Does Occupational Mobility Influence Health among Working Women? Comparing Objective and Subjective Measures of Work Trajectories

Abstract

Occupational mobility is highly valued in American society, but is it consequential to women’s health? Previous studies have yielded inconsistent results, but most measured occupational mobility by identifying transitions across occupational categories. Drawing from cumulative inequality theory, this study (1) compares objective and subjective measures of work trajectories and (2) examines the contributions of each to self-rated health. With 36 years of data from the National Longitudinal Survey of Mature Women (1967-2003), growth curve models are used to estimate the effects of middle-aged work trajectories on health among 2,503 U.S. women. Work trajectories as measured by the Duncan Socioeconomic Index predict health, but not after adjustment for perceived work trajectories and status characteristics. The findings reveal that subjective measures of occupational mobility provide important information for assessing health consequences of work transitions and that downward occupational mobility in middle age is deleterious to women’s health in later life.

Keywords

cumulative inequality longitudinal data occupational mobility perceptions self-rated health

Despite the vast stock of knowledge on the work-health relationship, research on the health consequences of occupational mobility yields inconsistent results. Some studies emphasize the influence of upward and downward occupational mobility on outcomes ranging from psychological well-being (Liljegren and Ekberg 2008) and long-term illness (Bartley and Plewis 1997) to longevity (Kern et al. 2009). By contrast, other studies show that occupational mobility is less consequential than being occupationally stable (e.g., Su 2009). For instance, Hart, Smith, and Blane (1998:1129) reveal “major” mortality differentials between the stable nonmanual and stable manual occupational groups, with “mobility making only a minor contribution.”

There are two distinguishing characteristics of the literature on occupational mobility and health that are relevant for the present investigation. First, in all of these studies, the measurement of occupational mobility is limited to categorical change in the occupational hierarchy such as transitions across classes of workers (e.g., Bartley and Plewis 1997; Hart et al. 1998). Identifying change in occupational groupings is very helpful for studying the relationship between occupational mobility and health, but none of these studies probe how the respondent actually views the occupational change.

Second, although there are studies of occupational mobility and health among men in the United States (e.g., Kern et al. 2009; Su 2009), the vast majority of studies that examine women or compare men and women are based on samples from Europe. Scholars have studied how U.S. women’s careers have long been constrained by occupational segregation, glass ceilings, and women’s more complex work trajectories (Moen 2005; Rosenfeld 1992), but the relationship between their occupational mobility and health has received relatively little attention. Given the considerable mobility in the United States, studying the effect of occupational mobility on health among American women may be illustrative for other nations with open stratification systems.

The purpose of this research is to address several gaps in the literature. We compare objective and subjective measures of occupational mobility in order to enhance our understanding of the health consequences of women’s work trajectories. It is possible that failure to account for the respondent’s view of occupational mobility may contribute to some of the disparate findings in the literature. Moreover, how women viewed their occupational mobility during the past half century is intriguing because it was a period of increasing occupational opportunity for women in the United States. For the present analysis, we examine the health consequences of occupational mobility on middle-aged American women followed over 36 years. We draw from cumulative inequality theory to inform the analysis.

Background

Cumulative inequality theory provides the framework for this analysis and helps formulate the specific research questions. The theory draws on a number of other theories and perspectives—including age stratification, cumulative disadvantage, life course, and stress-process—to address how stratification manifests itself over the life course (Ferraro, Shippee, and Schafer 2009). Specified in five axioms and 19 propositions, cumulative inequality theory emphasizes the manifestation and accumulation of inequality over the life course. According to cumulative inequality theory, differentiation in occupational and other adult statuses is shaped through social systems via demographic and developmental processes, including family functioning, childhood experiences, and cohort succession (axiom 1) (Ferraro et al. 2009; Riley 1987).

Although these processes are known to influence occupational status at any point in time, cumulative inequality theory can also be used to enhance our understanding of occupational mobility. Occupational stratification processes do not cease in early middle age but continue throughout the working years and influence labor force exits in later life. To understand intra-individual change in occupational status during adulthood, we find that axioms 2 and 3 of the theory are especially germane. Axiom 2 of cumulative inequality theory posits that disadvantage increases exposure to risk, whereas advantage leads to opportunity (Ferraro and Shippee 2009). We define opportunity as a promising configuration of circumstances that increases the likelihood of advancement or improvement. Thus, opportunity is greater when resources are available, able to be mobilized, and relevant for the life situation of the person (i.e., useful for desired improvement). Although some opportunities are tied to a single law or social condition, the theory emphasizes how the accumulation of resources and risks shape opportunities. When thinking of life course stratification, it is the constellation of conditions at various life stages that is essential to whether one rises or falls in the social hierarchy.

Linking the theory to occupational mobility, we conceptualize a person with higher occupational status as having greater opportunity than that afforded to a person with lower occupational status. The opportunities associated with higher status positions are diverse, but we see a greater chance of high-status achievement as well as both occupational and financial security. This is especially the case when a person reaches high status early in one’s career (proposition 2.c; see also Merton 1968).

Persons in lower status positions still have opportunity, but hundreds of sociological studies reveal that it is more limited. This is because people in lower status jobs are at greater risk of unemployment, financial strain (perhaps even due to lack of health insurance), and occupational injury or exposure to hazardous materials. This is not to say that persons in lower status jobs cannot achieve very high status in many modern societies but rather to acknowledge that the odds of doing so are very low. Moreover, the networks in which one interacts are also influential. Persons in higher status occupations are connected to networks that can more effectively aid upward mobility compared with individuals in lower status occupations. This expectation is consistent with the concept of biographical structuration introduced in cumulative inequality theory (Schafer, Ferraro, and Mustillo 2011) as well as Sampson and Laub’s (2005) depiction of situated choice. Also, according to fundamental cause theory, which posits that people of high socioeconomic status (SES) are more likely than those of low SES to have access to the best resources, one would anticipate that higher status persons are positioned for more opportunity and better health (Link and Phelan 1995).

In a departure from previous theories of advantage/disadvantage, axiom 3 of cumulative inequality theory specifies that perceptions of one’s life situation are important to stratification processes and may modify the long-term consequences of early disadvantage. In other words, subjective evaluations of life circumstances may redirect individuals’ life trajectories, with favorable evaluations leading to more positive outcomes. People reflect on the past and present to project into the future, and these temporal evaluations can alter lines of action and one’s sense of well-being (Carstensen 2006; Schafer et al. 2011).

Although objective circumstances in one’s life are essential to the accumulation of disadvantage over the life course, perceptions of relative advantage or disadvantage also contribute to one’s future trajectory (Ferraro et al. 2009). For instance, regardless of whether one is succeeding in his or her job, the perception of how one is doing may have real consequences for the future of one’s trajectory—career or otherwise (Mandal, Ayyagari, and Gallo 2011). Such perceptions involve a complex set of ideas about oneself derived from processes that allow for individuals to assess both the “good” and “bad” in their lives (Festinger 1954). This assessment is based on (1) perceptions of progress relative to others (Festinger 1954) and (2) knowledge of past trajectories (Pearlin et al. 2007). Although comparisons can occur in any arena, work is a particularly important platform on which people learn about themselves and others (Moen 2005). The workplace is a hierarchical system where actors can compare themselves to co-workers and, more generally, to others elsewhere who hold the same position. These comparisons may occur at any point in time, but the actor is also splicing together these evaluations into a perceived work trajectory.

Thus, cumulative inequality theory seeks to integrate perceptions of life conditions and trajectories into the study of inequality along with the objective measures. Our aim is not to replace the objective measures but to consider both objective and subjective domains when evaluating the health consequences of occupational mobility. Both resources and perceived trajectories are seen as important phenomena that may alter trajectories in one or more domains of life (Ferraro et al. 2009). In a changing world of work, the availability of resources is important for career advancement; however, perhaps equally important is one’s interpretation of occupational mobility. Our empirical research asks whether such subjective interpretations of occupational mobility are consequential to health trajectories, above and beyond the objectively measured change in occupational status.

The Thomas Theorem and Women’s Work Trajectories

More generally, judging one’s career in light of others is consistent with the Thomas theorem (Thomas and Thomas 1928): situations defined as real are real in their consequences. Regardless of whether women move from one occupational class to another, the critical issue is how they view any change. Is upward mobility fulfilling, reflecting a milestone of personal achievement? Or might upward mobility feel somewhat hollow if perceived as delayed or lacking in commensurate responsibility? And is downward mobility harmful to women’s health? Especially for studies of the health consequences of occupational mobility, how the actor views the mobility seems critically important. Favorable views of the change may add to the salubrious quality of higher status, but unfavorable views of the change may amplify the stress of achieving higher status (Schieman, Whitestone, and Van Gundy 2006).

Previous research reveals that subjective perceptions of social status—sometimes referred to as the SES ladder—are equal to or more powerful predictors of health than are objective indicators of SES (e.g., Demakakos et al. 2008). One explanation is that individuals’ perceptions of their social status may in fact be a more precise measure of their actual social status (Singh-Manoux, Marmot, and Adler 2005). Moreover, when assessing occupational mobility, subjective evaluations are probably more critical because actors are repeatedly judging themselves in comparison to others and melding those judgments into a perceived work trajectory (Settersten 2004).

For the present research, we contend that these perceived trajectories merit greater attention in order to understand the links between occupational mobility and health. Perhaps the inconsistency in findings noted earlier is partially due to the exclusive use of objective indicators of occupational mobility. Indeed, a woman may perceive that she is progressing at work due to rising wages, greater responsibilities within the same job, improved skills so that the same work feels easier, or even fewer frustrations or disciplinary issues at work, but these would not be captured by conventional measures of occupational mobility. We seek to bring the woman’s view of her work situation to the forefront in hopes of giving a fuller picture of how occupational mobility may be related to health.

Research Questions

Guided by cumulative inequality theory and gaps in the existing literature, we examine the link between occupational mobility and women’s health. A strong test of the theory requires objective information as well as some type of perceptual information on women’s work trajectories. To use the former only would neglect how the woman sees her work situation; to use the latter only may misrepresent the actual change experienced. We consider both measurements in this research and ask two main questions:

What is the association between women’s objective work trajectories and their self-perceived trajectories? We compare an objective measure of work trajectory (based on observed change in the Duncan Socioeconomic Index) with a subjective measure of work trajectory. We anticipate some overlap between the objective and subjective evaluations but also view them as separate and distinct measures. Subjective evaluations, for instance, may capture change not measured by using objective indicators and vice versa. Drawing on cumulative inequality theory, we anticipate that women with higher SES and resources will be more likely to hold favorable views of their work trajectories (i.e., advantage leads to opportunity).

Do subjective assessments of work trajectories have independent effects on women’s health net of objective measures? Here we ask whether subjective evaluations of one’s work trajectories add anything to what can be known from the objective measures alone. Reflecting how the person evaluates the situation, we expect that women’s perceptions of their work trajectories will be associated with health changes, over and above the contribution of objective measures. We anticipate consistency across the objective and subjective measures; the question is what role each measure has on health over three decades of women’s lives.

Data and Methods

Sample

The study is based on data from The National Longitudinal Survey of Mature Women (NLSMW). Beginning with a representative sample of 5,083 women aged 30 to 44 in 1967, respondents were interviewed a total of 21 times over a 36-year period. By 2003, most of the sample attrition was due to mortality (n = 651 or 26%); other sources of attrition were refusal (n = 305 or 12%) and failure to locate (n = 87 or 3%).

The sample for this analysis was limited to women who were employed during the first decade of the panel study (1967-1977) and had valid responses on both perceived work trajectory variables, resulting in an initial sample of 2,503 working women. Women were asked about their perceived work trajectories in 1972 and 1977 only if they were employed in the previous five years. Compared with those excluded, the analytic sample was, as expected, somewhat younger, more educated, and less likely to be married. They also had higher wages and better self-ratings of health in 1967 and at the conclusion of the study in 2003. Preliminary analyses showed no difference in results based on whether the sample was comprised of women who had valid scores at one versus two time points.

Measures

Self-Rated Health

In this article, we model trajectories of self-rated health. Research shows that self-rated health is a reliable predictor of mortality and other health outcomes above and beyond objective health status measures and other known risk factors (Idler and Benyamini 1997). Also, self-rated health is consistently asked over the duration of the study unlike most other health-related variables in the NLSMW.

Data on women’s self-rated health were collected at eight time points over the course of the study—1967, 1986, 1992, 1995, 1997, 1999, 2001, and 2003. Respondents were asked to rate their health in comparison to other women their age as excellent, good, fair, or poor. The variables range from 1 to 4, with higher values corresponding to better self-rated health. This wording of self-rated health has been used in a number of studies where respondents are asked to compare their health to that of their peers (e.g., Appels et al. 1996). The findings are very consistent with the more widely used measures of self-rated health, leading Idler and Benyamini (1997) to conclude that “the consistency of the effects seems to show that the concept of self-rated health is relatively insensitive to the semantic variations in the questions eliciting it” (p. 22). Women’s health trajectories are estimated with growth curve models capturing eight measurement occasions for survivors as well as the available measurements from those women who did not complete all waves.

Work Trajectories

Occupational mobility may be defined in several ways, but this analysis examines intragenerational mobility and only over a limited period of the life course. The NLSMW contains both objective and subjective measures of such mobility during the study’s first decade. Although it is a limitation that both measures are not available over the duration of the study, being able to compare the two measures over 10 years is invaluable for enhancing our understanding of the relationship between occupational mobility and health. To clarify that the NLSMW does not measure lifetime occupational mobility, we refer to the objective and subjective measures of occupational change spanning 10 years only as work trajectories. For context, the women ranged in age from 30 to 44 at the beginning of the study; thus, the 10-year period for measuring work trajectories is as the women reach ages 40 to 54. This period captures women during middle age when they are less likely to exit the labor force indefinitely due to marriage or childbearing but observes women during a period when they are more likely to be occupationally mobile (Rosenfeld 1992).

Duncan Work Trajectories

Our objective measure of work trajectories is based on the Duncan Socioeconomic Index (Duncan 1961). Prestige scores (two-digit codes) have a theoretical range of 0 to 97 and are based on income and education distributions associated with occupations identified in the 1960 Census (U.S. Department of Labor 2001). Prestige scores were assigned to respondents’ current or most recent occupation. The analyses below make use of the initial (1967) Duncan score as well as change in Duncan scores over time. The variable for Duncan work trajectories was created as a change score by subtracting women’s occupational prestige scores in 1967 from their scores in 1977. The variable ranges from −76 to 77 but was categorized and coded as follows: 1, decline of 10 or more prestige points; 2, decline less than 10 points; 3, Duncan score unchanged; 4, increase less than 10 prestige points; and 5, increase of 10 or more points.

Perceived Work Trajectories

The subjective assessment of work trajectories was based on questions measured in 1972 and 1977. The indicator for perceived work trajectories is derived from the question “All in all, so far as your work is concerned, would you say that you’ve progressed during the past five years, moved backward, or just about held your own?” Response categories range from 1 (moved backward) to 3 (progressed). The two questions, each tapping a 5-year period, were used to create a variable for perceived work trajectories from 1967 to 1977. It was coded as follows: 1, decline; 2, limited decline; 3, unchanged; 4, limited progress; and 5, progress. (Supplementary analyses with alternative coding schemes for both Duncan and perceived work trajectories, including using different thresholds on Duncan work trajectories, yielded conclusions similar to those presented below.)

Employment Characteristics

Work trajectories are contextualized on employment history, which may also have important effects on women’s health. Thus, we control for two additional employment characteristics. First, to account for labor force entries and exits, the analysis adjusts for spells of unemployment during the first decade. In 1967, 1972, and 1977, women were asked about time spent unemployed in the previous year. The indicator for unemployment is a continuous variable, measured in weeks, equal to the duration of time spent unemployed. Unemployment in the NLSMW refers to those without a job or actively seeking work. Second, the analysis differentiates full versus part-time employment because the impact of occupation on health may vary for part-time workers (Ross and Mirowsky 1995). A binary variable is used to account for full-time employment, with 1 equal to women who worked 35 or more hours per week and 0 equal to those who were nonemployed or worked fewer than 35 hours per week at baseline.

Status and Health Characteristics

Based on the literature examining occupational mobility and health, we adjusted for several additional variables. Chronological age is measured in years. Black is a binary variable coded 1 for black respondents and coded 0 for nonblack respondents (hereafter referred to as white). South is a binary variable coded 1 for respondents who live in that region of the United States (0, otherwise). Both the education level of the respondent and her parent (head of household) are measured in years of formal education attended. Because income and assets are important to how a woman judges her work trajectory, we specified personal income and household wealth as time-varying covariates. In doing so, we account for changes in wages and accumulated assets, especially because the latter may be influenced by changes in household structure. Data for personal income are drawn from eight different waves—1967, 1986, 1992, 1995, 1997, 1999, 2001, and 2003—and coincide with reports of self-rated health. Data on wealth were not available in 1986 or 1992; for that reason, data for household wealth were drawn from adjacent years at those time points. The analysis also accounts for health limitations affecting function by using a three-category variable that ranges from 1 “health does not affect work or housework” to 3 “health prevents respondent from working.”

Life Course Events

Given the duration of the observation period, the analysis also considers selected life course events that may influence respondents’ perceptions of their life situation, work, household need, and resources. First, we adjust for marital status in 1967 when women’s work trajectories were first observed; married is a binary variable coded 1 for married women and 0 for otherwise. Second, of those who were married in 1967, women who divorced or separated over the course of the study were identified with a binary variable (1 = divorced, 0 = otherwise). Third, the analysis also adjusts for number of children in 1977, which provides a near-complete childbearing history (U.S. Department of Labor 2001).

Table 1 presents the ranges, means, and standard deviations for the variables used.

Table 1.

Means and Standard Deviations of Variables in the NLSMW, 1967-2003 (N = 2,503)^a

	Range	Mean	Standard Deviation
Dependent Variable ^b
Self-rated health, 1967	1-4	3.307	.720
Self-rated health, 2003	1-4	2.724	.892
Independent Variables
Age, 1967	30-44	37.088	4.318
Duncan prestige score, 1967	0-97	34.909	21.750
Duncan work trajectories, 1967-1977	1-5	3.148	1.308
Perceived work trajectories, 1967-1977	1-5	3.821	1.099
Employment characteristics
Full-time, 1967	0/1	.515
Unemployed, 1967-77	0-97	3.627	10.166
Status and health characteristics
Black, 1967	0/1	.293
South, 1967	0/1	.407
Education, 1967	2-18	11.437	2.649
Parents’ education, 1967	0-18	8.120	3.921
Personal income, 1967 ($)^c	0-15	1.974	2.270
Household wealth, 1967 ($)^c	−46.275-76	8.923	14.193
Health limitations, 1967	1-3	1.171	.435
Life course events
Married, 1967	0/1	.789
Divorced, 1969-2003	0/1	.161
Number of children, 1977	0-17	3.278	2.493

Notes: All dichotomous variables are scored 0 and 1 (0 = no or otherwise). The standard deviation of a dichotomous variable is omitted because it is a function of the mean.

N varies depending on the wave.

Self-rated health was measured on eight occasions during the study; the first and last measurements are displayed.

Time-varying covariate; measured in thousands.

Analytic Plan

Our first research question asks whether there is an association between objective work trajectories and women’s perceived work trajectories. We use two different procedures in addressing this question. First, we compare Duncan work trajectories and perceived work trajectories in a simple cross-tabulation to examine the degree of association. Second, we use ordered logistic regression to identify which women held more favorable perceived work trajectories after adjusting for initial differences in work status as well as Duncan work trajectories.

Our second research question focuses on whether women’s perceived work trajectories have independent effects on health net of objective measures. The data array over 36 years permitted the use of nested growth curve models to examine women’s health trajectories over time. Whereas self-rated health is an ordinal variable, modeling was completed with the Stata program gllamm (Rabe-Hesketh and Skrondal 2008). The modeling consists of two-levels of analysis: occasions represent Level 1 units, and subjects represent Level 2 units. Adaptive quadrature estimation techniques were used in the analysis. Mean-centered age was used as the time metric; a quadratic term was also used to test for nonlinearity but was nonsignificant in preliminary results and therefore excluded from subsequent analyses.

Multiple imputation was used to handle missing data in Stata. We used pooled results from 10 imputed data sets to estimate the fixed effects and a single imputed data set to estimate the variance components. Since the variance function is quadratic in nature, the variance components cannot be averaged across the imputed data sets, as is the case with the linear fixed effects.

The growth curve analyses were divided into two main stages. First, random-intercept proportional-odds models were used to estimate the effects of occupational mobility on women’s self-rated health. A model including age only was used to evaluate the change in respondents’ health trajectories over time. Second, random-intercepts and slopes-as-outcomes proportional-odds models were used to estimate the effects of occupational mobility adjusting for covariates and a series of interaction terms. Interactions between mean-centered age (the time metric) and perceived work trajectories and Duncan work trajectories were first used to assess the long-term consequences of occupational mobility on health. Next, we used an interaction term between mean-centered age and race given a fairly extensive literature that testifies to differences in black and white women’s experiences at work (e.g., Beal 2008). This interaction term examines the effects of race on changes in women’s self-ratings of health over the 36-year period.

Results

As shown in Table 1, the mean for self-rated health in 1967 was 3.307, with the largest proportion of women reporting excellent health in comparison to other women their age. Women’s health, however, declined by 2003 (from 3.307 to 2.724). This decline is to be expected because respondents were, on average, 37 years of age at the beginning of the study but were 73 by 2003. Duncan work trajectories largely reflect stability, with a mean of 3.148 and a mode of 3 (i.e., unchanged). The mean for perceived work trajectories was more favorable, with a mean of 3.821 and a mode of 5 (i.e., progress).

Our first research question focuses on whether objective work trajectories are associated with women’s self-perceived trajectories. Duncan work trajectories represent change in the socioeconomic index, but do women’s evaluations of their work trajectories capture change not measured by the Duncan variable? To gain a sense of the degree of association, Table 2 presents a cross-tabulation of these two variables. The diagonal represents women who evaluated their work trajectories in a way that might be expected on the basis of Duncan work trajectories. Those women off the diagonal reflect subjective evaluations that would not be expected on the basis of Duncan work trajectories. Although the measures are significantly associated with one another (p < .001), it is clear from the table that many women judged their work trajectories in ways that would not be predicted by using the Duncan measure alone.

Table 2.

Percentage Distribution of Duncan Work Trajectories and Perceived Work Trajectories Over a 10-Year Period (n = 2,440)

	Duncan Work Trajectories
Perceived Work Trajectories	Decline	Limited Decline	Unchanged	Limited Progress	Progress
Decline	6.95	5.88	4.44	7.02	4.03
Limited decline	6.95	4.15	3.56	3.37	3.84
Unchanged	25.67	29.41	29.78	28.09	16.12
Limited progress	32.35	33.91	30.22	31.18	34.36
Progress	28.07	26.64	32.00	30.34	41.65
n	374	289	900	356	521

Notes: Column percentages are shown; χ² = 63.28, p < .001.

Further examination of Table 2 reveals that there are relatively more women in the lower triangle of the table than in the upper triangle, which is indicative of generally more favorable views of work trajectories than one would expect from Duncan work trajectories alone.

Although the bivariate analysis in Table 2 is informative, Table 3 presents ordered logistic regression models of women’s perceived work trajectories to account for other variables that may be related to work trajectories. Model 1 includes initial prestige scores and work trajectories based on the Duncan index and employment characteristics. Results reveal that initial Duncan scores and Duncan work trajectories had positive effects on women’s perceived work trajectories (p < .001). Women in initially higher status positions and women who experienced upward occupational mobility were more likely to perceive their work trajectory in a favorable way. Beyond the two Duncan measures, only spells of unemployment influenced perceived work trajectories; predictably, unemployment spells were associated with more negative perceived work trajectories.

Table 3.

Ordered Logistic Regression of Perceived Work Trajectories on Independent Variables in the NLSMW, 1967-1977 (N = 2,503)

Independent Variables	Model 1	Model 2
Duncan prestige score, 1967	1.027***^a	1.017***
	(1.024, 1.031)^b	(1.012, 1.022)
Duncan work trajectories, 1967-77	1.341***	1.291***
	(1.263, 1.423)	(1.214, 1.372)
Full-time, 1967	1.118	1.107
	(.968, 1.291)	(.930, 1.319)
Unemployed, 1967-77	.986***	.987**
	(.979, .993)	(.980, .995)
Age, 1967		.986
		(.969, 1.003)
Black, 1967		.796*
		(.656, .967)
South, 1967		1.373***
		(1.167, 1.614)
Education, 1967		1.105***
		(1.062, 1.151)
Parents’ education, 1967		1.008
		(.982, 1.035)
Personal income, 1967 ($)		1.018
		(.974, 1.063)
Household wealth, 1967 ($)		1.002
		(.995, 1.008)
Health limitations, 1967		.868
		(.731, 1.032)
Married, 1967		1.096
		(.904, 1.328)
Number of children, 1977		1.011
		(.979, 1.044)

Odds ratio.

Confidence interval.

p < .05. **p < .01. ***p < .001 (two-tailed tests).

Model 2 includes all covariates, and the effect of Duncan work trajectories remains significant albeit attenuated slightly from Model 1 (from odds ratio [OR] = 1.341 to OR = 1.291). Three of the additional covariates are significant in Model 2: black, South, and education. Compared with white women, black women were less likely to have positive perceptions of their progress at work (OR = 0.796). Women residing in the South and those with higher education were more positive in their perceived work trajectories.

Having identified that there are considerable differences between objective and subjective work trajectories, we now turn to growth curve modeling to estimate whether each is consequential to health during the 36-year observation period. In Table 4, Models 1 through 3 represent random-intercept proportional-odds models, allowing individuals to vary in terms of their overall levels of self-rated health. Model 4 uses random intercepts and slopes, using interaction terms between time and select covariates to allow the slopes for those predictors to vary across individuals over time. At each step, we compare log-likelihood ratios between models to test whether adding more predictors actually improves model fit.

Table 4.

Proportional-Odds Growth Curve Models for Self-Rated Health in the NLSMW, 1967-2003

	Model 1	Model 2	Model 3	Model 4
Fixed Effects
Age	.938***^a	.938***	.936***	.924***
	(.935, .941)^b	(.935, .941)	(.932, .939)	(.907, .942)
Duncan prestige score, 1967	1.038***	1.034***	1.011***	1.010***
	(1.034, 1.043)	(1.030, 1.039)	(1.006, 1.016)	(1.005, 1.016)
Duncan work trajectories, 1967-77	1.181***	1.135***	1.043	1.045
	(1.103, 1.265)	(1.059, 1.216)	(.976, 1.115)	(.968, 1.128)
Perceived work trajectories, 1967-77		1.294***	1.198***	1.221***
		(1.193, 1.403)	(1.110, 1.294)	(1.115, 1.337)
Employment characteristics
Full-time, 1967	.845	.833*	.891	.906
	(.713, 1.001)	(.704, .986)	(.757, 1.050)	(.767, 1.069)
Unemployed, 1967-77	.986**	.988*	.992	.995
	(.977, .995)	(.979, .998)	(.983, 1.001)	(.985, 1.004)
Status and health characteristics
Black, 1967			.671***	.592***
			(.543, .829)	(.464, .755)
South, 1967			.714***	.751**
			(.598, .853)	(.627, .899)
Education, 1967			1.144***	1.168***
			(1.096, 1.194)	(1.118, 1.220)
Parents’ education, 1967			1.021	1.032*
			(.994, 1.050)	(1.003, 1.061)
Health limitations, 1967			.361***	.245***
			(.300, .434)	(.203, .297)
Personal income, 1967-2003^c			1.021***	1.021***
			(1.016, 1.026)	(1.015, 1.026)
Household wealth, 1967-2003^c			1.001***	1.001***
			(1.000, 1.001)	(1.000, 1.001)
Life course events
Married, 1967			1.321*	1.315*
			(1.070, 1.632)	(1.061, 1.630)
Divorced, 1969-2003			.821	.854
			(.659, 1.022)	(.686, 1.064)
Number of children, 1977			1.002	.991
			(.968, 1.038)	(.957, 1.027)
Interaction terms
Duncan Work Trajectories × Age				1.001
				(.998, 1.005)
Perceived Work Trajectories × Age				1.001
				(.997, 1.005)
Black × Age				.992
				(.983, 1.002)
Fixed Effects: Thresholds ^d
k1	−2.066***	−1.329***	−2.063***	−2.504***
k2	.406*	1.143***	.431	.291
k3	3.614***	4.350***	3.646***	3.797***
Random Effects: Variances and Covariances ^d
ψ11	3.579***	3.502***	2.979***	4.175***
ψ12				.095
cor(2,1)				.679
ψ13				.005***
Log likelihood	−14587.351	−14567.954	−14381.886	−14085.172
n	2,499	2,499	2,499	2,499

Odds ratios.

Confidence interval.

Time-varying covariate.

Fixed effects thresholds and random effects are derived from a single imputed data set.

p < .05. **p < .01. ***p < .001 (two-tailed tests).

Prior to estimating the analyses presented in Table 4, an unconditional growth model revealed health decline as women aged (results not shown). The random-intercept variance was estimated at 4.263, with a standard error of .180. This represents an intraclass correlation of .56, indicating that 56% of the total variance was between individuals rather than within (Rabe-Hesketh and Skrondal 2008).

Model 1 estimates the effects of Duncan work trajectories and other work characteristics on women’s self-rated health over 36 years. Results reveal that both initial Duncan scores and Duncan work trajectories have positive effects on women’s self-rated health. Of note, for every unit increase in Duncan work trajectories, the overall level of self-rated health increased by 18 percent: Upward occupational mobility is associated with better health trajectories (p < .001). Women reporting time spent unemployed were also less likely to report better self-ratings of health (p < .01). The random-intercept variance, which represents individual differences in the overall mean level of response after controlling for covariates, was estimated at 3.579 and was lower in between-person heterogeneity compared with the unconditional growth model. Supplementary analyses adding occupational categories (six binary variables with a reference group) to the model revealed the same basic conclusions.

In Model 2, perceived work trajectories were added to Model 1, improving model fit. Perceived work trajectories were a significant and positive predictor of self-rated health: for every unit increase in perceived work trajectories, the odds of being in a higher category of self-rated health increased by nearly 29 percent (p < .001). In other respects, results were generally similar to those found in Model 1: Being unemployed and having a lower initial Duncan score decreased the odds of being in a higher category of self-rated health. The effect of Duncan work trajectories on women’s health, however, was slightly attenuated when perceived work trajectories were added to the model: For every unit increase in Duncan work trajectories, the odds of being in a higher overall category of self-rated health increased by 14 percent (compared with 18 percent in Model 1). The attenuation effect was significant at the .001 level (Sobel 1982). The random-intercept variance in Model 2 was slightly lower than in Model 1, estimated at 3.502, with a standard error of .153.

Model 3 incorporated demographic characteristics, health, and life course events and, in doing so, rendered the effect of Duncan work trajectories nonsignificant. Interestingly, perceived work trajectories remained a significant and positive predictor of overall self-rated health: For every unit increase in perceived work trajectories, the odds of having better health increased by nearly 20 percent (p < .001). Educational attainment was associated with higher odds of a better category of self-rated health (p < .001), whereas health limitations reduced the likelihood of a better health category by almost 64 percent (p < .001). In addition, being black and residing in the South decreased the odds of being in a higher category of self-rated health by 33 percent and 29 percent, respectively (p < .001). Changes in personal income and household wealth had significant and positive effects on self-rated health (p < .001), along with being married (p < .05). The random-intercept variance in Model 3 was markedly lower than in Model 2 and was estimated at 2.979, with a standard error of .133.

In supplementary analyses, we added covariates in more blocks to better understand how these variables relate to Duncan work trajectories (results not shown). We found that demographic characteristics and not income, wealth, or life course events rendered the effect of Duncan work trajectories nonsignificant. In particular, education had the greatest attenuating effect.

Model 4 includes all covariates and adds random slopes. These interactions are for Duncan work trajectories, perceived work trajectories, and race, each multiplied by the time metric. The random coefficients model allows us to estimate individual slopes and examine whether particular covariates influence the average rate of change in health over time. In Model 4, race had significant effects on initial levels of self-rated health but not on the slopes. Specifically, black women had lower initial odds of being in a higher category of self-rated health (41 percent lower, p < .001) but did not significantly influence trajectories over time, compared with white women. Indeed, none of the interaction terms were significant in Model 4. The estimated slope variance, however, was .005 indicating significant between-person variation in slopes (p < .001). A likelihood ratio test of Model 4 in comparison to the random-intercept model showed that the random coefficients model had significantly better fit.

Discussion

Occupational mobility is highly valued in open stratification systems, but is it consequential to women’s health? Given the inconsistency in empirical answers to this question, we sought to address it by giving special attention to measurement issues. We were struck by the fact that none of the studies of occupational mobility and health used any indicators that tapped how the respondent felt about her work trajectory. Our purpose, therefore, was to examine the health consequences of women’s work trajectories while comparing two approaches to measuring these trajectories. We specified two main research questions to guide the analysis.

Our first research question focused on measuring intragenerational mobility by systematically comparing women’s work trajectories assessed by the Duncan Socioeconomic Index with those assessed by the women themselves (i.e., objective and subjective approaches). We found that women’s perceptions of their work trajectories during middle age did not neatly map onto observed changes in the Duncan Socioeconomic Index. Rather, the perceived work trajectories captured degrees of “progress” or “decline” that one would not expect on the basis of the Duncan Socioeconomic Index alone. Thus, objective and subjective work trajectories represent two important and distinct ways of measuring occupational mobility.

Women’s occupational trajectories, however, should be interpreted in light of major historical trends, including those in politics, marriage, and domestic labor. The 1970s were a critical time for the women’s liberation movement, when growing numbers of women entered the world of work for career advancement. The women in our study were in their late 30s and 40s during this time and experienced markedly different roles and responsibilities than did earlier cohorts, which likely resulted in distinct effects of work on their health and well-being (Pavalko, Gong, and Long 2007). During this time, the roles of marriage and motherhood were being redefined and influenced how women perceived their work advancement. Women began to dedicate more time to paid labor and reported shorter exits from the labor force compared with earlier cohorts (Pavalko and Smith 1999). Women also experienced more pressure with demands for productivity at work, coupled with nonstandard schedules and their competing roles as mothers and caregivers (Carr and Springer 2010; Pavalko et al. 2007). In addition, the late 1970s saw an increase in female-headed households, declines in marriage, lower fertility, and greater educational attainment for women (Blau 1998). These changing historical trends may have resulted in different reasons for women to enter the labor force (or remain at home) as well as changes in women’s agency to make these choices.

Also, the period during which the NLSMW asked women about perceived work trajectories was marked by growing economic uncertainty in the United States. When the NLSMW started in 1967, the U.S. economy was fairly solid (e.g., unemployment was less than 4%). The next decade, however, was marked by economic concerns, during which U.S. unemployment nearly doubled, oil prices spiked in 1973, and inflation rose considerably (U.S. Department of Labor 2009). Given the economic tenor of the 1970s, perhaps simply holding a job and making modest progress were evaluated more favorably than would be the case during a period of “runaway prosperity.” Adding some credence to this interpretation, women who experienced longer spells of unemployment during the first 10 years of this survey were much more negative about their work situation. All of these factors could affect how women perceived their work trajectories and the effect that work had on their health.

The analyses also revealed how other status characteristics and resources influenced perceived work trajectories. First, women with higher levels of education and with higher Duncan prestige scores at the start of the study generally reported more favorable views of their work trajectories—a finding that is consistent with the expectations of cumulative inequality theory (i.e., advantage leads to opportunity). The consequences of education are many, including viewing one’s work situation more positively, which may, in turn, bolster work performance and appraisals by others. Beyond these differences due to status at the beginning of the study, the analysis showed clearly that changing Duncan scores were associated with women’s evaluations of their work. Finally, black women were more likely than white women to perceive their work trajectories in a negative light, which may reflect the many structural disadvantages faced by black women. For instance, we know that black women in the NLSMW were more likely than white women to lack a college education, work in service or private household jobs, suffer spells of unemployment, and be in poorer health. Granted, white women generally had higher status positions, but the more negative view of one’s work trajectory among black women was observed even after controlling for occupational differences as well as personal income and wealth.

Our second question examined the relationship between work trajectories and health over 36 years while extending the comparison of the objective and subjective evaluations of work trajectories. If one uses either Duncan work trajectories or perceived work trajectories, the basic conclusion is the same: Downward trajectories are associated with worse health. Unlike some studies that suggest that positive lifestyle changes may exact a toll on health (Ganzel, Morris, and Wethington 2010), there is no evidence from these analyses that upward mobility is hazardous to women’s health. Rather, upward trajectories were associated with better health over more than three decades—a finding that is consistent with studies of European women (Liljgren and Ekberg 2008; Swaen et al. 2002).

In analyses comparing the prognostic ability of the objective and the subjective work trajectories on health, however, the latter proved to be more consequential. Results from the NLSMW revealed that negative perceived work trajectories were associated with poorer self-rated health over the duration of the 36-year study. Although Duncan work trajectories predict health when perceived work trajectories are omitted from the analysis, Duncan work trajectories are nonsignificant when perceived work trajectories are added to the specification along with status characteristics.

The literature on occupational mobility to date has relied on objective measures (i.e., change in occupation) to capture women’s movement in the labor force. This study, however, sheds light on the importance of considering both objective and subjective measures of mobility in predicting future health and further elucidates the work-health relationship. Compared with studies that focus on women’s movement into and out of the labor force, our findings describe the effects of work trajectories among women employed at multiple points in time. Using two different measures of occupational mobility, we were able to provide a more complete interpretation of the effects of occupational mobility on women’s health, showing that perceived work trajectories predict health over and above the contribution of objective measures. To our knowledge, this is the first study to compare the predictive ability of objective and subjective indicators of occupational mobility on health. The findings presented herein suggest that this is a fruitful line of investigation.

Cumulative inequality theory is an emergent theory, and there are few empirical tests of it. As such, this research sought to test selected elements of the theory; doing so enables us to offer some conclusions about it. There are several findings that are quite consistent with elements of cumulative inequality theory; we highlight three. First, the fact that women with more education had more favorable views of their work trajectories than what one would predict on the basis of Duncan trajectories is consistent with the theoretical statement that advantage leads to opportunity (Ferraro et al. 2009). Second, consistent with the theory, the power of perceived trajectories was manifest in these data. In a departure from cumulative disadvantage theory, cumulative inequality theory gives explicit attention to perceived life trajectories, and the statistical analyses confirm the centrality of such subjective evaluations. Third, given that perceived work trajectories cover only the first 10 years of the study, there is some evidence of an enduring effect due to occupational mobility in middle age: “Early disadvantage shapes the trajectory, most often bringing additional risks” (Ferraro et al. 2009:424).

In contrast, there are spots where cumulative inequality theory did not help interpret findings. Most notably, the analysis in Table 3 revealed that women dwelling in the South had more favorable views of their work trajectories than what one would expect on the basis of Duncan work trajectories. The analysis from Table 4, however, showed that women dwelling in the South actually had poorer health. This raises questions: Why do southern women have more favorable perceived work trajectories when objectively they did not show much upward mobility? Do perceived work trajectories affect health similarly for southern and non-southern women? In supplementary analyses, we tested an interaction between South and perceived work trajectories, but it was nonsignificant. Thus, the poorer health of southern women was not moderated by perceived work trajectories.

Although this topic merits further study and the theory needs to specify likely conditions, whether cultural or otherwise, under which perceived trajectories have more or less influence, we propose that research on reference groups and individual well-being offers one way to understand these findings. Previous studies have found that the most meaningful groups are often those within one’s network (e.g., Hodson 1985). Research has also discussed the role of psychological comparisons with one’s aspiration level as important for job satisfaction. Specifically, Stouffer et al. (1949) showed that individuals assess their situation in life not as much by objective circumstances (whether they got promoted) but by comparisons with their peers. The classic study revealed that soldiers who worked in the Military Police Unit (MP), which had the slowest rate of promotions in the Army, had more favorable views of the promotion system than did their counterparts in the Army Air Corps unit, which had the fastest promotions. The argument for this finding is that those in the MP unit, even when not promoted for a long time, felt the system was fair because others among them were also not promoted.

This work provides one possible interpretation for our finding that women in the South had more favorable perceived work trajectories than their counterparts living in other parts of the country. Although the finding is somewhat counterintuitive, given that women in the South had lower wages and lower rates of employment than their counterparts, based on the reference group theory, they would not be comparing themselves to women in the North; rather, southern women’s social comparisons are with their reference groups: those also living in the South with whom they had regular interaction. Thus, their perception of work progress may be based on comparisons to nonemployed women or those who had even fewer opportunities for advancement.

The current study is clearly not without limitations. First, it was only during the first decade of the study when we could compare the objective and subjective measures of work trajectories. Additional measures of work perceptions would provide a more complete life course picture of occupational mobility and across a variety of economic seasons. The enduring effect of perceived work trajectories is noteworthy, but these results based on the first 10 years may underestimate the true effect of perceived work trajectories on health over the duration of the study. Second, self-rated health is an excellent summary measure of health, but the analysis raises questions about whether specific health domains are more influenced than others by work trajectories. Unfortunately, we are unable to address these questions with the NLSMW. Upward trajectories were not harmful to self-rated health, but other domains of health, including mental health, merit further study.

Despite the limitations, the findings from this study provide a clear and consistent picture of the link between occupational mobility and health. Drawing on rich data from 36 years of the NLSMW, we were able to assess the influence of perceived work trajectories on women’s long-term health, adjusting for objective measures of occupational mobility and other employment characteristics. The findings are consistent with an emergent literature that emphasizes the essential role of subjective evaluations, relative to one’s objective circumstances, for predicting future physical and mental health outcomes (Demakakos et al. 2008; Mandal et al. 2011). Specifically, if one seeks to understand the health consequences of occupational mobility, probing how the respondent views her work trajectory adds considerably to the quest.

There is no evidence from the present study that upward occupational mobility is hazardous to women’s health over the long run. Although there is clearly some personal cost associated with rising occupational status, the health benefits appear to outweigh any costs. Downward mobility in middle age, however, was associated with long-term declines in health that extended into later life.

Footnotes

Acknowledgements

We thank Sarah Mustillo, Kevin Stainback, Mike Vuolo, Jacob Hibel, and Ann Howell for their helpful comments on this manuscript, in addition to four anonymous reviewers.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the National Institute on Aging (T32AG025671-02) and the Purdue Center on Aging and the Life Course. Support to the second author was provided by the Fessler-Lampert Chair of Aging; the University of Minnesota Center on Aging; and the National Center for Research Resources of the National Institutes of Health to the University of Minnesota Clinical and Translational Science Institute (1UL1RR033183).

Bios

Lindsay R. Wilkinson is a dual-title PhD candidate in sociology and gerontology at Purdue University. Her research interests focus on health inequalities over the life course, with particular emphasis on women’s health, socioeconomic status, and race-ethnicity.

Tetyana P. Shippee is assistant professor, division of health policy and management, school of public health, University of Minnesota. Dr. Shippee’s substantive areas of expertise pertain to long-term care, racial disparities in health care utilization, and conceptual/theory development on health and social inequality. More recently, her work has focused on quality of life in nursing homes and long-term-care decision making.

Kenneth F. Ferraro is Distinguished Professor of sociology and director of the Center on Aging and the Life Course at Purdue University. With interests in how stratification processes unfold over the life course, he has developed cumulative inequality theory for the study of human development, aging, and health. He is currently testing elements of the theory in projects examining minority health and the cumulative effects of early misfortune on adult health.

References

Appels

Bosma

Grabauskas

Gostautas

Sturmans

1996. “Self-Rated Health and Mortality in a Lithuanian and a Dutch Population.” Social Science and Medicine 42:681–89.

Bartley

Mel

Plewis

Ian

. 1997. “Does Health-Selective Mobility Account for Socioeconomic Differences in Health? Evidence from England and Wales, 1971 to 1991.” Journal of Health and Social Behavior 38:376–86.

Beal

Francis M.

2008. “Double Jeopardy: To Be Black and Female.” Meridians: Feminism, Race, and Transcendentalism 8:166–76.

Blau

Francine D.

1998. “Trends in the Well-Being of American Women, 1970-1995.” Journal of Economic Literature 36:112–65.

Carr

Deborah

Springer

Kristen

. 2010. “Advances in Families and Health Research in the 21st Century.” Journal of Marriage and Family 72:744–62.

Carstensen

Laura L.

2006. “The Influence of a Sense of Time on Human Development.” Science 312:1913–15.

Demakakos

Panayotes

Nazroo

James

Breeze

Elizabeth

Marmot

Michael

. 2008. “Socioeconomic Status and Health: The Role of Subjective Social Status.” Social Science and Medicine 67:330–40.

Duncan

Otis Dudley

. 1961. “A Socioeconomic Index for all Occupations.” Pp. 109–38 in Occupations and Social Status, edited by Reiss

A. J.

Jr . New York: Free Press.

Ferraro

Kenneth F.

Shippee

Tetyana P.

2009. “Aging and Cumulative Inequality: How Does Inequality Get Under the Skin?” Gerontologist 49:333–43.

10.

Ferraro

Kenneth F.

Shippee

Tetyana P.

Schafer

Markus H.

2009. “Cumulative Inequality Theory for Research on Aging and the Life Course.” Pp. 413-33 in Handbook of Theories of Aging, 2nd ed., edited by Bengston

V. L.

Gans

Putney

N. M.

Silverstein

New York: Springer.

11.

Festinger

Leon

. 1954. “A Theory of Social Comparison Processes.” Human Relations 7:117–40.

12.

Ganzel

Barbara L.

Morris

Pamela A.

Wethington

Elaine

. 2010. “Allostatis and the Human Brain: Integrating Models of Stress from the Social and Life Science.” Psychological Review 117:134–74.

13.

Hart

Carole L.

Smith

George Davey

Blane

David

. 1998. “Social Mobility and 21 Year Mortality in a Cohort of Scottish Men.” Social Science and Medicine 47:1121–30.

14.

Hodson

Randy

. 1985. “Workers’ Comparisons and Job Satisfaction.” Social Science Quarterly 66:266–80.

15.

Idler

Ellen I.

Benyamini

Yael

. 1997. “Self-Rated Health and Mortality: A Review of Twenty-Seven Community Studies.” Journal of Health and Social Behavior 38:21–37.

16.

Kern

Margaret L.

Friedman

Howard S.

Martin

Leslie R.

Reynolds

Chandra A.

Luong

Gloria

. 2009. “Conscientiousness, Career Success and Longevity: A Lifespan Analysis.” Annals of Behavioral Medicine 37:154–63.

17.

Liljegren

Mats

Ekberg

Kerstin

. 2008. “The Longitudinal Relationship Between Job Mobility, Perceived Organizational Justice, and Health.” BMC Public Health 8:164.

18.

Link

Bruce G.

Phelan

. 1995. “Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior 35:80–94.

19.

Mandal

Bidisha

Ayyagari

Padmaja

Gallo

William T.

2011. “Job Loss and Depression: The Role of Subjective Expectations.” Social Science and Medicine 72:576–83.

20.

Merton

Robert K.

1968. “The Matthew Effect in Science.” Science 159:56–63.

21.

Moen

Phyllis

. 2005. “Beyond the Career Mystique: ‘Time In,’ ‘Time Out,’ and ‘Second Acts.’” Sociological Forum 20:189–208.

22.

Pavalko

Eliza K.

Gong

Fang

Long

J. Scott

. 2007. “Women’s Work, Cohort Change, and Health.” Journal of Health and Social Behavior 48:352–68.

23.

Pavalko

Eliza K.

Smith

Brad

. 1999. “The Rhythms of Work: Health Effects of Women’s Work Dynamics.” Social Forces 77:1141–62.

24.

Pearlin

Leonard I.

Nguyen

Kim B.

Schieman

Scott

Milkie

Melissa A.

2007. “The Life-Origins of Mastery among Older People.” Journal of Health and Social Behavior 48:164–79.

25.

Rabe-Hesketh

Sophia

Skrondal

Anders

. 2008. Multilevel and Longitudinal Modeling Using Stata, 2nd ed. College Station, TX: StataCorp LP.

26.

Riley

Matilda W.

1987. “On the Significance of Age in Sociology.” American Sociological Review 52:1–14.

27.

Rosenfeld

Rachel A.

1992. “Job Mobility and Career Processes.” Annual Review of Sociology 18:39–61.

28.

Ross

Catherine E.

Mirowsky

John

. 1995. “Does Employment Affect Health?” Journal of Health and Social Behavior 36:230–43.

29.

Sampson

Robert J.

Laub

John H.

2005. “A Life-Course View of the Development of Crime.” Annals of the American Academy of Political and Social Science 602:12–45.

30.

Schafer

Markus H.

Ferraro

Kenneth F.

Mustillo

Sarah A.

2011. “Children of Misfortune: Early Adversity and Cumulative Inequality in Perceived Life Trajectories.” American Journal of Sociology 116:1053–91.

31.

Schieman

Scott

Whitestone

Yuko Kurashina

Van Gundy

Karen

. 2006. “The Nature of Work and the Stress of Higher Status.” Journal of Health and Social Behavior 47:242–57.

32.

Settersten

Richard A.

Jr . 2004. “Age Structure and the Rhythm of the Life Course.” Pp. 81–98 in Handbook of the Life Course, edited by Mortimer

J. T.

Shanahan

M. J.

New York: Springer.

33.

Singh-Manoux

Archana

Marmot

Michael G.

Adler

Nancy E.

2005. “Does Subjective Social Status Predict Health and Change in Health Status Better Than Objective Status?” Psychosomatic Medicine 67:855–61.

34.

Sobel

Michael E.

1982. “Asymptotic Confidence Intervals for Indirect Effects in Structural Equation Models.” Sociological Methodology 13:290–312.

35.

Stouffer

Samuel A.

Suchman

Edward A.

DeVinney

Leland C.

Star

Shirley A.

Williams

Robin M.

Jr . 1949. The American Soldier: Adjustment During Army Life (Studies in Social Psychology in World War II, Vol. 1). Princeton, NJ: Princeton University Press.

36.

Dejun

. 2009. “Occupational Career and Risk of Mortality among U.S. Civil War Veterans.” Social Science and Medicine 69:460–68.

37.

Swaen

Gerhard M. H.

Kant

IJmert

van Amelsvoort

Ludovic G. P. M.

Beurskens

Anna J. H. M.

2002. “Job Mobility, Its Determinants, and Its Effects: Longitudinal Data from the Maastricht Cohort Study.” Journal of Occupational Health Psychology 7:121–29.

38.

Thomas

William I.

Thomas

Dorothy S.

1928. The Child in America: Behavior Problems and Programs. New York: Knopf.

39.

U.S. Department of Labor. 2001. NLS of Mature Women User’s Guide. Washington, DC: Bureau of Labor Statistics and Center for Human Resource Research.

40.

U.S. Department of Labor. 2009. Consumer Price Index. Washington, DC: Bureau of Labor Statistics.