Population Research and Policy Review

, Volume 30, Issue 1, pp 59–79

Are Generation X’ers Different than Late Boomers? Family and Earnings Trends among Recent Cohorts of Women at Young Adulthood

Authors

    • Office of Retirement PolicyU.S. Social Security Administration
  • Howard M. Iams
    • Office of Research Evaluation and StatisticsU.S. Social Security Administration
Article

DOI: 10.1007/s11113-010-9178-x

Cite this article as:
Tamborini, C.R. & Iams, H.M. Popul Res Policy Rev (2011) 30: 59. doi:10.1007/s11113-010-9178-x

Abstract

This article examines emerging trends in childbearing, marital status, and earnings for U.S. women over young adulthood across recent birth cohorts spanning the late baby boom and Generation X. We use a unique dataset that matches the 1990, 1996, and 2004 fertility and marital history modules of the Survey of Income and Program Participation with Social Security Administration longitudinal earnings records derived from survey respondents’ own tax records. While there have been some cohort-level changes, we find little empirical evidence of large-scale shifts in the family and earnings histories of young adult women born toward the end of Generation X, particularly college graduates, relative to their late baby-boom counterparts at the same stage of the life course. The broader implications of our findings and directions for further research are discussed.

Keywords

Generation XWomenFamily and earningsYoung adulthoodSocial Security longitudinal earnings data

Introduction

The last half of the twentieth century witnessed dramatic changes in women’s roles in the workplace and family in the United States. Such changes have been marked, on the one hand, by rising education levels and labor force participation (Blau et al. 2006; U.S. Census Bureau 2008, Table 568), and on the other hand, by declining fertility rates, rising divorce rates in the 1960s and 1970s that have since leveled off, and increasing incidence of delayed or forgone marriage (Goldstein and Kenny 2001; Tamborini 2007; Waite 1995).

While changes in women’s family and labor force activities in the baby-boom generation have been relatively well documented (Butrica et al. 2003; Easterlin et al. 1993), less attention has focused on post-boomer cohorts such as Generation X, primarily because of their age. In recent years, however, discussion about emergent trends has become more prevalent as recent cohorts begin to enter prime childbearing and earning years. In the popular media, much of the attention has centered on professional women who are characterized as increasingly “opting out” of the paid labor force to focus on childrearing or other family responsibilities (Belkin 2003; Brenner 2001; Gross 2005; Story 2005). Kuperberg and Stone (2008) provide valuable recent analysis of this narrative in the media.

Empirical research on women’s family and work patterns in post-boomer cohorts has been more varied. A prominent example of a cohort analysis supporting a change thesis is Vere (2007), which argues that unlike their counterparts of the baby-boom generation who aspired to “have it all,” meaning balancing a career with full family responsibilities; women of more recent cohorts have different attitudes toward family and work. Using various years of the CPS and Natality Detail Files, the author provides evidence that Generation X, particularly college-educated women born in the mid to late 1970s, are having more children and working fewer cumulative hours at the start of their careers than women born 10–15 years earlier.1

Recent economic research also has suggested modest change in women’s paid work. Following sharp rises since the 1970s, women’s labor force participation growth rate has subsided since the late 1990s and early 2000s (Blau and Kahn 2007; Hotchkiss 2006).2 Bradbury and Katz’s (2005) analysis of CPS data finds reduced labor force participation for prime-age women (age 25–54) with a college degree between 1995 and 2004 (from 84.7 to 81.8%), most prominently among married mothers with a young child or a high-earning husband.

However, the extent to which women’s family and labor market behavior has changed among recent cohorts remains ambiguous. Using Census Bureau and American Community Survey (ACS) data, Percheski (2008) finds stable employment and fertility patterns among college-educated women from Generation X cohorts (1966–1975). DiNatale and Boraas (2002) show that Generation X women have more attachment to the labor force between age 25 and 34 years old compared to women born in the 1940s. Furthermore, observed changes in women’s labor market participation may be linked to other factors, such as a weak labor market conditions or changing demographics (i.e., rising share of Hispanics), rather than an increasing prevalence of women that voluntarily choose to forgo paid work (Boushey 2008).

Against this backdrop, this article examines the family and earnings histories of young adult women from three birth cohorts, from the late baby boom through late Generation X.3 We draw data from a restricted-use dataset that matches wave two of the 1990, 1996, and 2004 fertility and marital history topical modules of the U.S. Census Bureau’s Survey of Income and Program Participation (SIPP) to Social Security Administration (SSA) longitudinal earnings records. The SIPP data provide women’s retrospective fertility and marital histories, while the linked administrative records provide survey respondents’ longitudinal earnings histories based on tax records submitted for all jobs to the Internal Revenue Service (IRS). The matched administrative data are especially well suited to this study as they permit examination of women’s earnings characteristics over young adulthood, at ages 23–28 and a narrower range age 27–28.

Using a cohort analysis approach, we consider whether Generation X women, particularly college graduates, are following different family and labor patterns at young adulthood than their predecessors from the late baby boom. The main analytic sample consists of women aged 27 and 28 in the aforementioned SIPP panels. The selection of this age range reflects a research strategy to compare the most recent birth cohort of Generation X given available data (age 27 and 28 in 2004 SIPP) with earlier Generation X (age 27 and 28 in the 1996 SIPP) and late baby-boom cohort (age 27 and 28 in the 1990 SIPP) at the same stage of the life course. As with Vere (2007), we choose ages 27 and 28, rather than a younger age, in order to estimate family and economic outcomes after most members of the cohort have completed schooling. Of course, women’s subsequent experience may change before they reach middle age, but these young women suggest whether changes in life history patterns are occurring among recent cohorts of Generation X. Existing studies, particularly those focused on labor force participation, tend to lump large age groups of women together (e.g., Bradbury and Katz 2005), which while useful, may conceal cohort differences in life events over specific stages of the life course; in our case, young adulthood.4

The next section elaborates on the study’s data and analytic approach, followed by the presentation of the results. The final section considers implications of our findings and directions for future research.

Data and Methods

The SIPP is a nationally representative household panel survey of the civilian noninstitutionalized U.S. population, with a sample size ranging from approximately 14,000 to 46,500 households depending on the panel. The survey is designed to measure the economic situation of persons and households in the United States and to provide a tool for managing and evaluating government transfer and service programs. Interviews (also referred to as waves) are conducted every 4 months for approximately 30–48 months, depending on the panel. The survey can be used as a longitudinal or cross-sectional data source.

The data source for family trends analyzed in this study is the fertility and marital history topical modules in wave 2 of the 1990, 1996, and 2004 SIPP panels. In addition to the core questionnaire, the SIPP contains various topical modules assigned to different waves within each panel. The fertility and marital history modules provide complete retrospective fertility and marital histories for every person in the household aged 15 or older from a nationally representative sample. The modules were administered in the second interview in all three of the SIPP panels used in this study, and use essentially the same questionnaire.

Matched Administrative Earnings Data

For labor trends, our data source is Social Security longitudinal earnings records matched to the aforementioned survey records. With the assistance of the U.S. Census Bureau, SIPP respondents who provided permission were matched to their Social Security earnings records. In this research, we use SSA’s Detailed Earnings Records (DER), which are based on SSA’s Master Earnings File, a primary repository of earnings data for the U.S. population sent to SSA by employers. Appendix A discusses the administrative data in greater detail.

In essence, a matched SIPP-SSA dataset combines the accuracy and longitudinal coverage of SSA’s earnings records with the rich demographic and socioeconomic information available in the SIPP (Davies and Fisher 2009). A particular advantage for this study is the ability to trace the earnings streams of real birth cohorts over young adulthood. This method is often superior to synthetic cohorts, that is, a cohort created by summing the experience of different age groups across or within a cross-sectional survey. Because individuals move in and out of the labor force at different points of their life cycle, the attribute of the cohort does not necessarily cumulate over time. Members of a cohort in one cross-sectional sample may be different than the same cohort represented in a later snap-shot.

Linked SSA earnings data also provide clear advantages over economic information reported in the SIPP. First, SSA earnings offer superior life history earnings information. In this study’s framework, SSA earnings provide 6 years of respondents’ earnings, from ages 23–28, whereas SIPP earnings would only permit around 2.5–4 years. Second, the administrative data are subject to significantly less sample attrition than multi-year earnings obtained by running through a SIPP panel. The SIPP data suffer from substantial sample attrition after the second interview containing the fertility and marital histories. Toward the final waves, around 35–40% of eligible households have fallen out of the survey. Third, because SSA earnings data come from respondents’ tax records sent by employers, they contain little measurement error relative to survey-reported earnings, which are subject to misreporting (Pedace and Bates 2000).

The matched SIPP-SSA data are not without its drawbacks. Administrative earnings records omit earnings in the underground economy. More importantly, not all SIPP respondents are able to be matched to their SSA earnings records. Some SIPP respondents provide inaccurate matching information (e.g. address), which preclude the efforts of the Census Bureau to link survey fields with SSA’s earnings records; others may choose to opt out of the matching program. Somewhat analogous to survey response rates, there is the potential for a match selectivity problem. However, the final sample’s match rates are high: 92% for the 1990 SIPP, 82% for the 1996 SIPP, and 77% for the 2004 SIPP. A comparison of the SIPP (wave 2) and matched SIPP-SSA samples, which is reported in a subsequent section, shows strong similarities between the samples. Also noteworthy, a recent study by Mathematica Policy Research on potential bias in the matched samples for the 1996 and 2001 SIPP panels concluded that the bias was minimal (Czajka et al. 2008).

Analytic Approach

Our main analytic sample consists of women aged 27–28 in the three SIPP panels. The selection of these women allows a cohort analysis of family and work trends in the most recent cohort of Generation X that available SIPP data provide with earlier cohorts at the same stage of the life course. In the 1990 SIPP panel, women age 27 and 28 were born (as reported in survey) between 1961 and 1963 (referred here as late baby boomers); in the 1996 panel, they were born between 1967 and 1969 (referred here as early Generation X’ers); and in the 2004 panel, they were born between 1975 and 1977 (referred here as late Generation X’ers).5 The corresponding sample sizes are 972 late baby boomers, 1,273 early Generation X’ers, and 1,349 late Generation X’ers.

Our analysis first assesses cohort differences in women’s family characteristics at age 27 and 28 and includes childbearing and marriage outcomes. We calculate per capita births per cohort where the numerator is the total reported number of births among women age 27 and 28 up to the respective survey interview, and the denominator is the total number of women age 27 and 28 in the survey. We also examine how age-specific marriage outcomes, including median age of first marriage, vary across the cohorts.6

The second focus of the analysis is women’s paid work histories over young adulthood using the matched SIPP-SSA data. We harmonize women’s earnings histories in each cohort to reflect the same period of young adulthood, from ages 23 through 28. The analysis starts at age 23, rather than 18, because women are likely to have graduated and entered the workforce by this time.7 To gain more distance from traditional schooling years in higher education, we also track earnings at ages 27–28.

Because administrative data do not include a direct measure of labor supply (i.e., hours worked), we measure paid work involvement over young adulthood by calculating women’s total number of years from ages 23 through 28 with non-trivial earnings. In an effort to not treat minor earnings as an indicator of paid work, we define non-trivial as having earned full Social Security coverage (i.e., four quarters) for that year. Thus, in 2005, non-trivial earnings denote annual earnings equal to at least $3,680, the amount needed for full Social Security coverage in that year. When annual earnings meet this threshold, a binary variable for that year is coded 1; earnings below this threshold are coded 0. For earlier years, annual earnings are adjusted using the same wage-indexing method employed by Social Security. From this information, we create a summary measure that varies from 6 years (most engagement with paid work) to 0 years (least engagement with paid work).

We also measure earnings levels over young adulthood. For each of the selected cohorts, we calculate women’s average annual real earnings for years worked, and total cumulative real earnings, from ages 23 through 28 (and ages 27 through 28). All earnings are adjusted to 2005 dollars using the Consumer Price Index (CPI-W) series method.

We employ descriptive and regression techniques to examine cohort differences in the life history measures described above. Descriptive results are broken out by educational subgroups given the strong role of education in influencing women’s family and earnings characteristics, especially at young adulthood (Blossfeld and Huinink 1991; Vere 2007). To approximate differences between the true population and sample estimates, standard errors were computed (available upon request) using SUDAAN 9.0, a specialized statistical software package designed to produce unbiased variance estimates from complex sampling designs such as the SIPP (RTI 2005).

We use logistic and ordinary least squares (OLS) regression analysis to examine cohort differences in women’s family and labor experiences over young adulthood, while controlling for important factors, such as education, fertility, and race/ethnicity. Because our regression analyses intend to disentangle associations between cohort and family and paid work outcomes independent of a series of control variables, rather than to evaluate casual relationships, they can be viewed as similar to what Ginther and Pollak (2004) call “descriptive regressions.”

Comparison of Full- and Matched-SIPP Samples by Various Characteristics

Before turning to the study’s main results, Table 1 reports our SIPP (wave 2) and matched SIPP-SSA samples of women aged 27 and 28 across several key characteristics. The usefulness of these estimates is twofold. First, they highlight the representation of the SIPP sample in the matched SIPP-SSA datasets for each cohort. As can be observed, the characteristics of the matched sample are quite similar to the full sample across cohorts. Across most categories, differences do not exceed 1 percentage point and are not statistically different. The largest observed differences are by education and race/ethnicity in the 2004 panel, with the matched sample having slightly less high school drop-outs and Hispanics. This is largely due to the fact that matched survey respondents are slightly more likely to be native-born citizens (around 2 percentage points).
Table 1

Distribution of SIPP and matched SIPP-SSA samples (wave 2), by selected characteristics: women aged 27 and 28

Variable

Late baby boom

Early Gen. X

Late Gen. X

SIPP 1990

Matched 1990

SIPP 1996

Matched 1996

SIPP 2004

Matched 2004

Age

 27

47.4

47.3

47.6

47.1

51.4

51.3

 28

52.6

52.8

52.5

52.9

48.6

48.7

Birth year

 1961

20.4

21.0

    

 1962

50.0

49.3

    

 1963

29.6

29.7

    

 1967

  

17.0

17.2

  

 1968

  

48.6

49.3

  

 1969

  

34.4

33.5

  

 1975

    

23.0

23.2

 1976

    

48.8

49.4

 1977

    

28.2

27.3

Education

 H.S. drop

12.0

11.2

12.3

11.4

10.3

7.5*

 H.S. grad.

34.5

34.3

29.1

29.2

21.1

20.1

 Some Col.

28.3

28.7

33.3

34.2

35.8

37.5

 BA

25.2

25.8

25.3

25.2

32.9

35.0

Race/ethnicity

 White

76.7

78.2

68.6

71.2

60.3

64.6+

 Black

11.3

11.0

14.2

13.3

12.4

11.8

 Hispanic

8.3

7.5

13.3

12.0

19.6

16.1+

 Othera

3.7

3.4

3.9

3.5

7.7

7.5

N

972

875

1,273

1,045

1,349

1,051

Notes: Data are weighted with survey-specific weights. We test for statistical difference between the SIPP and matched SIPP-SSA sample for the same survey year, + p < 0.10, * p < 0.05. Some Col. some college; BA bachelor’s degree or higher

aIncludes American Indian, Aleut or Eskimo, Asian or Pacific Islander

Second, Table 1 reveals important changes in the population demographics across the cohorts. Late Generation X women (2004 SIPP) have sharply higher proportions of college graduates by age 27 and 28, and correspondingly, lower proportions of high school graduates and drop-outs, than the late boomers (1990 SIPP). Furthermore, our analysis reveals a substantial rise in advanced degree holders (Master’s, Professional school, and Doctorate degrees), 7.3% in late Generation X compared to 4.5% in early Generation X and 4.1% in late baby boom (results not shown in table). Generation X women, particularly members of the more recent cohort, are also more racially/ethnically diverse, namely Hispanic. Such compositional changes, in turn, may underlay cohort differences in women’s family and earnings experiences over young adulthood.

Results

Childbearing and Marriage at Young Adulthood: Late Boomers and Generation X

Fertility outcomes shed light on how family patterns might be changing among recent cohorts of women at young adulthood. As Fig. 1 shows, per capita births for 27 and 28 year old women remained stable between the late baby boom (1.16) and Generation X cohorts (1.19, 1.17). Table 2, which documents the distribution of births among these women, is consistent with this picture.
https://static-content.springer.com/image/art%3A10.1007%2Fs11113-010-9178-x/MediaObjects/11113_2010_9178_Fig1_HTML.gif
Fig. 1

Per capita births of women at ages 27 and 28, by birth cohort and educational status

Source: Authors’ calculations using the 1990, 1996, and 2004 SIPP fertility history module (wave 2). Notes: Data are weighted with survey-specific weights. Estimates with superscripts differ significantly from the comparable estimate for the late baby boom, + p < 0.10, * p < 0.05 level, ** p < 0.01 (two-tailed tests)

Table 2

Motherhood and marriage at young adulthood, by cohort and education: women aged 27 and 28

 

All

BA or higher

No college degree

L. baby boom

Early Gen. X

Late Gen. X

L. baby boom

Early Gen. X

Late Gen. X

L. baby boom

Early Gen. X

Late Gen. X

Number of births (in %)

 0

37.3

39.1

41.0

64.0

74.0*

69.7

28.3

27.3

27.0

 1

26.9

22.7*

22.5*

22.4

13.9*

19.2

28.4

25.7

24.1*

 2 or more

35.8

38.2

36.5

13.6

12.2

11.1

43.5

47.0

48.9*

 Totalsa

100

100

100

100

100

100

100

100

100

Marital status (in %)

 Married

64.8

59.1*

58.1**

65.7

56.3+

57.7+

64.5

60.1+

58.2*

 Never mar.

27.3

31.6*

37.8**

33.0

41.1+

40.9+

25.4

28.4

36.2**

 Divorced

7.6

9.2

4.0**

1.3

2.7

1.2

9.7

11.5

5.3**

 Widowed

0.3

0

0.2

0

0

0.2

0.4

0

0.3

 Totalsa

100

100

100

100

100

100

100

100

100

Median age of 1st marriage

 Overall

23.5

24.3

25.6**

25.5

27.0*

26.6*

22.5

23.4*

24.8**

 Ever-mar.

21.9

22.2

23.2**

23.8

24.5*

24.5*

20.8

21.4+

22.3**

Source: Authors’ calculations using data from the 1990, 1996, and 2004 SIPP (wave 2 modules)

Notes: Data are weighted with survey-specific weights. Estimates with superscripts differ significantly (two-tailed) from the comparable estimate in the late baby boom at + p < 0.10, * p < 0.05 level, ** p < 0.01

aMay not equal exactly 100 due to rounding

Among college-educated women, per capita births at ages 27 and 28 have changed little across the cohorts. In fact, Fig. 1 indicates a decline in per capita births among college graduates between the late baby boom (0.53) and early Generation X (0.41). Table 2 reveals that this decline has been driven by higher proportions of childless college-educated women in the early Generation X cohort (74%) compared with the late baby boom (64%). At the same time, we note a slight uptick in childbearing among college graduates between early and late Generation X, but such differences are not statistically significant. Sensitivity analysis removing women with graduate degrees did not change the substantive results. Further research on this point, perhaps when late Generation X’ers have reached 40, would help clarify possible childbearing differences within different waves of Generation X.8

Among women without a college degree, Fig. 1 shows increasing per capita births at ages 27 and 28 in late (1.54) and early (1.46) Generation X relative to late boomers (1.37). This increase, Table 2 shows, can be attributed to a rising share of non-college graduates with two or more children. Higher order births by age 27 and 28 is likely associated with changing population demographics, namely the increase in Hispanics and immigrants among Generation X, rather than behavioral changes. Kreider and Elliott (2009, Table 2) provide recent evidence of continuing higher fertility among Hispanic and foreign-born women in the U.S.

Marriage is another arena to assess the family choices of Generation X women by their late 20s. On the one hand, the data (Table 2, bottom rows) reveal continuing prevalence of marriage at young adulthood, as a clear majority in each cohort had married by age 27 and 28. On the other hand, the data indicate a sharp rise in never-married young adult women, from 27.3% in the late baby boom to 31.6% in early Generation X, up to 37.8% in late Generation X. This trend holds for college and non-college graduates.9 Marriage rates at ages 27 and 28 are substantially lower for college graduates; however, this differential is sensitive to the age range under analysis. Goldstein and Kenney (2001, pp. 513–514), for example, show that the marriage gap between college and non-college graduates among more recent cohorts narrows substantially by age 30 and is likely to converge by older ages.

The data also reveal lower proportions of divorced young adult women in late Generation X, namely among non-college graduates. This outcome, however, should not be interpreted as evidence of less divorce given that late Generation women have been married for fewer years, and less often, by ages 27 and 28 than earlier cohorts. The median age of first marriage increased from 23.5 years among late boomers to 25.6 years among late Generation X’ers. Not surprisingly, the median age of first marriage varies sharply by educational subgroup, with the typical female college graduate marrying at least two years later than non-college graduates.

Paid Work Involvement and Earnings over Young Adulthood: Late Boomers and Generation X

The preceding analysis focused on outcomes reflecting women’s family decisions at young adulthood; this section turns to outcomes reflecting women’s labor market decisions. Like the previous section, our sample is restricted to women aged 27 and 28 from wave two of the three SIPP panels. However, only respondents whose SSA earnings records have been successfully linked with their survey fields are examined.

Table 3 displays total years with non-trivial earnings at ages 23 through 28, and 27 through 28 to gain more distance from traditional college years. The columns differentiate by birth cohort and educational attainment and the rows report the distribution, mean and median. As previously noted, to obtain non-trivial earnings in a given year, annual earnings must equal the amount needed for four quarters of Social Security coverage in that calendar year.
Table 3

Years of non-trivial earnings over young adulthood, women, by cohort, educational status, and two age ranges

 

All

BA or higher

No college degree

L. baby boom

Early Gen. X

Late Gen. X

L. baby boom

Early Gen. X

Late Gen. X

L. baby boom

Early Gen. X

Late Gen. X

Number of years from ages 23 to 28 (max. 6)

 0

10.1%

11.4%

8.2%

0.8

3.5*

2.7+

13.4

14.0

11.2

 1

5.2

3.9

3.6

2.1

1.6

1.8

6.3

4.7

4.5

 2

5.9

5.8

5.2

2.3

2.1

1.9

7.2

7.0

7.0

 3

8.1

7.8

6.1

6.1

4.7

5.8

8.8

8.8

6.2

 4

9.0

7.7

8.5

7.8

5.0

7.4

9.4

8.6

9.1

 5

9.9

12.2

10.6

11.5

10.6

9.7

9.3

12.7*

11.0

 6

51.8

51.3

57.9*

69.4

72.5

70.7

45.6

44.2

51.1+

 Totalsa

100

100

100

100

100

100

100

100

100

 Mean

4.4

4.4

4.7**

5.3

5.3

5.3

4.1

4.1

4.4*

 Median

6

6

6

6

6

6

5

5

6

Number of years from ages 27 to 28 (max. 2)

 0

19.0%

21.5%

17.2%

6.0

7.0

7.9

23.5

26.4

22.2

 1

13.0

9.1*

10.2

11.0

4.9+

6.8

13.7

10.5

12.0

 2

68.0

69.4

72.7*

83.0

88.1

85.4

62.8

63.1

65.9

 Totalsa

100

100

100

100

100

100

100

100

100

 Mean

1.5

1.5

1.6+

1.8

1.8

1.8

1.4

1.4

1.4

 Median

2

2

2

2

2

2

2

2

2

Source: Authors’ calculations using SSA earnings records matched to the 1990, 1996, and 2004 SIPP (wave 2)

Notes: Data are weighted with survey-specific weights. Estimates with superscripts differ significantly (two-tailed) from the comparable estimate in the late baby boom, + p < 0.10, * p < 0.05 level, ** p < 0.01. Non-trivial annual earnings are defined as having at least four quarters of Social Security coverage in that year

aMay not equal exactly 100 due to rounding

Late Generation X women (1975–1977) had significantly higher mean total years of non-trivial earnings from ages 23 through 28, marked by greater proportions with non-trivial earnings in all 6 years (58% compared with 51 and 52% for earlier cohorts). Interestingly, there are no cohort differences among college graduates in mean total years with non-trivial earnings over young adulthood. One observable change is an increase in the percentage of college graduates with zero years of non-trivial earnings from age 23 to 28, from 0.8% in the late baby boom to 3.5% in early Generation X, to 2.7% in late Generation X. This pattern attenuates at ages 27–28. Removing women with graduate degrees diminishes the observed pattern, but it does not entirely take it away. While suggesting a slight rise in female college graduates with little engagement in paid work over young adulthood, the trend is small compared to the overall pattern. Moreover, more women still attending graduate school (e.g., PhD training) by age 28 in recent cohorts could contribute to the observed increase.

Looking at women without a college degree, we observe greater paid work activity from age 23 to 28 in the late Generation X cohort. This increase is marked by a rising share with non-trivial earnings in all 6 years of young adulthood, from approximately 46% in the late baby boom to 51% in late Generation X (significant at the 0.10 level).

Earnings levels are another useful gauge of women’s paid work behavior across recent cohorts. Table 4 reports women’s average annual real earnings for years worked over young adulthood by birth cohort and educational subgroup. To focus more clearly on earnings levels rather than the presence of paid work, we exclude cases with trivial/zero earnings over the corresponding age range, as well as individual years with trivial (or zero) earnings.
Table 4

Women’s average annual real earnings (2005 dollars) for years worked over young adulthood, by birth cohort, educational status, and two age ranges

 

L. baby boom

Early Gen. X

Late Gen. X

From ages 23 to 28a

 All

$20,952

20,463

23,680**

 Bachelor’s degreeb

$28,817

27,498

30,938

 No college degree

$17,820

17,808

19,410*

From ages 27 to 28c

 All

$23,613

24,044

27,580*

 Bachelor’s degree

$33,233

33,905

37,229

 No college degree

$19,499

19,914

21,444

Source: Authors’ calculations using SSA administrative earnings records matched to the 1990, 1996, and 2004 SIPP (wave 2)

Notes: Data are weighted with survey-specific weights. Estimates with superscripts differ significantly (two-tailed) from the comparable estimate in the late baby boom, * p < 0.05 level, ** p < 0.01

aWomen with zero years of non-trivial earnings between ages 23 and 28 excluded

bBachelor’s degree includes women with post-graduate education (sensitivity analysis excluding women with graduate degrees show similar results)

cWomen with zero years of non-trivial earnings between ages 27 and 28 excluded

Women’s average annual real earnings for years worked from age 23 to 28 are higher (at 0.01 level) in the late Generation X cohort ($23,680) compared to the late baby boom ($20,952). Among college graduates, cohort differences are not significant. Sensitivity analysis excluding women with graduate degrees (results not shown) show consistent patterns. Also interesting, late Generation X women without a college degree modestly increased their average earnings for years worked between ages 23 and 28. This increase is likely driven, in part, by greater percentages of women with “some college” in the non-college graduate category, as well as lower percentages of high school drop-outs.

Figure 2 displays the median (and 25th and 75th percentiles) cumulative real earnings among women with non-trivial earnings at ages 27 through 28, by cohort. Overall, Generation X women in the labor force had significantly higher median cumulative earnings over ages 27 and 28, as well as the bottom and top quartile, than their late boomer counterparts. For college graduates, cumulative earnings over age 27 and 28 were flat across the cohorts, with the highest median recorded for late Generation X ($70,092). Non-college graduates had higher cumulative earnings at the top and bottom quartile in late Generation X, which again may be partly associated with growing proportions of some college in this category.10 Analysis of women’s cumulative earnings from ages 23 to 28 (results not shown), and cumulative earnings adjusted for wage growth, are generally consistent with the patterns presented in Fig. 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs11113-010-9178-x/MediaObjects/11113_2010_9178_Fig2_HTML.gif
Fig. 2

Cumulative real earnings of women ages 27 through 28, by cohort and education: median, 25th and 75th percentile

Source: Authors’ calculations using SSA administrative earnings records matched to the 1990, 1996, and 2004 SIPP (wave 2). All estimates are in 2005 dollars. Notes: Data are weighted with survey-specific weights. Respondents with non-positive or trivial annual earnings at ages 27 and 28 are excluded. Estimates with superscripts differ significantly from the comparable estimate for the late baby boom cohort, + p < 0.10, * p < 0.05 level, ** p < 0.01 (two-tailed tests)

Multivariate Regression Analysis: Family and Labor Outcomes

To test if our descriptive results hold when controlling for many of the variables discussed above, we employ logistic and OLS regression analysis. The regressions pool the three cohort samples used in the descriptive analysis and include dummy variables for each cohort, which also reflect the survey year. Given recent focus on college graduates from Generation X (Percheski 2008; Vere 2007), we include an interaction term between cohort and education. Readers may refer back to Table 1 for descriptive statistics of the separate-year samples.

Table 5 presents results for several logistic regressions estimating women’s family characteristics at age 27 and 28. Motherhood (odds of childless) has no significant association with cohort at age 27 and 28 (left column), independent of other control variables. Being without a high school diploma, relative to high school graduates, reduce the odds of being childless, while having a bachelor’s degree increases the odds. A significant interaction reveals that early Generation X college graduates had higher odds of being childless at age 27 and 28 than their counterparts from the other cohorts. This is consistent with Table 2’s descriptive results.
Table 5

Logistic regression estimates: motherhood and marriage at young adulthood, women ages 27 and 28, selected cohorts (pooled SIPP sample)

Variable

Motherhood

Marriage

No child

Never married

Currently married

Early Gen. X (ref. = L. baby boom)

−0.091 (.118)

0.114 (.125)

−0.186 (.114)

Late Gen. X

−0.112 (.118)

0.507** (.123)

−0.272* (.114)

Less HS (ref. = HS Grad.)

−0.954** (.264)

0.020 (.225)

−0.018 (.204)

Bachelor’s degree

1.417** (.168)

0.564** (.174)

−0.064 (.167)

Less HS × Early Gen. X

−0.003 (.365)

0.224 (.294)

−0.232 (.270)

Less HS × Late Gen. X

−0.021 (.383)

0.264 (.296)

−0.198 (.278)

BA × Early Gen. X

0.438* (.226)

0.092 (.226)

−0.114 (.217)

BA × Late Gen. X

0.230 (.214)

−0.167 (.214)

−0.108 (.208)

Black (ref. = non-Hisp. White)

−0.513** (.119)

1.534** (.107)

−1.320** (.108)

Hispanic

−0.565** (.128)

0.078 (.115)

0.106 (.109)

Other

−0.063 (.164)

0.134 (.159)

0.023 (.154)

Intercept

−0.667** (.090)

−1.332** (.093)

0.762** (.084)

Likelihood ratio

644.82**

276.53**

192.10**

Pseudo R2

0.164

0.074

0.052

Sample size

3,594

3,594

Source: Authors’ calculations using the 1990, 1996, and 2004 SIPP (wave 2 modules)

Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01. HS Grad. high school graduate; Less HS no high school diploma; BA at least bachelor’s degree

In terms of marriage, the results are inconsistent with the notion that women of more recent cohorts are ‘more committed’ to family, at least by 27 and 28 years old. Late Generation X women relative to late boomers had higher odds of being never married at ages 27 and 28, and reduced odds of being currently married, holding all other variables constant. The insignificant bachelor’s degree-cohort interaction suggests a similar effect of college degree across cohorts.

Table 6 evaluates the relationship between cohort and women’s paid work histories over young adulthood using OLS. We start by estimating total years of non-trivial earnings from ages 23 to 28 and 27 to 28. Results show no significant cohort shifts away from paid work. In fact, late Generation X women relative to late boomers had significantly higher years of paid work activity from ages 23 through 28. Meanwhile, education has the expected effects; a college degree significantly increases, and less than high school significantly decreases, the number of years with earnings. The cohort–college graduate interactions yield insignificant coefficients, suggesting a relatively stable effect of college degree on the dependent variable across cohorts. As expected, number of children is negatively associated with women’s paid work involvement over young adulthood.
Table 6

OLS regression estimates: women’s paid work involvement and earnings over young adulthood, two age ranges, selected cohorts (pooled matched SIPP-SSA sample)

 

Total years with non-trivial earnings

Average annual real earnings, years worked

Age range

23–28a

27–28b

23–28c

27–28d

Early Gen. X (ref. = L. baby boom)

0.073 (.108)

−0.019 (.043)

−259 (620)

−539 (903)

Late Gen. X

0.338** (.110)

0.079 (.044)

832 (634)

1,388 (918)

Less HS (ref. = HS Grad.)

−1.079** (.194)

−0.212** (.079)

−1,579 (1,245)

−3,309 (1,138)

Bachelor’s degree

0.544** (.156)

0.214** (.063)

7,750** (868)

9,856** (1,228)

Less HS × Early Gen. X

−0.368 (.264)

−0.222* (.107)

50 (1,759)

727 (2,562)

Less HS × Late Gen. X

−0.149 (.294)

−0.152 (.119)

129 (1,896)

1,686 (2,818)

BA × Early Gen. X

−0.254 (.198)

0.030 (.084)

−1,247 (1,154)

979 (1,627)

BA × Late Gen. X

−0.036 (.207)

−0.044 (.080)

805 (1,106)

1,791 (1,563)

Number of children

−0.562** (.033)

−0.175** (.013)

−1,340** (214)

−2,441** (305)

Currently divorced (ref. = currently married)e

0.252 (.135)

0.176** (.055)

−656 (767)

364 (1,101)

Never married

−0.112 (.082)

0.015 (.033)

−687 (474)

142 (668)

Black

−0.130 (.110)

0.027 (.045)

−1,191 (653)

−1,309 (929)

Hispanic

0.018 (.113)

0.070 (.046)

679 (673)

669 (975)

Other

−0.791** (.159)

−0.135* (.064)

2,487** (946)

3,099* (1,353)

Years of non-trivial earnings

 Ages 23–28

  

2,698** (138)

 

 Ages 27–28

   

11,752** (853)

Self-employment earnings

 Ages 23–28 (1 = yes, 0 = no)

  

−1,204* (596)

 

 Ages 27–28

   

−3,565** (1,046)

Intercept

5.025** (0.096)

1.671** (0.039)

7,592** (912)

1,949 (1,797)

Adjusted R2

0.216

0.142

0.323

0.286

Sample size

2,971

2,971

2,654

2,353

Source: Authors’ calculations using SSA administrative earnings records matched to the 1990, 1996, and 2004 SIPP (wave 2)

Notes: Standard errors in parentheses. * p < 0.05; ** p < 0.01. HS high school; Less HS no high school diploma; BA at least bachelor’s degree

aMaximum value of 6

bMaximum value of 2

cWomen with zero years of non-trivial earnings between ages 23 and 28 excluded

dWomen with zero years of non-trivial earnings between ages 27 and 28 excluded

eWidow coefficient not included due to small sample size

To assess earnings, our regression models estimate average annual real earnings for years worked, from ages 23 through 28 and ages 27 through 28. Because having greater experience in the workforce would likely increase a person’s earnings potential, the models introduce years of non-trivial earnings over the age range as a control variable. Moreover, to control for cohort differences in self-employment and its potential effect on earnings, the models include a dummy variable denoting the presence of self-employment earnings.

The results (Table 6, right-hand columns) are consistent with our prior descriptive findings. Cohort is not significantly associated with women’s average annual earnings for years worked over young adulthood, holding other factors constant. Having a bachelor’s degree increased annual earnings at both age ranges. Importantly, the association between bachelor’s degree and earnings did not vary significantly across the cohorts. With respect to other factors, number of children significantly decreased average earnings over young adulthood, as did having self-employment income.

Conclusions

Family and labor decisions at young adulthood have important implications for women’s life-course trajectory. This article offers insight into young adult women’s childbearing, marriage, and paid work characteristics across three cohorts. Our substantive interest in Generation X stems from increasing focus placed on the possibility of pending changes in family and labor market behavior of women entering adulthood in the late 1990s to date.

Using several panels of the SIPP marital and fertility history modules matched to SSA earnings records, our analysis addresses the question of whether Generation X women, particularly college graduates, are following different family and labor patterns than late baby boomer women at young adulthood. There are several noteworthy findings. First, consistent with Percheski (2008), our results show stability in childbearing patterns across the three cohorts at age 27 and 28. Second, Generation X women, relative to late boomers, were less likely to have married by age 27 and 28, particularly college graduates. Third, in terms of labor force activity, late Generation X women (1975–1977) had greater total engagement with paid work, and higher average annual earnings when they worked, from age 23 to 28 than their late baby boom counterparts over the same age range. Among college graduates, cohort differences in labor force engagement and earnings levels (conditional on work) over young adulthood were, on the whole, absent.

Put together, our findings speak to broader debates about family and labor market behavior among U.S. women. The dominant narrative in the popular press has been that professional women are increasingly ‘opting out’ of the paid labor force to focus on family and childrearing. Some empirical research has supported this view. Constructing synthetic cohorts using adjacent cross-sectional surveys in the CPS, Vere (2007) provides evidence that college-educated women born toward the end of Generation X have higher fertility and work fewer cumulative hours up to age 27 than women born 10–15 years earlier. Using longitudinal and retrospective life history data lead us to different conclusions, which do not support the characterization of a substantive reversal in family and labor behavior among young adult women born toward the end of Generation X. Our findings thus add to recent studies showing little evidence that recent cohorts of professional women are actually ‘opting out’ (Boushey 2008; Percheski 2008).

Our results also provide support for a methodological approach based on life course analysis of longitudinal data for individuals in different birth cohorts at the same stage in life. Our analysis, which focuses on individuals’ lives at ages 23 through 28 for three different cohorts, highlight some of the advantages of using longitudinal and retrospective cohort data to address life course questions, rather than synthetic cohort data from a series of cross-sections. While valuable in its own right, cross-sectional surveys refer to samples in particular years which may yield different populations over time due to selective entry and exit in the labor force, as well as mortality and migration patterns. Furthermore, our results may help inform demographic projections and micro-simulation models designed to analyze proposals and evaluate the population served by Social Security and other government programs. This research suggests that so far Generation X women reflect rather similar family and earnings behavior to late baby boomers, at least over young adulthood. Notable differences by ages 27 and 28 include increasing prevalence of graduate degrees, as well as a downward trend in marriage among college and non-college graduates.

There are limitations to our analysis that merit further study. First, without a parallel longitudinal measure of labor hours, the present study cannot directly evaluate whether Generation X women are working fewer or longer hours over young adulthood than their late boomer counterparts. A key factor influencing observed earnings levels is the amount of labor supplied and whether work is part-time or full-time. Other factors include the types of jobs women hold, the corresponding wage rate, and the business cycle, among other structural factors in the economy that may differentiate each cohort. Second, the worsening economic environment since the early to mid-2000s suggests that the earnings levels of late Generation X women over young adulthood might reflect somewhat of a high point.

Third, we use a narrow age range, which limits the scope of our analysis. While selecting ages 27 and 28 provided an opportunity to compare the most recent cohorts of Generation X that available SIPP data allow, our findings only begin to fill the gap in terms of understanding women’s family and labor experiences in post-boomer cohorts. Change in family and paid work trends among Generation X, such as ‘opting out,’ may only become evident as cohort members age. For example, given the younger age range of our analysis, it is premature to know how much of observed trends may change due to later childbearing or marriage, particularly for highly educated women (Schoen 2004; Sullivan 2005).

As new data become available, the literature would benefit from additional research on post-boomer cohorts. Further research is necessary to determine potential cohort differences in the relationship between family structure, education, and earnings. Research focused on possible cohort differences in the labor behavior of recent mothers across different education or race/ethnic subgroups may be a particularly useful avenue to further consider the ‘opt out’ thesis. The interplay between women’s rising educational credentials and family and earnings outcomes beyond young adulthood also deserves empirical attention. For example, analysis on Generation X women at later phases of the life course, such as when they have reached the end of childbearing years, would further clarify possible cohort-level changes in women’s family and employment characteristics.

Footnotes
1

Typically, the baby-boom cohort is defined as persons born between 1946 and 1964. Generation X has been conditionally defined by the U.S. Census Bureau as persons born between the years 1968 and 1979 (Crowley 2003).

 
2

A variety of factors have contributed to the growth in women’s labor participation rate over recent decades, such as rising wages and educational attainment, increasing opportunities in the workplace, changes in the American family, and shifting socio-cultural attitudes about women’s role in work and family (see Blau et al. 2006; Thornton and Young-DeMarco 2001).

 
3

A birth cohort reflects the year or group of years in which a person was born. In essence, grouping by birth year connects persons born at the same time with the historical context shaping the life course (Elder et al. 2003).

 
4

Young adulthood is defined in diverse ways. We use the concept here to emphasize a life course stage often dense in transitions to ‘adulthood’, such as finishing education, having a child, marriage, and full-time employment. For a valuable perspective on the features of ‘emerging adulthood’, from the late teens to mid- to late 20s, see Arnett (2004).

 
5

Our two-year age groups correspond to three-year birth cohorts because persons born over three calendar years can be 27 or 28 years old at the time of the survey.

 
6

Unlike 1990 and 1996, the 2004 SIPP Marital History Public-Use File suppresses marital event dates. Through an agreement with the U.S. Census Bureau, SSA has access to a Restricted-Use File with such information. All users must be authorized by the U.S. Census Bureau.

 
7

Our method of observing earnings from ages 23 to 28 across the cohorts is as follows. For the 2004 SIPP panel, annual earnings from Social Security records are observed for N sample of women who are age 28 from 1999 through 2004; for women who are age 27, administrative earnings are observed from 2000 through 2005. For the 1996 panel, young adult earnings histories consist of annual earnings from 1991 through 1996 for women age 28, and from 1992 through 1997, for women age 27. In the 1990 panel, administrative earnings are observed from 1985 through 1990 for women aged 28 and from 1986 through 1991 for women age 27.

 
8

For more information on the complex relationship between educational attainment and childbearing see Kuperberg (2009) and Schoen (2004).

 
9

One underlying reason for this pattern may be associated with higher incidence of cohabitation in more recent cohorts (Bumpass and Lu 2000; Manning et al. 2007; Raley 2000)

 
10

Cohort differences in cumulative earnings over young adulthood could be influenced by economic cycles, among other factors, not controlled for in our descriptive analysis.

 

Acknowledgements

We thank the Editor and the anonymous reviewers of Population Research and Policy Review for helpful comments and suggestions. The authors also thank Hilary Waldron, Irena Dushi, and Glenn Springstead for valuable comments. The findings and conclusions presented in this paper are those of the authors and do not represent the views of the U.S. Social Security Administration. The administrative earnings data used in this paper are restricted use; users must receive approval of the Social Security Administration and the U.S. Census Bureau.

Copyright information

© US Government 2010