The Influence of Past Unemployment Duration on Symptoms of Depression Among Young Women and Men in the United States

Abstract

Objectives. I examined whether unemployment while looking for a job and being out of the labor force while not seeking work have distinct effects on symptoms of depression among young women and men in the United States. I also investigated whether past unemployment duration predicts depressive symptoms.

Methods. I used ordinary least squares regression to analyze data from the 1979–1994 National Longitudinal Survey of Youth.

Results. Cross-sectional results suggested that current unemployment status and out-of-the-labor-force status were significantly associated with depressive symptoms at ages 29 through 37 years. The association between being out of the labor force and depressive symptoms was stronger for men. Longitudinal results revealed that past unemployment duration across 15 years of the transition to adulthood significantly predicted depressive symptoms, net of demographics, family background, current socioeconomic status, and prior depressive symptoms. However, duration out of the labor force did not predict depressive symptoms.

Conclusions. Longer durations of unemployment predict higher levels of depressive symptoms among young adults. Future research should measure duration longitudinally and distinguish unemployment from being out of the labor force to advance our understanding of socioeconomic mental health disparities.

Numerous studies have documented a relationship between disadvantaged socioeconomic status (SES) and symptoms of depression.^1,2 The primary indicators of SES are education, employment status, income, occupation, and wealth.^1–3 What warrants further inquiry is whether unemployment when looking for a job and being “out of the labor force” when not seeking work have distinct effects on symptoms of depression⁴ and whether these effects vary by gender.⁵ Furthermore, longitudinal studies are necessary to understand the mental health consequences of long-term unemployment, because the majority of studies have measured SES at a single point in time.^1–6

Some groundbreaking longitudinal studies have demonstrated that past unemployment at multiple points in time predicts symptoms of depression. A Chicago study found that prior job disruption (being fired, laid off, downgraded, or leaving work because of illness) during a 4-year interval was associated with subsequent depressive symptoms.⁷ A Southeastern Michigan study discovered that unemployment at some time in the past 5 years was a risk factor for depression.⁸ According to a study of women in Sweden, 2 years of unemployment with no realistic expectation of getting a job was associated with depressive symptoms.⁹ Research that used data from the US National Longitudinal Survey of Mature Women revealed that being continuously out of the labor force from 1982 to 1989 predicted depressive symptoms in 1989, regardless of earlier emotional health in 1982.¹⁰ A US National Longitudinal Survey of Youth study found that changing from employment status in 1992 to unemployment status or out-of-the-labor-force status in 1994 was associated with depressive symptoms in 1994, net of prior depressive symptoms in 1992.⁴ In a comparison of average levels of depressive symptoms at ages 19 to 24 years, a South Australian study discovered that unemployment for 9 months or more was a significant threshold.¹¹ According to a meta-analysis of longitudinal studies, there is international evidence that changing from employment status to unemployment status measured between 6 months and 3 years leads to worse mental health in general.¹²

Taking into account earlier mental health acknowledges the selection hypothesis, which argues that mental health problems could lead to job loss or prolonged unemployment.^4,13 A study of young adults in New Zealand found that unemployment for greater than 6 months did not increase the odds of depression after the researchers controlled for earlier psychological adjustment problems.¹³ Longitudinal research on young adults in the United States is necessary to examine whether failure to find employment for longer durations is more psychologically distressing than being out of the labor force when not looking for a job, independent of earlier mental health.

Studying the influence of unemployment histories on mental health during the transition to adulthood is important because employment is a marker of adulthood,¹⁴ and early adulthood is the stage at which the onset of depression usually occurs.¹⁵ Moreover, whether the mental health ramifications of long-term unemployment and out-of-the-labor-force status differ for young women and men needs further investigation. Our knowledge remains limited about women's psychological reactions to the time spent unemployed when looking for work or out of the labor force while not seeking employment.^5,9,10

Using data from the National Longitudinal Survey of Youth, I examined whether the duration of past unemployment status and out-of-the-labor-force status across 15 years of the transition to adulthood predicts symptoms of depression among young women and men in the United States. To evaluate the strength of these relationships, I controlled for gender, age, race/ethnicity, marital status, family socioeconomic background, multiple indicators of current SES, and earlier depressive symptoms. Finally, I also investigated whether the influence of unemployment status and out-of-the-labor-force status on depressive symptoms varies by gender.

METHODS

I used data from the 1979 to the 1994 waves of the National Longitudinal Survey of Youth (NLSY), which is based on a nationally representative probability sample of young adults in the United States.¹⁶ The first wave of the NLSY was in 1979 and sampled 12 686 individuals aged 14 to 22 years. The NLSY respondents were interviewed annually from 1979 to 1994; after 1994, the interviews were conducted biennially. The respondents were aged 29 to 37 years in 1994. Members of the NLSY were born between January 1, 1957, and December 31, 1964, and are part of the younger Baby Boom cohort. The NLSY is sponsored by the Bureau of Labor Statistics of the US Department of Labor.

Missing Data

Compared with many other longitudinal surveys, the retention rate for the NLSY is high (95.7% for the year after the survey began and 89.2% for the 1994 wave). Because of attrition from the 1994 wave, 1050 respondents were excluded; 2 population subsamples (2923 total respondents from the military sample and the economically disadvantaged non-Black and non-Hispanic sample) were also excluded because the NLSY discontinued sampling them before the 1994 wave, decreasing the eligible sample size to 8713.

The data for the regression analyses were weighted to readjust for attrition using the 1994 sample weight created by the National Opinion Research Center.¹⁶ Mean imputation was used to replace the missing cases for family socioeconomic background (498 cases) and current education (90 cases). Because imputation techniques can produce biased results, sensitivity analyses confirmed that the results did not substantively vary when listwise deletion was used rather than mean imputation. I used listwise deletion for the missing cases from the other independent variables and the dependent variable. The final sample size was 8290 for the ordinary least squares regression models that examined the cross-sectional association between unemployment and depressive symptoms. The final sample size for the longitudinal analyses of the effect of past unemployment duration (1979–1993 waves) on depressive symptoms (1994 wave) was smaller (n = 6891) mainly because of attrition across multiple waves.

Depressive Symptoms

The dependent variable was the 7-item Center for Epidemiologic Studies Depression (CES-D) scale (1994 wave).¹⁷ The respondents were asked, “How often have you felt this way during the past week?”: (1) depressed, (2) sad, (3) everything was an effort, (4) you could not get going, (5) your appetite was poor, (6) you had trouble keeping your mind on what you were doing, and (7) you had restless sleep. The frequency of depressive symptoms ranged from 0 = rarely or none of the time or 1 day; 1 = some or a little of the time or 1 to 2 days; 2 = occasionally or a moderate amount of the time or 3 to 4 days; to 3 = most or all of the time or 5 to 7 days. The items were summed and coded so that higher values signified higher levels of depressive symptoms, yielding a range of 0 to 21. The internal consistency of the scale was high (Cronbach α = 0.81).

Duration of Unemployment

I measured duration unemployed and duration out of the labor force from the 1979–1993 waves. This time span was chosen to control for current unemployment status (1994 wave). The NLSY distinguishes unemployment status when seeking work from “out of the labor force” status when not seeking work. For each wave, dichotomous variables were created that coded unemployment as “1” if the respondent was unemployed when looking for a job. These variables were summed across the 15 waves to measure the number of years of unemployment. Dichotomous variables were created that coded “1” as “out of the labor force” for different reasons, such as keeping house, going to school, being unable to work because of long-term physical or mental illness, or for another reason. These variables were summed across the 15 waves to measure the number of years out of the labor force. The time spent unemployed or out of the labor force was not necessarily over consecutive years.

Socioeconomic Status

Current SES was measured in the 1994 wave. Two dichotomous variables measured unemployment status (unemployed = 1; out of the labor force = 1) with employed as the reference category. The number of years of education was assessed. Poverty was measured with a dichotomous variable (poverty = 1) to assess low income. Although there has been debate about the poverty threshold being set too low, poverty is an empirically clear-cut demarcation useful for policymaking.¹⁸ The poverty variable was created by the Center for Human Resource Research, based on annual household income, family size, and the yearly poverty guideline from the US Department of Health and Human Services.

Because income is a sensitive topic for survey respondents, 1878 respondents from the 1994 wave had missing values for poverty. Fortunately, respondents who had missing values for poverty answered a question about receipt of Aid to Families With Dependent Children, food stamps, supplemental security income, or other public or welfare assistance. Following the work of other scholars, I assigned “poverty status” to those respondents who received any of these forms of income assistance, and assigned respondents who had not received income assistance in the past year to the “not in poverty” category.¹⁹ Wealth was measured with net worth, which is considered to be one of the optimal indicators of economic well-being.^3,6 To determine net worth, respondents were asked: “If you and your spouse/partner were to sell all your major possessions (including your home), turn all of your investments and other assets into cash, and pay all of your debts, would you have money left over, break even, or be in debt?” Three dichotomous variables were created to measure net worth. “Break even” is zero net worth, “in debt” is negative net worth, and “have money left over” is positive net worth (reference category).

Mental health studies on young adults often measure SES with family background.^20–22 Family background (1979 wave) included parental education and parental occupation. The respondents were asked the highest grade or year of school completed by their mothers and fathers. The highest level achieved by either parent was used. The respondents were also asked about the occupational status of the adult male and female in the household when they were aged 14 years. I matched the Hodge-Siegel-Rossi occupational prestige scale with the 3-digit 1970 US Census Bureau occupational classifications.²³ The median level of parental occupational prestige was 35, which was equal to occupations in sales.

Demographics

Demographic variables included gender, age, race/ethnicity, and marital status. Gender was a dichotomous variable (female = 1). Age ranged from 29 to 37 years (1994 wave). Dichotomous variables measured race and ethnicity (Black = 1; Hispanic = 1) with White as the reference category. The sample identification code variable (1979 wave) was used, which identified NLSY subsamples by Black, Hispanic, and White respondents. This is the standard variable used by NLSY researchers to measure race and ethnicity.¹⁶ Dichotomous variables for marital status were constructed from the 1994 wave (previously married = 1; never married = 1) with married as the reference category.

Prior Depressive Symptoms

Prior depressive symptoms were measured using the 20-item CES-D scale (1992 wave).¹⁷ The first NLSY wave to measure depression was 1992, and depression was not measured in 1993. The frequency of depressive symptoms ranges from 0 = rarely or none of the time or 1 day; 1 = some or a little of the time or 1 to 2 days; 2 = occasionally or a moderate amount of the time or 3 to 4 days; to 3 = most or all of the time or 5 to 7 days. Items were summed and coded so that higher values signified higher levels of depressive symptoms, yielding a range of 0 to 60 (Cronbach α = 0.88).

RESULTS

Table 1 presents the means, standard deviations, and ranges for the variables. As a sensitivity analysis, Table 1 compares the descriptive statistics for the sample (n = 8290) in column 1 used for the cross-sectional analyses to the smaller sample (n = 6891) in column 2 used for the longitudinal analyses. The levels of depressive symptoms were within the range to be expected for a general population.¹⁷ According to the first column of Table 1, 5% of the sample was unemployed when seeking work, 17% was out of the labor force when not looking for a job, and 78% was employed in 1994 at age 29–37 years. The second column shows that the average unemployment duration was 1.47 years (women = 1.39 years; men = 1.55 years), with a minimum of 0 years and a maximum of 13 years. The average duration out of the labor force was 3.21 years (women = 4.25 years; men = 2.02 years), with a range of 0 to 15 years. Rates of unemployment from 1979 to 1993 are displayed in Table 2. In general, rates of out-of-the-labor-force status decreased more over time than did unemployment.

TABLE 1.

Sample Descriptive Statistics, by Analytical Method: National Longitudinal Survey of Youth, 1979–1994

Variables	Cross-Sectional Analyses Sample,a Mean (SD)	Longitudinal Analyses Sample,b Mean (SD)	Response Range
Past unemployment (1979–1993 waves)
Duration unemployed, y		1.47 (1.78)	0–13
Duration out of the labor force, y		3.21 (3.42)	0–15
Socioeconomic status (1994 wave)
Unemployed	0.05 (0.23)	0.05 (0.22)	0–1
Out of the labor force	0.17 (0.37)	0.16 (0.36)	0–1
Employed	0.78 (0.41)	0.79 (90.4)	0–1
Years of education	12.93 (2.42)	13.06 (2.39)	0–20
Poverty statusc	0.16 (0.37)	0.16 (0.36)	0–1
Zero net worthd	0.21 (0.41)	0.21 (0.40)	0–1
Negative net worthd	0.13 (0.34)	0.13 (0.33)	0–1
Positive net worthd	0.66 (0.47)	0.66 (0.47)	0–1
Family background (1979 wave)
Years of parental education	11.65 (3.39)	11.74 (3.34)	0–20
Parental occupatione	33.94 (18.5)	34.39 (18.5)	0–86
Demographics
Female	0.52 (0.50)	0.53 (0.50)	0–1
Age in y (1994 wave)	33.00 (2.23)	33.00 (2.23)	29–37
Black	0.30 (0.46)	0.30 (0.46)	0–1
Hispanic	0.19 (0.39)	0.18 (0.38)	0–1
White	0.51 (0.50)	0.52 (0.50)	0–1
Previously married (1994 wave)	0.18 (0.39)	0.18 (0.38)	0–1
Never married (1994 wave)	0.26 (0.44)	0.26 (0.44)	0–1
Married (1994 wave)	0.56 (0.50)	0.56 (0.50)	0–1
Depressive symptoms
20-item CES-D (1992 wave) score	9.75 (9.25)	9.67 (9.13)	0–60
7-item CES-D (1994 wave) score	3.80 (4.10)	3.75 (4.06)	0–21

Note. CES-D = Center for Epidemiologic Studies Depression scale.

^aFor the ordinary least squares regression models that examined the cross-sectional association between unemployment and depressive symptoms, the sample size was n = 8290.

^bFor the longitudinal analyses of the influence of past unemployment duration (1979–1993 waves) on depressive symptoms (1994 wave), the sample size was n = 6891, mainly because of attrition across multiple waves.

^cPoverty level determined by the US Department of Health and Human Services.

^dNet worth was determined by asking: “If you and your spouse/partner were to sell all your major possessions (including your home), turn all of your investments and other assets into cash, and pay all of your debts, would you have money left over, break even, or be in debt?”

^eMeasured by Hodge-Siegel-Rossi prestige scale.

TABLE 2.

Rates of Unemployment: National Longitudinal Survey of Youth (NLSY), 1979–1993

NLSY wave	Age, Range	% Unemployed	% Out of the Labor Force
1979	14–22 y	15.57	39.67
1980	15–23 y	15.28	33.36
1981	16–24 y	14.56	29.28
1982	17–25 y	14.09	27.37
1983	18–26 y	14.47	24.39
1984	19–27 y	11.49	20.69
1985	20–28 y	10.59	18.28
1986	21–29 y	9.03	16.89
1987	22–30 y	6.57	16.22
1988	23–31 y	5.78	15.93
1989	24–32 y	5.85	16.21
1990	25–33 y	5.02	15.54
1991	26–34 y	6.40	15.85
1992	27–35 y	6.33	15.12
1993	28–36 y	5.65	16.06

Note. N = 6891.

In Table 3, cross-sectional ordinary least squares regression results for model 1 suggested that compared with being currently employed, unemployment status (b = 1.42; P < .001) and out-of-the-labor-force status (b = 1.03; P < .001) were significantly associated with depressive symptoms at age 29–37 years, net of current SES, family background, and demographics. In model 2, the effects of unemployment (b = 1.09) and out-of-the-labor-force status (b = 0.80) were reduced, but maintained their statistical significance (P < .001) when prior depressive symptoms were controlled. In model 3, a significant interaction effect was observed (b = –1.39; P < .01), signifying that out-of-the-labor-force status had a stronger association with depressive symptoms among men. Supplementary analyses indicated that 23.9% of women compared with only 8.7% of men were out of the labor force.

TABLE 3.

Ordinary Least Squares Regression Models of Socioeconomic Status, Family Background, Demographics, and Prior Depressive Symptoms Predicting Depressive Symptoms at Ages 29 to 37 Years (1994 Wave): National Longitudinal Survey of Youth, 1979, 1992, and 1994

Variables	Model 1, b (SE)	Model 2, b (SE)	Model 3, b (SE)
Socioeconomic status (1994 wave)
Unemployeda	1.42 (0.30)***	1.09 (0.28)***	1.20 (0.43)**
Out of the labor forcea	1.03 (0.18)***	0.80 (0.16)***	1.85 (0.38)***
Education	−0.18 (0.02)***	−0.11 (0.02)***	−0.11 (0.02)***
Poverty statusb	1.11 (0.20)***	0.73 (0.18)***	0.70 (0.18)***
Zero net worthc	0.30 (0.15)*	0.04 (0.14)	0.03 (0.14)
Negative net worthc	1.74 (0.19)***	1.24 (0.18)***	1.23 (0.18)****
Family background (1979 wave)
Parental education	0.01 (0.02)	0.01 (0.02)	0.01 (0.02)
Parental occupation	−0.01 (0.00)	−0.01 (0.00)	−0.01 (0.00)
Demographics
Female	1.12 (0.10)***	0.88 (0.10)***	1.04 (0.10)***
Age, y	0.01 (0.02)	−0.00 (0.02)	−0.00 (0.02)
Blackd	−0.20 (0.12)	−0.31 (0.12)**	−0.32 (0.12)**
Hispanicd	−0.05 (0.14)	−0.14 (0.13)	−0.14 (0.13)
Previously marriede	1.04 (0.17)***	0.57 (0.15)***	0.53 (0.15)**
Never marriede	0.45 (0.13)***	0.18 (0.12)	0.12 (0.12)
Prior depressive symptomsf (1992 wave)		0.16 (0.01)***	0.16 (0.01)***
Female × unemployed			−0.18 (0.56)
Female × out of the labor force			−1.39 (0.41)**
Intercept	4.49	2.90	2.82
R2	0.13	0.24	0.24

Notes. For these analyses, the sample size is n = 8290. Data are weighted.

^aReference category is employed.

^bPoverty level as determined by the US Department of Health and Human Services.

^cReference category is positive net worth. Net worth was determined by asking: “If you and your spouse/partner were to sell all your major possessions (including your home), turn all of your investments and other assets into cash, and pay all of your debts, would you have money left over, break even, or be in debt?”

^dReference category is White.

^eReference category is married.

^f20-item Center for Epidemiologic Studies Depression Scale.

*P < .05; **P < .01; ***P < .001 (2-tailed tests).

In Table 4, longitudinal analyses for model 1 showed that the duration (1979–1993 waves) of past unemployment (b = 0.13; P < .01) significantly predicted depressive symptoms (1994 wave), with adjustment for current SES, family background, and demographics. However, the influence of the duration of being out of the labor force on depressive symptoms was not statistically significant. Supplementary analyses (available on request) of variance inflation factors confirmed that multicollinearity was not an issue for the models that included multiple socioeconomic indicators.

TABLE 4.

Ordinary Least Squares Regression Models of Unemployment Duration, Socioeconomic Status, Family Background, Demographics, and Prior Depressive Symptoms Predicting Depressive Symptoms at Ages 29 to 37 Years (1994 Wave): National Longitudinal Survey of Youth, 1979–1994

Variables	Model 1, b (SE)	Model 2, b (SE)	Model 3, b (SE)
Past unemployment (1979–1993 waves)
Duration unemployed	0.13 (0.04)**	0.08 (0.03)*	0.06 (0.05)
Duration out of the labor force	0.01 (0.02)	−0.01 (0.02)	−0.01 (0.03)
Socioeconomic status (1994 wave)
Unemployeda	1.09 (0.33)**	0.85 (0.31)**	0.85 (0.31)**
Out of the labor forcea	0.87 (0.21)***	0.72 (0.20)***	0.72 (0.20)***
Education	−0.17 (0.03)***	−0.11 (0.02)***	−0.11 (0.02)***
Poverty statusb	0.93 (0.23)***	0.68 (0.21)**	0.67 (0.21)**
Zero net worthc	0.31 (0.17)	0.06 (0.16)	0.06 (0.16)
Negative net worthc	1.71 (0.21)***	1.22 (0.20)***	1.23 (0.20)***
Family background (1979 wave)
Parental education	0.01 (0.02)	0.01 (0.02)	0.01 (0.02)
Parental occupation	−0.01 (0.00)	−0.01 (0.00)	−0.01 (0.00)
Demographics
Female	1.14 (0.11)***	0.93 (0.11)***	0.87 (0.15)***
Age, y	0.02 (0.02)	0.00 (0.02)	−0.00 (0.02)
Blackd	−0.33 (0.14)*	−0.41 (0.13)**	−0.41 (0.13)**
Hispanicd	−0.12 (0.15)	−0.23 (0.14)	−0.23 (0.14)
Previously marriede	1.14 (0.19)***	0.65 (0.17)***	0.65 (0.17)***
Never marriede	0.40 (0.14)**	0.14 (0.13)	0.15 (0.13)
Prior depressive symptomsf (1992 wave)		0.16 (0.01)***	0.16 (0.01)***
Female × duration unemployed			0.04 (0.06)
Female × duration out of labor force			0.01 (0.04)
Intercept	3.81	2.69	2.73
R2	0.13	0.23	0.23

Notes. For these analyses, the sample size is n = 6891. Data are weighted.

^aReference category is employed.

^bPoverty level as determined by the US Department of Health and Human Services.

^cReference category is positive net worth. Net worth was determined by asking: “If you and your spouse/partner were to sell all your major possessions (including your home), turn all of your investments and other assets into cash, and pay all of your debts, would you have money left over, break even, or be in debt?”

^dReference category is White.

^eReference category is married.

^f20-item Center for Epidemiologic Studies Depression Scale.

*P < .05; **P < .01; ***P < .001 (2-tailed tests).

Model 2 (Table 4) revealed that the significant effect of unemployment duration (1979–1993 waves) on depressive symptoms (1994 wave) was independent of prior depressive symptoms (1992 wave). The magnitude of the effect was reduced in model 2 (b = 0.08; P < .05) compared with model 1 (b = 0.13; P < .01) but continued to be statistically significant. As mentioned previously, depressive symptoms were first measured in the 1992 wave. With regard to earlier mental health, supplementary analyses confirmed that the influence of unemployment duration (1979–1993 waves) on depressive symptoms (1994 wave) maintained its statistical significance when self-esteem at age 15 to 23 years (1980 wave) was controlled. Consistent with other research on mental health, self-esteem had a robust (P < .001) inverse relationship with depressive symptoms.⁷ In addition, model 3 indicated that the influence of past unemployment duration on depressive symptoms did not significantly vary by gender.

Subsidiary analyses showed that the strength of the effect of past unemployment duration on depressive symptoms was maintained when age at departure from the parental home was controlled. A large proportion of the sample was unemployed (15.57%) and out of the labor force (39.67%) at age 14 to 22 years (Table 2), which is to be expected for those who are younger and living with their parents. There were noticeable drops in these rates (unemployed = 11.49% and out of the labor force = 20.69%) at age 19 to 27 years, which likely represents younger respondents leaving the parental home and entering the labor force after age 18. Finally, to address whether the mental health consequences of unemployment duration vary by age,²⁴ supplementary analyses confirmed that the influence of unemployment duration on depressive symptoms did not significantly interact with age.

DISCUSSION

Unemployment is a major economic problem, yet little is known about its long-term costs to mental and physical health, and the extent of its impact as a public health problem.^5,25,26 Rather than measuring unemployment as a dichotomous comparison with those who are currently working, it is important to distinguish unemployment when seeking work from being out of the labor force when not looking for a job.⁴ I advance previous research by analyzing longitudinally whether the duration of unemployment and out-of-the-labor-force status across 15 years of the transition to adulthood predicts symptoms of depression.

Cross-sectional results suggested that, compared with being currently employed, unemployment status and out-of-the-labor-force status were significantly associated with depressive symptoms at ages 29–37 years, independent of demographics, current SES, family background, and prior depressive symptoms. Yet causal inferences cannot be made about the direction of cross-sectional relationships. Cross-sectional analyses of interaction effects further suggested that the relationship between out-of-the-labor-force status and depressive symptoms was stronger for men. Because it is more common for women to be out of the labor force, it is plausible that it may be less stigmatizing and psychologically distressing for them compared with men.

In exploring temporal ordering, longitudinal analyses revealed that past unemployment duration (1979–1993 waves) significantly predicted depressive symptoms (1994 wave), net of demographics, current SES, family background, and prior depressive symptoms. In contrast to the cross-sectional findings, the time spent out of the labor force did not have a significant influence on subsequent depressive symptoms. Furthermore, the influence of the duration of unemployment status or out-of-the-labor-force status on depressive symptoms did not significantly vary by gender. An important topic for future research is whether symptoms of depression among wives who are in the labor force or out of the labor force are affected by their husbands' duration of unemployment status or out-of-the-labor-force status.

Examining the depressive effect of unemployment duration contributes to research on chronic stress, which has demonstrated that the duration of stress is crucial for understanding the extent of its harmful mental health effects.^7,27,28 The economic hardship associated with unemployment when looking for work during young adulthood can be chronically stressful even if it is intermittently experienced, which has been referred to as “living on the edge.”^27(p313) In my study, the years of unemployment were not necessarily consecutive and unemployment was measured annually. To better assess a chronic experience, future longitudinal research should measure consecutive years and continuous unemployment within each year. Longitudinal studies on mental health should further explore the timing of unemployment entries and exits for different cohorts,¹⁰ and “underemployment,” which refers to poverty-level wages and involuntary part-time work.⁴

The NLSY focuses on a young adult cohort of the Baby Boom generation. Thus, my findings could be specific to this life stage or historical period. The economic recessions in the early 1980s and 1990s could have impacted the rates or durations of unemployment and psychological well-being of this cohort.^29,30 An advantage of the NLSY data for the purpose of my study is the measurement of a long duration (1979–1993 waves) of both unemployment status and out-of-the-labor-force status. A disadvantage of the NLSY is that the first measure of depression was in the 1992 wave. Measuring depression earlier would have better tested mental health selection effects.

However, with regard to the possibility of depression prolonging unemployment, those respondents who were unable to work because of mental illness were included in the NLSY category of out-of-the-labor-force status. Interestingly, results showed that the duration out of the labor force (1979–1993 waves) did not significantly predict depressive symptoms (1994 wave). Furthermore, supplementary analyses indicated that unemployment duration (1979–1993 waves) significantly predicted depressive symptoms (1994 wave) net of prior self-esteem (1980 wave), which is a conservative proxy for mental health at ages 15 to 23 years.

Future longitudinal research should continue to investigate the temporal ordering of the relationship between unemployment and depression. Another study that used data from the NLSY discovered that depressive symptoms (1992 wave) significantly increased the odds of changing from employment status to unemployment status (1994 wave), but did not predict being out of the labor force (1994 wave).⁴ Other research has demonstrated that clinical depression does not lead to unemployment,³¹ and that elevated levels of depressive symptoms increased the probability of reemployment.³² Overall, research on the causation–selection issue suggests that the selection process may be stronger for schizophrenia and conduct disorder than for depression and, thus, depressive symptoms may be less likely to lead to unemployment and downward socioeconomic mobility.^20,33

In conclusion, longer durations of unemployment predict symptoms of depression among young adults in the United States. Research on risk factors for depression earlier in the life course can concomitantly add to our understanding of the origins of physical health disparities. Studies have shown that depressive symptoms in early adulthood can predict trajectories of physical health problems in later life.^22,34 Therefore, medical interventions, social welfare initiatives, and public health policies targeted to counteract spells of unemployment and protect mental health during the transition to adulthood could ultimately improve population health in the future. More longitudinal studies are necessary to capture socioeconomic disadvantages at multiple points in time to advance our knowledge of the lasting influence of past social conditions on mental health in the United States and internationally.^{3,6,19,35–37}