. For women's receipt of a degree, a Weibull function fit the data best, while for women's enrollment the Gompertz function provided the best fit.4 Models predicting college enrollment are presented in Table 2, and models predicting receipt of a college degree at midlife are presented in Table 3. Separate models for men and women are presented, and significant gender differences across the models are noted. We began our analysis by estimating pooled models with all independent variables interacted with gender, yet have presented only the sex-specific models here for ease of interpretation. Full gender-interaction models are available from the authors. FINDINGS

Before reviewing results of the multivariate analyses, it is informative to describe patterns of adult educational transitions in the WLS. Women are substantially more likely than men to enroll in college after age 35, a pattern consistent with prior studies (see Jacobs and Stoner-Eby 1998 for a review). Results in Table 1 show that 19 percent of women and 12 percent of men returned to college at midlife. Similarly, a higher proportion of women than men completed a degree after age 35 (5.5 percent and 2 percent, respectively). Most educational transitions occurred during the 1980s, when respondents were in their 40s. The average dates of both returning to school, and receiving a degree occurred in 1984 for men and 1985 for women.

[Table 1 about here] Linked Lives

Findings from the multivariate analyses will be organized according to the themes of the life course paradigm.

4We use the Continuous Time Models (CTM) software (version 5, Yates 1987) to estimate our models, which allows us to specify the Exponential, Weibull, Gompertz, quadratic, and cubic baseline functions.

[Tables 2 and 3 about here] Given the vast body of research linking women's work pursuits to childrearing

responsibilities, we expected children to be inversely related to both college enrollment and degree receipt. We did not find support for this hypothesis; each additional child increases the risk of women's obtaining a college degree in later life (by 9 percent), but this result is only marginally significant. Number of children did not significantly affect men's transitions.

Spouse's employment and education characteristics significantly influence the risk of educational transitions - confirming our hypothesis that persons with greater family-level economic and social resources will be more likely to return to school. Specifically, women whose husbands have graduate-level education have a significantly higher risk of both enrolling in college and receiving a degree. Interestingly, a more direct indicator of a family's economic resources - the income characteristics of a spouse's job - are only a marginally significant predictor of women's enrollment at midlife. For men, being married to a woman who attended college increases the risk of returning to college in midlife by 53 percent, compared to men whose wives are high school graduates or less. Planful Activity

Our analyses confirm the previously documented finding that higher education at baseline is associated with greater risk of returning to school and earning a degree at midlife. This pattern is particularly strong for women. Women's educational attainment is positively associated with educational transitions, although the relationship shifts for women with graduate education at baseline. Having some graduate education increases the risk of returning to school, relative to high school graduates, but does not increase the risk of obtaining a degree in midlife. Women's mental ability, another indicator of personal resources, increased the risk of both returning to school and receiving a degree; each 1-point increase in IQ score increased the risks by 22 and 40 percent, respectively. The relationship between measured intelligence and midlife educational transitions is substantially stronger for women then men, sex by IQ interaction terms are significantly different at thepc.05 level.

Interestingly, baseline education has a substantially weaker impact on midlife educational attainment among men. Men who attended some college before 1975 had an 82 percent faster return to college than high school graduates, but this is the only significant effect of educational attainment for men. Likewise, men's IQ is unrelated to their returns to college. As we expected, men's educational transitions are more responsive to work conditions; none of the work characteristics is significant for women at thep 5 . 0 5 level. Occupational education had a marginally significant impact on men's educational transitions. As the average educational level of men's occupation rises, their risk of both enrolling in college and receiving a degree increases. Being self-employed reduces men's risk of returning to school by 48 percent, while military service quadruples men's risk of enrollment, and increases men's risk of degree receipt by a factor of 12. This powerful effect is likely due to men's receipt of military benefits, which facilitate the completion of a college education for veterans (Greenberg 1997; Bennett 1996). Likewise, given the place of military service in the life course, these men would not have attended college between the ages 1 8-2 1 as their high school classmates did; rather, they postponed the experience (Elder, Shanahan and Clipp 1994).

Although women's employment experiences are not significant predictors of their educational transitions, their occupational goals are quite strong predictors. Compared to women who (at age 35) said they wanted to remain in their current job, women who sought a new job are 25 percent faster to return to school, and receive their degrees 5 1 percent faster. In contrast, women who did not want to work or who desired to be homemakers in 1975 returned to school (p5.05) and received degrees (not significant) 36 percent slower than those women who wanted to remain in their 1975 jobs. Women who aspired to occupations requiring high levels of education had increased risk of both returning to school and earning a degree.

The occupational aspirations of the WLS men, however, have little effect on midlife educational transitions. Men who sought a different job (at age 35) had a 23 percent higher risk of returning to school than those men who wanted to remain in their 1975 job, but this result is only marginally significant. No other aspiration variables have a significant effect on men's risk of educational transitions. Transitional EventsIuTurning Points"

Both normative and non-normative life events may be turning points if they spark a redirection in the life course. We expected that selected family and career transitions might also be turning points if they 'triggered" a return to school or attainment of a college degree at midlife. We found several such turning points in our analysis - but for women only. Interestingly, markers of turning points in their children's lives, such as starting school or turning age 18, have no significant effect on men's and women's risk of either enrolling in school or earning a degree at midlife. Contrary to our starting hypothesis, transferring money to a child to defray college expenses increases women's risk of college enrollment.

The transitions which have the strongest effects on women's educational transitions are related to a woman's marital history, not her childbearing history. Divorce increases by 50 percent women's risk of both returning to school and receiving a degree. Widowhood substantially increases women's risk of both returning to school and receiving a degree but these results are only marginally significant. Remarriage is unrelated to men's and women's educational transitions.

Providing care to a family member or friend is unrelated to men's and women's adult college attendance, although transferring money to one's parents to defray living expenses substantially increases men's risk of returning to school. None of the variables relating to children or marriage significantly affect men's risk of returning to college or receiving a college degree in midlife. Overall, family transitions are unrelated to men's educational transitions, although marital dissolution is an important turning point in the life course of women.

Finally, we expected that voluntarily leaving one's longest job would be associated with a higher risk of returning to college, for both men and women. We also expected that an involuntary loss of longest job would significantly reduce the risk of college attendance for midlife men, given the burden placed on men to provide economically for their families; seeking a new job might be preferable to returning to school. Instead, we found that job loss is unrelated to men's educational transitions, although loss of longest job is significantly associated with women's enrollment and degree receipt. Women who lost their longest job involuntarily have a 3 1 percent higher risk of returning to school, compared to women who stayed in their longest-

held job, while women who voluntarily left their longest job have a 23 percent increase in the risk of returning to college. DISCUSSION

Taken together, our findings reveal important linkages among family characteristics, human capital investments, workplace experiences and midlife educational transitions. Yet our findings also have broader implications for the study of the life course. First, our findings reveal that life transitions are subject not only to structural influences but also to personal preferences and aspirations. Second, our results suggest that some normative and non-normative family events may also be considered "turning points," which trigger a re-direction in adults'lives. Finally, the study of midlife college attendance highlights gender differences in the shape and content of the life course. In a period of rapid social change, distinguished by shifts in the work and family roles held by women, old norms and cultural expectations about women's life course disappear, and a more fluid life course emerges. Men's life course, still guided largely by expectations that the life course will unfold in an orderly fashion, appears to abide by a much more rigid timetable. Work and Family Influences on Educational Transitions

We expected that family and spouse characteristics would have a much stronger effect on women's midlife college attainment than men's; however, neither sex evidenced the very strong effects we were expecting. Having a spouse with education beyond high school generally increased the hazard of returning to school for both women and men, while there was little effect of spouse's education for either sex on the hazard of obtaining a degree (except if a woman's husband had some graduate school). While our results do not necessarily refute the importance of family economic resources in guiding adult educational transitions, our findings do suggest that other resources, such as a spouse who values education and encourages educational pursuits, may be equally important. Having a highly educated spouse or a spouse holding an occupation with highly educated incumbents does not only indicate higher income, but also implies that one has a social network in which educational pursuits are encouraged.

Consistent with past research (Felmlee 1988; Bradburn et a1 1995), one's own levels of education prior to age 35 had strong effects on later educational attainment, with stronger effects for women than men. Compared to high school graduates, WLS women who completed at least some college are substantially more likely to return to college and to receive a degree at midlife. The one exception is women who attended graduate school prior to 1975; women who attended graduate school prior to 1975 did not differ significantly from high school graduates in their rates of receiving a college degree in midlife, yet they were substantially more likely to return to college. This finding emphasizes the importance of examining both college enrollment and degree attainment at midlife. Women with at least some graduate education might have returned to college for personal enrichment or to keep up in their field, but because they had already attained a high level of education are unlikely to change career paths and obtain another degree.

Work characteristics were among the few significant predictors of men's educational transitions. Men holding jobs which require higher education were more likely to return to school, suggesting that certain occupations might require that workers "keep up" with the changing technology in the workplace. Although we expected that working in declining

industries during the 1970s and 1980s would be associated with an elevated risk of educational transitions, neither working in farming nor manufacturing impacted educational transitions.

Owning one's own business impacted men and women quite differently. Self-employed men had sharply decreased rates of returning to college and earning a degree, while self- employed women had somewhat higher rates of midlife college enrollment. This finding is not surprising. Among self-employed women and men, the gender gap in earnings - even among full- time year-round workers - is much starker than among wage and salary workers (Devine 1994). Women who have their own businesses may seek to expand their educational credentials in order to increase their future earnings. The Role of Aspirations

One important distinction of our paper is the inclusion of direct measures of career aspirations. If, as the life course paradigm suggests, people are planful and make choices in their lives, then including measures of what respondents plan to do in the future may help us to better understand midlife educational transitions. Career aspirations were a strong predictor for women's but not men's educational transitions. Women who wanted a career change at age 35 (in 1975), and who wanted a job requiring higher levels of education, returned to college significantly faster than other women, and also received degrees much faster. These women might believe that the way to achieve their goal of a different job or a more challenging job is to obtain hrther education. At the same time, women who did not want to work in the paid labor force were significantly slower to return to college in midlife. Increased educational attainment would probably have relatively little value for individuals who did not want to work for pay.

For men, however, occupational aspirations had little effect on adult college attendance. In contrast to the measures of current occupational status that had significant effects on midlife returns to school, aspirations had little effect. This might reflect the fact that men and women have very "different life courses" (Kohli 1986). For many men, their occupational aspirations at age 35 simply reflect their current circumstances, and relatively few may hope to redirect their lives, given that they recognize the limits on their own occupational mobility (Jacobs et al. 1991). For women of the WLS cohort, at age 35 they simultaneously found their childrearing obligations diminishing while workforce opportunities expanded for women. At age 35, women may have viewed their hture as filled with possibilities for new occupational and educational pursuits. Turning Points

Next, we turn to the question of whether important life transitions might also serve as turning points in the life course. Two life transitions - youngest child starting school and youngest child turning age 18 - did not impact educational transitions. This might be attributable to measurement issues in the WLS. While having the youngest child turn age 6 is a good proxy for having all children in school, over half of the WLS women's youngest children turned age 6 before the 1975 interview, which decreases the amount of variance in this variable. Using the 1 8th birthday of the youngest child as a proxy for the "empty nest" stage is more problematic. We cannot ascertain whether the child's 18th birthday actually coincides with that child's leaving of the family home, for either college, or the establishment of one's own home - although we partially addressed this issue by including measures of economic transfers for educational expenses.

Yet the weak linkages between children's transitions and parental educational transitions also reflects shifting patterns in the life course of women and children. Nearly all of the children of WLS parents were age 21-35 at the time of the 1992-93 interview. As members of the "Baby Bust" or "Generation X cohort, these children are substantially more likely than young adults of prior generations to live with their parents in young adulthood, or to return to their parents'home following divorce, or difficulties in the labor force (Aquilino 1991, Goldscheider and Goldscheider 1993). Thus, a child's 18th birthday or entrance to college does not necessarily mean that a child has been "launched" successfully into adulthood. Moreover, for the WLS cohort of women, the "empty nest"phase may be less meaningful than it was for earlier cohorts of women. Nearly all (98 percent) of the WLS women had ever worked during their lives, and most worked while they were raising children. In the time period from 1957 to 1992-93, only 12 percent of the WLS women had worked fewer than 10 years. Although women of earlier cohorts may have been reluctant to return to school before their youngest child reached adulthood, for most women of the WLS, children might not have been an impediment to either work or educational pursuits. Interestingly, we have the unexpected finding that transferring money to a child to defray hislher educational expenses increases women's risk of returning to college. Perhaps the children's pursuit of higher education provides encouragement for their mothers to do the same; adult children may socialize their parents and expose them to new opportunities (Alwin 1991). This observation is merely speculative, however, and warrants further investigation.

Marital transitions, in contrast, appear to be turning points in women's lives. Our results show that both widowhood and divorce increase the rate of women's educational transitions. Newly-single women may return to college in order to obtain skills and credentials (Collins 1979) necessary to support themselves financially. Returning to school also may represent an adaptive strategy in the face of adversity, with women channeling their energies into self- improvement and personal growth, following the loss of the spousal role. Most sociological research has focused narrowly on the economic consequences of martial dissolution for women and children (Smock and Holden 1991, McLanahan and Sandefur 1994). Less is known about the behavioral and psychological adaptations made by men and women following marital dissolution. Future research should focus on the variety of behavioral responses women make following marital dissolution; where such behaviors may include returning to school, or making a major change in one's work life.

Although we expected caregiving to reduce women's risk of educational transitions, we found no significant relationship between caregiving and returns to school. We believe this may be attributable to the fact that we cannot ascertain how time-intensive the caregiving activity was; presumably caregiving that required at least ten hours a week may be a hindrance to women's labor force transitions. Ruhrn (1 996) found that married women who provide more than 10 hours of caregiving per week are less likely to be employed than those who provide no care, but there are no differences among those who provide fewer hours of care. Moreover, Pavalko and Artis (1 997) found that caregiving behavior among midlife women was associated with a reduction in the number of hours a woman worked.

Implications for the Life Course Overall, we found that women's college enrollment and degree receipt are tied to a

complex constellation of family events and personal aspirations. For men, educational transitions were guided largely by prior work experiences. These findings are consistent with some core assumptions of life course research. Scholars have suggested that the social meaning of age and time differ for men and women, and that men and women may have different needs for predictability and order in their lives (Hagestad 199 1 ; Hagestad and Neugarten 1985). These differences are linked to the fact that women's lives traditionally have been more firmly bound to the family sphere, which operates on a nonlinear time line, while the traditional lives of men are tightly bound to spheres outside the family, which operate on a linear time line (Maines and Hardesty 1987; Settersten and Lovegreen 1998). The general assumption of life course researchers is that men's lives are more rigid, while women's lives are more flexible.

For current cohorts of midlife men, the status attainment process is one which occurs along a fairly orderly time line. If we believe that men's life course is more "rigid" than that of women's (Settersten and Lovegreen 1998), then it is not surprising that most of men's education is completed in the first 'third" of the life course (Kohli 1986). Men are less likely than women to complete their education later in the life course, and few factors - whether enduring family characteristics or unexpected family transitions - will affect their final level of education; that is, men's early life education shapes the outcomes in later life, rather than life outcomes shaping men's educational attainment. Our data suggest that the timing and content of men's life course are constrained by rigid norms and social institutions. Not surprisingly, the one covariate in our analysis that had a large impact on men's college attendance exemplifies the normative sequencing of the male life course. Men who were in the military (and who presumably enlisted immediately following high school graduation) returned to college at the "appropriate" time in the life course - after they had completed their military obligations.

The story for women is far more complex, at least among the WLS cohort of women. Divorce and widowhood may serve as turning points, raising a woman's propensity to return to college, and setting her on a new life course trajectory (Gotlib & Wheaton 1997). Women's educational transitions also are a response to occupational aspirations. The WLS women may actively seek higher education in their late 30s, 40s and early 50s as a strategy for reaching new goals. The women of this cohort experienced a substantial amount of change in both family and work roles during their early and middle adult years. Such social changes are often accompanied by a re-examination of the content, timing, and sequencing of one's roles. Settersten and Lovegreen (1 998) argue that current cohorts of adult women have 'hew time budgets." Given a reduction in the number of years spent bearing and rearing children, and an increase in the number of 'fiealthyVyears to be experienced by these women in late adulthood (see Watkins, Menken and Bongaarts 1987), women (and men) may have many more years to spend in work and educational roles. As a result, age might be less important in determining one's experiences, and education, work and leisure pursuits might be allocated across the entire life course, rather than being limited to one "third" (Kohli 1988; Riley 1986). Education achieved at midlife and beyond may thus have important ramifications for understanding the well-being of future and current cohorts of adults at midlife and beyond.

REFERENCES Alwin, Duane. 1991. Political Attitudes Over the Life Span: The Bennington Women After 50

Years. Madison, WI: University of Wisconsin Press. Aquilino, William. S. 1996. "The Returning Adult Child and Parental Experience at Midlife.'Q

Pp. 423-458 in The Parental Experience in Midlife, edited by Carol D. Ryff and Marsha M. Seltzer. Chicago: University of Chicago.

Bardwick, Judith M. 1980. "The Seasons of a Woman's Life." Pp. 35- 57 in Women's Lives: New Theory, Research and Policy, edited by Dorothy G. McGuigan. Ann Arbor, MI: University of Michigan, Center for Continuing Education of Women

Baruch, Grace K., Rosalind Barnett, and Caryl Rivers. 1983. Llfeprints: New Patterns of Love and Work for Todayd Women . New York: New American Library.

Baumrind, Diana. 1968. "Authoritarian vs. Authoritative Parental Control."Adolescence:3(1 I): 255-272

Berk, S. 1985. The Gender Factory: The Apportionment of Work in American Households. NY: Plenum Press.

Bennett, Michael J. 1996. When Dreams Came True : the GI Bill and the Making of Modern America. Washington, D.C.: Brassey's.

Bianchi, Suzanne and Daphne Spain. 1996. "Women, Work, and Family in America." Population Bulletin 5 1 : 1-48.

Bourdieu, Pierre and Jean-Claude Passeron. 1977. Reproduction in Education, Society, and Culture. London: Sage.

Bowles, Samuel and Herbert Gintis. 1976. Schooling in Capitalist America. New York: Basic Books.

Bradburn, Ellen M., Phyllis Moen, and Donna Dempster-McClain. 1995. "Women's Return to School Following the Transition to Motherhood." Social Forces 73(4): 15 17-7 1.

Brooks-Gum, Jeanne and B. Kirsh. 1984. "Life Events and the Boundaries of Midlife for Women."Pp. 11-30 in Women in Midlfe, edited by G.K. Baruch and J. Brooks-Gum. New York: Plenum.

Burman, Patrick. 1988. Killing Time, Losing Ground: Experiences of Unemployment. Toronto: Wall and Thompson.

Camarigg, Valerie and Jennifer Glass. 1992. "Gender, Parenthood, and Job-Family Compatibility." American Journal of Sociology 98: 13 1-5 1.

Carr, Deborah. 1998. "The Influence of Career Change on Men's Midlife Mental Health." Paper presented at Population Association of America annual meeting. Chicago.

Cherlin, Andrew. 198 1. Marriage, Divorce and Remarriage. Cambridge, MA: Harvard University Press.

Clausen, John. 1991. American Lives. New York: Free Press. Collins, Randall. 1979. The Credential Society. New York: Academic Press. Coontz, Stephanie. 1992. The Way We Never Were: American Families and the Nostalgia Trap.

New York: Basic Books. Davis, Nancy J. and Larry Bumpass. 1976. "The Continuation of Education After Marriage

Among Women in the United States, 1970." Demography 13(2): 16 1-74.

Desai, Sonalde and Linda Waite. 199 1. "Women's Employment During Pregnancy and After the First Birth: Occupational Characteristics and Work Commitment." American Sociological Review 56:55 1-66.

Devine, Theresa. 1994. "Characteristics of Self-Employed Women in the United States." Monthly Labor Review 1 17: 20-34.

Easterlin, Richard. 1980. Birth and Fortune: The Impact of Numbers on Personal Welfare. New York: Basic Books.

Elder, Glen H., Jr., 1994. "Time, Human Agency, and Social Change: Perspectives on the Life Course." Social Psychology Quarterly 57(1):4- 15.

Elder, Glen H., Jr., and Angela ORand. 1995. "Adult Lives in a Changing Society."Pp. 452-475 in Sociological Perspectives on Social Psychology. Boston: Allyn & Bacon.

Elder, Glen H Jr., Michael J, Shanahan, Elizabeth Colerick Clipp. 1994. "When War Comes to Men's Lives: Life-course Patterns in Family, Work, and Health." Psychology and Aging 9:5-16.

Elman, Cheryl and Angela M. ORand. 1998. "Midlife Work Pathways and Educational Entry." Research on Aging 20(4):475-505.

Farrell, M.P. and S.D. Rosenberg. 1981. Men at Midlife. Boston: Auburn House. Featherman, David and Robert M. Hauser. 1978. Opportunity and Change. New York:

Academic Press. Felmlee, Diane. 1988. "Returning to School and Women's Occupational Attainment." Sociology

of Education 6 1 :29-4 1. Gerson, Kathleen. 1985. Hard Choices: How Women Decide About Work, Career and

Motherhood. Berkeley: University of California Press. Gerson, Kathleen. 1993. No Man's Land: Men's Changing Commitment to Family and Work.

New York: Basic Books. Goldscheider, Francis K. and Calvin Goldscheider. 1993. Leaving Home Before Marriage:

Ethnicity, Familism, and Generational Relations. Madison, WI: University of Wisconsin Press.

Gotlib, Ian H. and Blair Wheaton. 1997. Stress and Adversity Over the Life Course: Trajectories and Turning Points. New York: Cambridge University Press.

Greenberg, Milton. 1997. The GI Bill : The Law That Changed America. New York : Lickle Publishing.

Hagestad, Gunhild. 1991. "Trends and Dilemmas in Life Course Research: An International Perspective."Pp. 23-57 in Theoretical Advances in Life Course Research, edited by W.R. Heinz. Weinheim, Germany: Deutscher Studien Verlag.

Hagestad, Gunhild and Bernice Neugarten. 1985. "Age and the Life Course."Pp. 36-61 in Handbook of Aging and the Social Sciences, 2"d ed., edited by E. Shanas and R. Binstock. New York: Van Nostrand and Reinhold.

Haraven, Tamara K. and Kanji Masaoka. 1988. "Turning Points and Transitions: Perceptions of the Life Course." Journal of Family History 13(3):271-289.

Hauser, Robert M. and John Robert Warren. 1997. 'Socioeconomic Indexes for Occupations: A Review, Update, and Critique." Pp. 177-298 in Sociological Methodology, edited by Adrian E. Raftery. Cambridge: Basil Blackwell.

Hauser, Robert M., Jennifer Sheridan, and John Robert Warren. 1998. 'Socioeconomic Achievements of Siblings in the Life Course: New Findings from the Wisconsin Longitudinal Study." Unpublished manuscript. (Earlier version was presented at the Gerontological Society of America meetings in Cincinnati, OH in November 1997).

Havighurst, R.J. 1948. Developmental Tasks and Education. New York: David McKay. Hill, Reuben. 1970. Family Development in Three Generations. Cambridge, MA: Schenkman. Hogan, Dennis P., David J. Eggebeen, and Clifford C. Clogg. 1993. "The Structure of

Intergenerational Exchanges in American Families." American Journal of Sociology 98(6): 1428-58.

Holden, Karen C. 1988. 'Foverty and Living Arrangements Among Older Women: Are Changes in Economic Well-Being Underestimated?" Journal of Gerontology: Social Sciences 43: S22.

Holmes, J.H. and R.H. Rahe. 1967. "The Social Readjustment Scale." Journal of Psychosomatic Research 1 1 : 2 13-28.

Jacobs, Jerry A. and Scott Stoner-Eby. 1998. "Adult Enrollment and Educational Attainment." The Annals of the American Academy ofPolitica1 and Social Science 559:91-108.

Jacobs, Jerry A., D. Karen, K. McLelland. 1991. "The Dynamics of Young Men's Career Aspirations." Sociological Forum 6: 609-639.

Kasworm, Carol E. 1990. "Adult Undergraduates in Higher Education: A Review of Past Research Perspectives." Review of Educational Research 60(3):346-72.

Kohn, Melvin and Kazimierz M. Slomcyznski. 1993. Social Structure and Self-Direction: A Comparative Analysis of the United States and Poland. Cambridge, MA: Blackwell.

Kohli, Martin 1986. "The World We Forgot: A Historical Review of the Life Course."Pp. 271- 303 in Later Life, edited by Victor Marshall. Beverly Hills, CA: Sage.

Kutscher, R. and H. Fullerton. 1990. "The Aging Labor Force." In The Aging of the American Work Force: Problems, Programs, Policies, edited by R. Montgomery and J. Owen. Detroit: Wayne State University Press.

Lee, G.R. 1992. "Gender Differences in Family Caregiving: A Fact in Search of a Theory." In J.W. Dwyer and R.T. Coward (editors)., Gender, Families and Elder Care. Newbury Park, CA: Sage

Maines, David R. and Monica J. Hardesty. 1987. "Temporality and Gender: Young Adults' Career and Family Plans." Social Forces 6 1 : 102- 120.

McLanahan, Sara and Gary Sandefur. 1994. Growing Up With a Single Parent: What Hurts, What Helps. Cambridge: Harvard University Press.

Mergers and Acquisitions. 1987. "Almanac and Index" 2 l(6). Mason, Karen 0 . and Karen Kuhlthau. 1992. "The Perceived Impact of Child Care Costs on

Women's Labor Supply and Fertility." Demography 29:523-43.

Moen, Phyllis, and Martha Moorehouse. 1983. Overtime Over the Life Cycle: a Test of the Llfe Cycle Squeeze Hypothesis. Research in the Interweave of Social Roles. Greenwich, CT: JAI Press.

Pappas, Gregory, Susan Queen, Wilbur Hadden, and Gail Fisher. 1993. "The Increasing Disparity in Mortality Between Socioeconomic Groups in the United States, 1960 and 1986." New England Journal of Medicine 329(2): 103-109.

Pascarella, Ernest and Patrick T. Terenzini. 1991. How College Affects Students. San Francisco: Jossey-Bass.

Pavalko, Eliza and Julie E. Artis. 1997. "Women3 Caregiving and Pai8d Work: Causal Relationships in Life Midlife." Journal of Gerontology: Social Sciences 52B(4): S 170-9.

Powell, Douglas H. and Paul F. Driscoll. 1979. "Middle-Class Professionals face Unemployment." In Socialization and the Life Cycle, edited by Peter I. Rose. New York: St. Martin's Press.

Presser, Harriet B. 1988. 'Shift Work and Child Care Among Young Dual-Earner American Parents." Journal ofMarriage and the Family 50: 133-48.

Riley, Matilda W. 1986. "Men, Women and the Lengthening of the Life Course.'?Pp. 333-47 in Gender and the Life Course, edited by A. Rossi. New York: Aldine.

Ross, Catherine and Chia-ling Wu. 1995. "The Links Between Education and Health." American Sociological Review 60:7 19-45.

Rubin, Lillian. 1979. Women of a Certain Age. New York: Basic Books. Ruhrn, Christopher J. 1996. "Gender Differences in Employment Behavior During Late Middle-

Age." Journal of Gerontology: Social Sciences 5 1 B: S 1 1 - 17. Salvo, Joseph J. 1984. Lifetime Work Experience and its Effect on Earnings: Retrospective Data

from the 1979 Income Suwey Development. Washington, D.C.: Bureau of the Census. Settersten, Richard A., Jr. and Loren D. Lovegreen. 1998. "Educational Experiences Throughout

Adult Life: New Hopes or No Hopes for Life-Course Flexibility?'Research on Aging 20(4):506-538.

Sewell, William H. and Robert M. Hauser 1975. Education, Occupation and Earnings: Achievement in the Early Career. New York: Academic Press.

Smari, Jakob, Elvar Arason, Hafsteinn Hafsteinsson, and Snorri Ingimarsson. 1997 "Unemployment, Coping and Psychological Distress." Scandinavian Journal of Psychology 38(2) 151-156.

Smock, Pamela and Karen C. Holden. 1991. "The Economic Costs of Marital Dissolution: Why Do Women Bear a Disproportionate Cost?" Annual Review of Sociology 17: 5 1-78.

South, S.J. and G. Spitze. 1994. "Housework in Marital and Nonrnarital Households. American Sociological Review 59:327-47.

Spannard, Jan-Marie A. 1990. "Beyond Intent: Reentering College to Complete the Degree." Review of Educational Research 60(3):309-44.

Stone, Robyn, G.L. Cafferata and J. Sangl. 1987. "Caregivers and the Frail Elderly: A National Profile." The Gerontologist 27: 280-304.

Suitor, J. Jill. 1988. "Husband's Educational Attainment and Support for Wives'Return to School." Gender and Society 2(4):482-95.

Tolbert, C.M. 1982. 'fndustrial Segmentation and Men's Career Mobility. "American Sociological Review 47: 457-477.

Watkins, Susan C., Jane A. Menken and John Bongaarts. 1987. "Demographic Foundations of Family Change." American Sociological Review 52: 346-358.

Williams, David. 1990. "Socioeconomic Differentials in Health: A Review and Redirection." Social Psychology Quarterly 53 :8 1-99.

Table 1. Wisconsin Longitudinal Study Sample Selection

Original 1957 Sample

1992193 Survey Respondents

1975 and 1992193 Survey Respondents

Not in School in 1975

Returned to School in Mid-Life

Returned to School in Mid- Life and Received a Degree

Women Men Experienced Mean Century St. Dev. Experienced Mean Century St. Dev.

Eligible Event Month of Trans. (Months) Eligible Event Month of Trans. (Months) 5325 --- --- --- 4992 --- --- ---

4088 79 1 August 1985 59.06 3565 435 April 1984 62.09 (76.8%) [19.3%] (7 1.4%) [12.2%]

41 15 228 October 1985 54.91 3577 70 March 1984 56.52 (77.3%) [5.5%] (71.7%) [I .9%]

NOTE: Percent of original sample in parentheses. Percent of eligible sample in brackets.

Table 2. Parameter Estimates and Relative Risks of Enrolling in College at Midlife, Women and Men of the Wisconsin Longitudinal Study, 1957-1993.*

Women Men Covanate Beta S.E. RR Beta S.E. RR - - --- Intercept Gamma One

Linked Lives - Family

Number of Children Not Mamed in 1975 Spouse's 1975 Occupational Education Spouse's 1975 Occupational Income Spouse's Education - Attended College Spouse's Education - College Graduate Spouse's Education - Graduate School Spouse in Military in 1975 Spouse Not in Labor Force

Planful Activity - Human Capital

Mental Ability Educational Attainment-Attended College Educational Attainment-College Graduate Educational Attainment-Graduate School 1975 Occupational Education 1975 Occupational Income In Military in 1975 Farming Industry in 1975 Manufacturing Industry in 1975 Self-Employed in 1975

Planful Activity - Aspirations

1975 Aspirations, Occupational Education 0.14 0.04 1.15 0.00 0.05 1.00 a 1975 Aspirations, Occupational Income 0.04 0.04 1.04 -0.10 0.07 0.91 b 1975 Aspirations, Wanted Different Job 0.23 0.09 1.25 0.21 0.11 1.23 1975 Aspirations, Wanted Housework'No Work -0.45 0.20 0.64 -0.19 0.44 0.83 1975 Aspirations, Don't Know Aspirations -0.10 0.15 0.91 0.02 0.21 1.02

Transitional Events

Youngest Child turns Age 6 Youngest Child turns Age 18 Gave Money to Child for Education Divorce Widowhood Remarriage Caretaking Gave Money to Parent for Living Expenses Left Longest Job-Involuntarily Left Longest Job-Voluntarily

Total N 4088 3565 Number of Events 79 1 435 -2 Log Likelihood 5025.23 2973.88

Note: Social background characteristics (parental education, parental income, parental occupation, maternal employment, farm background, family structure, and number of siblings) and holding a teaching or nursing occupation in 1975 are controlled.

* Baseline hazard for women is Gompertz; baseline hazard for men is Exponential. **Could not be estimated due to small number of events. Bold coefficients are twice their standard error; coefficients in italics are 1.5 times their standard error. a Women and men significantly different at .05 level. b Women and men significantly different at .10 level.

Table 3. Parameter Estimates and Relative Risks of Receiving College Degree at Midlife, Women and Men of the Wisconsin Longitudinal Study, 1957-1993.*

Women Men Covariate Beta S.E. RR Beta S.E. RR

Intercept -12.80 1.30 -13.22 3.05 Gamma One 0.39 0.13

Linked Lives - Family

Number of Children Not Manied in 1975 Spouse's 1975 Occupational Education Spouse's 1975 Occupational Income Spouse's Education - Attended College Spouse's Education - College Graduate Spouse's Education - Graduate School Spouse in Military in 1975 Spouse Not in Labor Force

Planful Activity - Human Capital

Mental Ability Educational Attainment-Attended College Educational Attainment-College Graduate Educational Attainment-Graduate School 1975 Occupational Education 1975 Occupational Income In Military in 1975 Farming Industry in 1975 Manufacturing Industry in 1975 Self-Employed in 1975

Planful Activity - Aspirations

1975 Aspirations, Occupational Education 0.11 0.08 1.12 -0.08 0.22 0.93 1975 Aspirations, Occupational Income 0.08 0.08 1.08 0.01 0.27 1.01 1975 Aspirations, Wanted Different Job 0.41 0.20 1.51 0.01 0.44 1.01 1975 Aspirations, Wanted Housework/No Work -0.43 0.43 0.65 * * ** * * 1975 Aspirations, Don't Know Aspirations 0.07 0.32 1.07 0.22 0.60 1.24

Transitional Events

Youngest Child turns Age 6 Youngest Child turns Age I8 Gave Money to Child for Education Divorce Widowhood Remamage Caretaking Gave Money to Parent for Living Expenses Left Longest Job-Involuntarily Left Longest Job-Voluntarily

Total N 41 15 3577 Number of Events 228 70 -2 Log Likelihood 1693.61 576.30

Note: Social background characteristics (parental education, parental income, parental occupation, maternal employment, farm background, family structure, and number of siblings) and holding a teaching or nursing occupation in 1975 are controlled.

* Baseline hazard for women is Weibull; baseline hazard for men is Exponential. **Could not be estimated due to small number of events. Bold coefficients are twice their standard error; coefficients in italics are 1.5 times their standard error. a Women and men significantly different at .05 level. b Women and men significantly different at . I0 level.

Appendix A. Descriptive Statistics for Independent Variables Used in the Analysis: Men and Women of the Wisconsin Longitudinal Study, 1957-1993 *

Women Men

Covariate Mean St. Dev. N Mean St. Dev. N

Social Background

Mother's Educational Attainment

Father's Educational Attainment

Family Income

Head's Occupational Status

Mother Works

Farm Background

Non-Intact Family

Number of Siblings

Linked Lives - Family

Number of Children

Not Married in 1975

Linked Lives - Spouse

Spouse's 1975 Occupational Education** * Spouse's 1975 Occupational Income***

Spouse's Education - High School Graduate**

Spouse's Education - Attended College

Spouse's Education - College Graduate

Spouse's Education - Graduate School

Spouse in Military in 1975

Spouse Not in Labor Force

Planful Activity - Human Capital

Mental Ability Educational Attainment-High School Graduate** Educational Attainment-Attended College Educational Attainment-College Graduate Educational Attainment-Graduate School 1975 Occupational Education* * * 1975 Occupational Income*** In Military in 1975 Farming Industry in 1975 Manufacturing Industry in 1975 Self-Employed in 1975

Planful Activity - Aspirations

1975 Aspirations, Occupational Education* * * 0.44 (0.27) 3046 0.41 (0.29) 3091 1975 Aspirations, Occupational Income*** 0.16 (0.18) 3046 0.40 (0.21) 3091 1975 Aspirations, Wanted Different Job 0.51 (0.50) 4135 0.4 1 (0.49) 3591 1975 Aspirations, Wanted HouseworkPJo Work 0.10 (0.30) 4136 0.02 (0.14) 3594 1975 Aspirations, Don't Know Aspirations 0.13 (0.34) 4135 0.09 (0.28) 3591 (Continued on next page)

Appendix A. Descriptive Statistics for Independent Variables Used in the

Analysis: Men and Women of the Wisconsin Longitudinal Study, 1957-1993 *

Women Men Covariate Mean St. Dev. N Mean St. Dev. N

Transitional Events

Youngest Child tums Age 6

Youngest Chlld tums Age 18

Gave Money to Ch~ld for Educat~on

Mean Number of Transfers

Mean Year of Flrst Transfer

Dlvorce

Mean Number of Dlvorces

Mean Length of Dlvorce Spell (Months)

Mean Date of First D~vorce

Widowed

Mean Number of Wldow Penods

Mean Length of Wldow Spell (Months)

Mean Date of Flrst Widowhood

Remarrled

Mean Number of Mamages (~nc . Flrst)

Mean Length of Marrlages (inc. Flrst) (Months)

Mean Length of Marrlages (2+ Only) (Months)

Mean Date of F~rst Remarr~age

Caretaklng

Mean Date Spell Began

Mean Length of Spell (Months)

Gave Money to Parent for Llv~ng Expenses

Mean Year of Flrst Transfer

Left Longest Job Spell

Percent Never Left Longest Job

Mean Date Left Longest Job Involuntanly

Mean Date Lefi Longest Job Voluntarily

Other Controls

Teachlng Occupation In 1975

Nurslng Occupation in 1975

Aug. 1974

Aug. 1986

0.90

1987

0.25

91.77

Dec. 1978

0.06

60.49

Oct. 1983

1.17

300.84

140.19

Mar. 1977

Nov. 1989

36.50

1988

0.45

Oct. 1984

Dec. 1985

0.07

0.04

Sept. 1976

Sept. 1988

1.01

1988

0.29

69.32

Nov. 1979

0.02

36.23

June 1984

1.21

274.79

129.91

Oct. 1978

Dec. 1988

46.93

1989

0.42

Sept. 1984

May 1985

0.05

0.00

*Base sample is WLS Graduates who answered both 1975 and 1992193 surveys and were not in school in 1975.

For women, N=4 136; for men, N=3594.

**Reference variable.

***Actual percentages reported here are transformed into started logits in the following models--see text.