| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Department of Epidemiology and Public Health, Yale University School of Medicine
William T. Gallo, Yale University School of Medicine, Department of Epidemiology and Public Health, 60 College Street, New Haven, CT 06520-8034 E-mail: william.gallo{at}yale.edu.
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Using longitudinal data from the 1992 and 1994 waves of the Health and Retirement Survey, multivariate regression models were estimated to assess the impact of involuntary job loss on both physical functioning and mental health. Our analysis sample included 209 workers who experienced involuntary job loss between survey dates and a comparison group of 2,907 continuously employed workers.
Results. The effects of late-life involuntary job loss on both follow-up physical functioning and mental health were negative and statistically significant (p < .05), even after baseline health status and sociodemographic factors were controlled for. Among displaced workers, reemployment was positively associated with both follow-up physical functioning and mental health, whereas the duration of joblessness was not significantly associated with either outcome.
Discussion. The findings provide evidence of a causal relationship between job loss and morbidity among older workers. This relationship is reflected in both poorer physical functioning and mental health for workers who experience involuntary job loss. In addition to the economic consequences of worker displacement, there may be important health consequences of job loss, especially among older workers.
IN the United States, downsizing led to more than 10 million workers losing their jobs between 1989 and 1992 (Cascio 1998
). According to the United States Labor Department, Bureau of Labor Statistics, 43 million jobs were eliminated between 1979 and 1995 (Uchitelle and Kleinfield 1996). Although the overall rate of worker displacement has not changed appreciably in the past two decades, the composition of those suffering job loss has shifted. Whereas displacement was once concentrated among younger workers, its incidence is now greatest among older workers (Couch 1998; Farber 1993). Findings such as these indicate that involuntary job loss and workplace uncertainty are common experiences for American workers, particularly older workers. Although the economic impact of job displacement has been widely studied, the potential health effects of job loss are less well-understood, despite substantial research in this area.
Understanding the health consequences of job loss is important to discerning the complete impact of economic downturns, in which worker displacements are common, especially among older workers. In addition, if involuntary job loss has a deleterious effect on subsequent health, then appropriate services for displaced workers can be designed to address anticipated health needs during and after job separation.
There are a number of reasons why older U.S. workers may be particularly vulnerable to health problems following job loss. First, older workers commonly suffer longer jobless spells than do younger workers (Swaim and Podgursky 1991), forfeiting salaries and other benefits of permanent employment. In addition, the skills of older workers may not be transferable to new positions so that, frequently, displaced older workers experience substantial earnings losses in their postseparation employment (Carrington 1993; Carrington and Zaman 1994; Couch 1998; Farber 1993; Jacobson, LaLonde, and Sullivan 1993; Kletzer 1989; Neal 1995; Ong and Mar 1992; Ruhm 1990, Ruhm 1991). Finally, American workers accumulate a significant proportion of the wealth that will finance their retirements in the decade preceding retirement (Mitchell and Moore 1998). Involuntary job loss and prolonged unemployment in this period may therefore have a particularly devastating impact on economic well-being, and consequently on emotional and physical well-being.
Existing literature on the health consequences of job loss can be categorized into two classes of studies: cross-sectional research and longitudinal analyses. Spurred by the dramatic increase in unemployment in the 1970s and 1980s, many cross-sectional studies have identified an association between unemployment and various indicators of poor health, including minor psychiatric morbidity, major affective disorders, and a range of chronic physical health problems, as well as mortality (Arber 1987; Cook, Cummins, Bartley, and Shaper 1982; D'Arcy 1986; D'Arcy and Siddique 1985; Kasl and Jones in press; Mathers and Schofield 1998). Although these studies have provided reasonably consistent evidence of an association between unemployment and poorer health, they have frequently been unable to control adequately for health status prior to job loss. Thus, these studies have not clearly established whether those in poorer health are more likely to lose their jobs or, conversely, whether job loss itself leads to poorer health, or both (Kasl and Jones in press).
Longitudinal studies, which provide better methodology for assessing the direction of causality, have become reasonably common during the past decade. These studies have generally shown that job loss has some negative impact on physical health, although the nature of this impact is difficult to assess (Kasl and Jones in press), given the diversity of physical health measures considered and the sensitivity of results to the measures used. With reasonable agreement, however, the longitudinal studies have demonstrated a negative effect of job loss on a range of measures likely related to physical health, including the number of reported medical conditions, rates of medical services utilization, and pension disability use (Ferrie, Shipley, Marmot, Stansfeld, & Smith, 1998; Hamilton, Broman, Hoffman, and Renner 1990; Hamilton, Hoffman, Broman, and Rauma 1993; Kasl and Jones in press; Morris and Cook 1991; Westin 1990). In terms of mental health, longitudinal data analysis has strongly supported the hypothesis that job loss has an adverse effect on symptoms of poorer mental health (Brenner and Starrin 1988; Frese and Mohr 1987; G.A. Kaplan, Roberts, Camacho, and Coyne 1987; Warr, Jackson, and Banks 1988).
Studies of plant closings are potentially the best of the longitudinal designs. Even so, limitations of these studies have been described (Morris and Cook 1991). The populations studied tend to be fairly homogeneous, generally male, manual workers, and the control groups used for dislocated workers (frequently displaced from a single plant) are often not directly comparable because members of comparison groups are commonly selected from a different plant. Existing studies of plant closings have involved small numbers of subjects, and thus lack adequate statistical power to detect significant effects. Further, the vast majority have been conducted outside the United States, in Australia and the United Kingdom. The one previous study to our knowledge that used a national sample of American workers (Sorlie and Rogot 1990) focused on mortality, rather than physical morbidity or mental health. Given the differences in social, economic, and health programs, the observed effects of job loss in other countries may not be generalizable to the United States. Finally, we could find few studies that investigated the effect of job loss on the health of older workers (Frese and Mohr 1987; Warr and Jackson 1984, Warr and Jackson 1987; Warr et al. 1988).
The first of these studies (Frese and Mohr 1987) was conducted during 1975–1977 in Germany, and included 51 workers over 45 years of age all of whom lost their jobs, with no control group. The authors reported increased depression and reduced hope over time among workers in their sample. However, the validity and generalizability of the findings are limited by the study's methodological shortcomings.
The second set of these studies (Warr and Jackson 1984, Warr and Jackson 1987; Warr et al. 1988) examined 954 men aged 16 to 64 years who were continuously unemployed for 25 months in Britain. These studies found that, among continuously unemployed men, the mental health effects of unemployment vary in a curvilinear fashion with age. The youngest (aged 16–19 years) and oldest (aged 60–64 years) groups had significantly better mental health than did the remaining groups. Further, the middle groups (aged 20–59 years) showed significantly less improvement over time in mental health than did the younger or older groups (Warr and Jackson 1987). The authors suggest that the observed age-related variations in responses to unemployment may be due to differing role responsibilities throughout the life cycle. For instance, older workers approaching retirement may have fewer financial commitments and less family responsibility than do middle-aged workers. Older workers may therefore face less pressure to become reemployed, and may experience less stress by remaining unemployed than younger workers (Warr et al. 1988). Nonetheless, this investigation did not use a comparison group to contrast the unemployed men, which is necessary to describe the effects of job loss itself.
Given the limited number of studies on older workers, the health consequences of late-life job loss in the United States remain largely unknown. The objective of this research is therefore to estimate the health effects of involuntary job loss among older workers in the United States, using longitudinal data on a nationally representative sample of older Americans. This research examines the effect of involuntary job loss on measures of both physical and mental health. Unlike previous studies, this study includes an adequate comparison group of older workers, and therefore, is able to effectively control for baseline measures of physical functioning and mental health.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
).
Sample
HRS data are collected at 2-year intervals. At the baseline (1992), HRS participants included 12,652 individuals from 7,702 households (
). In-home, face-to-face interviews were taken from a random sample of individuals born between 1931 and 1941 (aged 51 to 61 years at the baseline interview) and their spouses. Blacks, Hispanics, and Florida residents were oversampled. In 1994, 11,602 respondents from 7,093 households were reinterviewed by mail or telephone (
). HRS data collection is conducted by the Institute for Social Research at the University of Michigan. The primary funding organization of the Survey is the National Institute on Aging.
For purposes of this analysis, the HRS sample was initially restricted to include all workers who were currently working at the 1992 baseline survey and had been working without interruption for at least 3 years with the same employer (
). By selecting individuals who had been continuously employed, we hoped to eliminate workers with less stable work histories. Such an approach is consistent with prior studies of job loss (Couch 1998; Jacobson et al. 1993) and helps to reduce unobserved heterogeneity in the sample that might be related to both future job loss and future health status. Neglecting to account for such differences could bias estimates of the effect of job loss on postseparation health.
To construct the analysis sample, we created a continuous monthly record of employment between interview dates for the 4,792 workers eligible for the analysis, using retrospective data collected at the 1994 survey. We next investigated employment information for workers with intervals of unemployment. Respondents reporting job exit between survey dates due to plant closure or layoff were characterized as experiencing involuntary job loss. Our sample was then limited to workers who were either continuously employed or had experienced involuntary job loss (
). Respondents who reported job departures for other reasons (
3) at the follow-up interview were excluded from the analysis sample. Other reasons for job exit and their frequencies were disability (
), family reasons (
), better job (
), quit (
), and retirement (
). In addition, we excluded 239 respondents reporting a temporary job exit. These were individuals who not only reported working for their Wave 1 (1992) employer at follow-up, but also reported some months of nonemployment between surveys. The subsample of retirees () was retained for sensitivity analysis.
Of the 3,939 possible participants, we eliminated 595 because of missing data, 96 self-employed individuals, and 132 age-ineligible spouses, leaving 3,116 individuals in our analysis sample. In all, 209 of these workers experienced involuntary job loss. The remaining 2,907 continuously employed individuals comprise our comparison group. Of the 209 workers who experienced involuntary job loss, 101 were reemployed by the time of the 1994 survey. Two of the 209 displaced workers in our analysis sample did not provide sufficient information for us to calculate the unemployment spell length. These two observations were not included in the estimation of the effect of unemployment duration on follow-up health.
To investigate whether there were systematic differences between respondents excluded because of missing data (
) and individuals in our analysis sample (
), we used differences of group means and t tests. The results of these analyses suggest that the respondents with missing data differed from the individuals in our analysis sample with regard to a number of characteristics. Significant differences (p < .05) were detected in race, occupation class, and baseline health. Respondents with missing data were more likely to be non-White, blue-collar, and in poorer physical and mental health at the baseline than members of the analysis group.
All analyses were conducted using weighted data to correct for the HRS oversampling of relevant subgroups. Data were weighted by the person-level analysis weight provided by the HRS.
Dependent Variables
Physical functioning was measured using an index based on responses to 15 performance-based, functional status questions. These questions include self-reported performance of Activities of Daily Living (ADLs) and mobility tasks. Activities used to construct the physical functioning index included the following: running or jogging a mile; walking several blocks; walking one block; walking across a room; climbing several flights of stairs; climbing one flight of stairs; getting up from a chair; getting in and out of bed; bathing; dressing; eating; lifting or carrying over 10 pounds; stooping, kneeling or crouching; pulling or pushing large objects; and extending arms above shoulders. Because response options for these items varied between the baseline and follow-up HRS interviews, comparable measurement of the level of difficulty in performing each task was not possible unless items were dichotomized to identify respondents with difficulty performing a particular activity (
). Dichotomized values were summed to calculate a single functioning score, ranging from 0 to 15, with higher values reflecting poorer physical functioning.
Mental health was measured using a depression score based on 8 items from the 20-item Center for Epidemiological Studies-Depression (CES-D) scale. Although the full scale would be preferable, the HRS only includes these eight items in both survey waves. We used all available data. Of the eight CES-D items, six reflect the presence of depression (in the last week, respondent felt depressed; felt everything he or she did was an effort; experienced restless sleep; "could not get going"; felt lonely; felt sad), and two suggest its absence (in the last week, respondent enjoyed life; was happy). For each of the former six statements, responses were dichotomized (the symptom was experienced "much," "most," or "all" of the time vs. "little" or "none" of the time). For the latter two statements, responses were reverse coded, yielding an index, ranging from 0 to 8, with higher scores reflecting poorer mental health. Dichotomizing response categories was again necessary because only two categories of response (symptom was experienced "most" vs. "none" of the time) were used in the 1994 follow-up survey.
We assessed the reliability of the 15-item physical functioning score and the 8-item depression score using the Cronbach alpha coefficient (Cronbach 1951). Cronbach alpha coefficients for both baseline and follow-up physical functioning scores were .74. For baseline and follow-up depression scores, the Cronbach alpha coefficients were adjusted for the more limited test length of the CES-D included in the HRS (8 questions vs. 20 questions in the original CES-D) using the approach recommended by Nunnally 1967(p. 223). The adjusted coefficients were .82 and .88, respectively. These are consistent with the alpha coefficients reported for the full CES-D in several previous studies (Radloff 1977). In each case, the Cronbach values suggest adequate internal consistency among the elements comprising the dependent variable scales.
We used exploratory factor analysis to evaluate the factor structure of our dependent variables. As our physical functioning score consists of a combination of mobility tasks and more traditional ADLs, our approach was first to explore the possibility that two distinct conceptual dimensions existed within the 15-item score. We therefore computed factor loadings using maximum likelihood estimation, constraining the number of factors to two. However, examination of factor loadings revealed no obvious factor structure. Such was the case for both oblique and orthogonal factor analyses. We then restricted the number of factors to one and reestimated the loadings. Factor loadings supported the premise that the 15 items comprising the physical functioning score are appropriately aggregated into a single dimension.
A similar procedure was followed to investigate the validity of the 8-item depression score. Once more we observed no apparent factor structure from fitting the two-factor solution. Such findings indicate one underlying concept present in the depression score. This is not surprising when one considers that the elements comprising our mental health measure were drawn from a well-established empirical tool designed specifically to measure the phenomenon of depression.
The use of a self-reported measure of physical functioning and a depression battery, rather than a single, global measure of self-assessed physical or mental health, as the basis for estimating health changes associated with job loss has empirical support. Numerous studies provide evidence that jobless workers justify their labor market status by understating their health in self-reports (see Chirikos, 1993). Hence, using self-assessed health, rather than the more detailed measures used in this study, could potentially result in an overstatement of the impact of job loss in our empirical analysis. In addition, it is less clear what the global assessment of health measures. Given the interest in physical function and depression, using scales designed to measure these more specific constructs, rather than overall health, is appropriate.
Independent Variables
Independent variables were chosen based on previous literature suggesting their importance in the estimation of physical disability and/or mental health and the primary research question concerning involuntary job loss.
The primary independent variable of interest was a dummy variable for involuntary job loss (i.e., job exit due to either plant/company closing or layoff). The time component of involuntary job loss was assessed using a continuous variable that captured the months elapsed between job loss and reemployment for the reemployed, and between job loss and the follow-up survey date for workers not reemployed. For analyses of the effects of reemployment on physical functioning and mental health, a dummy variable for reemployment was used. Those workers reemployed included all previously displaced workers who were working by the time of the follow-up interview, whether the new position was full-time or part-time.
Because baseline health was likely to be strongly associated with health 2 years later, we included as independent variables a number of measures of baseline health. Baseline physical functioning (in the regression estimating postloss physical functioning) and baseline mental health (in the regression estimating postloss mental health) were each included as covariates. Although it is plausible that physical functioning and mental health may be related, we found no evidence that baseline mental health predicted follow-up physical functioning or that baseline physical function predicted follow-up mental health. Therefore, we did not include baseline mental health in the estimation of physical functioning; nor did we include baseline physical functioning in the estimation of follow-up mental health. Their removal from the respective models did not materially change the parameter estimates on the remaining independent variables.
In addition to baseline physical functioning, a number of other baseline physical health measures known from previous literature to be predictive of various health outcomes, including physical functioning, were included in the models of post-loss physical functioning. This set of covariates included (a) obesity, which was determined as a body mass index of at least 30, the most recent criterion for obesity in this age group (Flegal, Carroll, Kuczmarski, and Johnson 1998), (b) whether a physician had ever told the respondent that he or she had high blood pressure or hypertension, heart disease, or cancer or a malignant tumor, (c) heavy smoking (at least 20 cigarettes per day), and (d) heavy drinking (at least 3 drinks per day). The effect of these covariates on mental health is less clear from previous literature. In addition, these covariates were not associated with follow-up mental health in this sample; their removal from the mental health estimation model did not materially change the parameter estimates on the remaining independent variables, so for both theoretical and empirical reasons, they were not included in the models of follow-up mental health.
We also selected several socioeconomic variables as covariates in our empirical models. Their selection was based on a large body of literature from medical sociology demonstrating the influence of socioeconomic factors on health (H.B. Kaplan 1989; Link and Dohrenwend 1989). The socioeconomic variables included age, gender, years of education, race (White vs. non-White), marital status (married vs. unmarried), occupation class (blue collar vs. white collar), income, and assets. Blue-collar occupations included farming, forestry, and fishing; production and operations; and the military. White-collar occupations included managerial and professional, sales, clerical, administrative, and service occupations. Income included pretax individual labor earnings for 1991. Assets included household nonhousing asset amounts reported by respondents at the 1992 survey. Nonhousing assets included 1992 values of checking and savings accounts; certificates of deposit, bonds, and Treasury bills; individual retirement accounts; stock and mutual funds; business equity; equity in real estate other than respondent's primary assets; and other reported nonhousing assets. Continuous predictor variables were dichotomized in the bivariate analyses.
Estimation Model and Empirical Approach
We observe health for all workers in our analysis sample at two fixed points in time, t and t+k. Let s represent the time at which a subset of the workers observed at time t suffer involuntary job loss, and let us restrict s so that t < s < t + k. This restriction implies that job loss occurs between the two points in time at which we observe health. We estimate the effect of job loss on health for worker i at time t+k with the following equation:
 | (1) |
is a stochastic error term. The parameter estimate denoted by 
captures the average effect of job loss for workers in the sample who experience involuntary job loss relative to those who remain continuously employed.
We estimated 1 using ordinary least squares (OLS). As our data are longitudinal, the possibility of serial correlation exists. We therefore tested for serial correlation, calculating Durbin's h statistic (Durbin 1970), a modified version of the Durbin–Watson statistic, that accounts for the lagged dependent variables in our models. For the physical functioning regression we find no evidence of serial correlation. For the mental health regression, however, we may not rule out serial correlation, as we were unable to calculate the h statistic. (One element of the statistic involves a square root; for mental health the number under the radical was negative.) An alternative test is proposed by Durbin 1970, but requires a second lag in the variables, which is not possible with two waves of data. It should be noted that, even if the error covariance matrix displays autocorrelated errors, the effect of such autocorrelation when only two waves of data are used is typically very limited (Diggle, Liang, and Zeger 1994).
Separate models were estimated to assess the effect of involuntary job loss on postseparation physical functioning and post-separation mental health. Consistent with 1, both models covaried the baseline value of the response variable and included a vector of control variables in addition to the dummy variable for displacement.
Additional models were estimated to investigate the effects of joblessness duration and of reemployment following involuntary job loss on health measures at the follow-up interview. We estimated these models using only respondents who had reported involuntary job loss at some point during the 2-year study period. In a first specification, a continuous variable representing number of months unemployed was substituted for the job loss dummy variable in 1. In the second specification, a dummy variable measuring reemployment during the study period (
) replaced the dummy variable for involuntary job loss in 1. The reemployment and duration effects were tested independently because of the dependence between the unemployment spell length and reemployment. All other covariates in these models were identical to those included in the regressions for the full sample.
Sensitivity analyses comparing the health effects of retirement with those of involuntary job loss were also performed. In these analyses (results not shown), we added to our full analysis sample () the individuals who reported being retired at the follow-up interview (), and indicated the inclusion of this subsample with an additional dummy variable. 1 was then reestimated with the dummy variable for retirement included as an independent variable. The results were suggestive of a protective effect (not statistically significant) of retirement on both physical functioning and mental health, consistent with the findings of other studies that have explored the relationship between retirement and health (Kasl and Jones in press). Inclusion of the dummy variable for retirement had virtually no effect on the estimated coefficient on involuntary job loss.
Although least squares regression with continuous outcome variables was the technique we ultimately chose to estimate the models of health in this study, several other procedures were considered. The reason for exploring such alternatives was that the dependent variables were somewhat nonnormally distributed (asymptotically declining). Two alternative methods to estimate 1 were therefore investigated, each of which addressed the shape of the outcome variable's distributions.
First, we estimated a log-linear variant of the model, transforming the dependent variables to their natural logarithms (adding 1 to each value of the dependent variable to avoid losing observations with 0 health scores) and estimating the model with OLS. Second, we estimated the model with Poisson regression, a maximum likelihood technique for estimating the likelihood of rare events (in this case, physical disability or poorer mental health status). In both cases, the results were not qualitatively different from the reported estimation results using least squares estimation.
Another important issue considered in this analysis was the potential bias resulting from endogeneity of explanatory variables. Endogeneity concerns derive from two sources: (a) declines in health may lead to involuntary job loss, and (b) unobserved factors that may be associated with both the likelihood of involuntary job loss and follow-up health status may have been omitted. Regardless of the source of this potential endogeneity, the consequence could be a biased estimate of the effect of job loss.
With regard to the first issue, we used a Hausman Specification Test (Hausman 1978) to assess the possible endogeneity of involuntary job loss. This test was done separately for physical functioning and mental health, and involved a two-step process. In the first step, we regressed the dummy variable for involuntary job loss on baseline health, all exogenous variables, and an instrumental variable, and calculated the residuals. The instrument chosen was the 1993 average weekly Unemployment Insurance (UI) benefit (U.S. Department of Labor 1993) in the geographical area in which the worker resides. Ruhm 1994 has demonstrated the efficacy of this instrument in an economic study of displacement. The average weekly UI benefit was associated with involuntary job loss ( p < .05), but not associated with either physical functioning or mental health ( p > .05). In the second step, we included the residuals calculated in the first step as an additional regressor on the right-hand side of the follow-up health regressions and considered the statistical significance of the estimated coefficients on the residuals. Our test results indicated that involuntary job loss is not endogeneously determined ( p > .05).
To account for the second potential source of endogeneity, we carefully selected the analysis sample to eliminate substantial sources of heterogeneity. In addition, we included a comprehensive set of independent variables, including reliable measures of baseline health.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
In Table 4 , results of independent tests of workers who reported involuntary job loss during the study period are displayed. Once again, in these analyses we investigated two concerns: whether either the unemployment spell length or reemployment was associated with follow-up physical functioning and mental health. The results indicate that, among workers suffering involuntary job loss, longer unemployment spells are suggestive of poorer physical functioning and mental health at follow-up. However, the observed effect was not statistically significant ( p > .05). Reemployment was positively associated with physical functioning and mental health at the follow-up interview ( p < .05). This result suggests that reemployment may be protective against the declines in both physical functioning and mental health associated with remaining unemployed. However, when considering the reemployment result, the possibility of reverse causality may not be eliminated, because physical functioning and mental health were not assessed after job loss and before reemployment for workers who had lost their jobs between the baseline and follow-up interviews.
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Previous research has suggested that, at least for those who are continuously unemployed, older individuals may adapt better than younger individuals to unemployment, perhaps due to reduced financial and family obligations of such older individuals (Warr and Jackson 1987; Warr et al. 1988). Our findings, however, indicate that the negative health effects of involuntary job loss are significant for older workers. In the United States, workers save heavily in the years immediately preceding retirement, often relying primarily on personal savings amassed in this period to finance the costs of retirement (Mitchell and Moore 1998). It is therefore easy to imagine that late-stage job loss could bear important consequences for the well-being of dislocated U.S. workers.
It has been debated in previous literature on the health consequences of job loss whether unemployment is a cause or merely the result of poorer health (Kasl and Jones in press). The findings of this study suggest that the observed effect of involuntary job loss on health includes a combination of selection and causality effects.
The selection interpretation (i.e., those who lose their jobs are more likely to have had poorer health prior to the job loss than those who do not lose their jobs, and thus, are observed to have poorer postloss health) is evidenced by the comparison of unadjusted and adjusted effects of job loss on post-loss physical functioning. The unadjusted effect of involuntary job loss on physical functioning was nearly double the adjusted effect once we controlled for baseline physical functioning, suggesting that the effect of involuntary job loss on post-loss physical functioning is due in part to the poorer baseline health of workers who experienced job loss. However, the magnitude of the association between baseline physical functioning and involuntary job loss was small. Moreover, the effect of involuntary job loss on physical functioning remains negative and significant even after we controlled for a comprehensive set of health, sociodemographic, and economic factors that might account for selection effects, suggesting that the observed association is not due only to selection effects.
The evidence of the effect of involuntary job loss on mental health is even more heavily weighted toward the causality interpretation. The unadjusted and adjusted associations between involuntary job loss and post-loss mental health are nearly identical, suggesting that the effect of job loss on post-loss mental health is not due to poorer baseline mental health among workers experiencing job loss. In addition, the association between baseline mental health and involuntary job loss was small. Further, the effect of involuntary job loss on mental health remains substantial and significant after we controlled for baseline measures of mental health, and sociodemographic and economic factors, suggesting again that the observed negative effects of involuntary job loss are not due only to selection effects.
Few previous studies have had rich enough data to assess how the health effects of involuntary job loss may vary by subgroups of respondents, an important factor in the planning of targeted interventions to buffer negative health effects. In analyses not shown, interaction models were estimated to test subgroup differences in the health consequences of involuntary job loss. Our results indicated that both older and unmarried individuals may be especially vulnerable to the negative mental health consequences of job loss. The finding regarding marital status is consistent with other unemployment studies that have shown that social support can buffer the health effects of job loss (Kasl and Jones in press). Neither interaction effect was detected to be significant for post-loss physical functioning, although the interaction effect of marriage was in the expected (buffering) direction.
This research has also demonstrated that for displaced workers, reemployment is positively associated with postseparation health. This was the case for both physical functioning and mental health. Our finding indicates that securing a new job may diminish the health impacts of job loss. In addition, our results are suggestive of an accumulation effect of protracted unemployment. This was again indicated for both health dimensions.
The study has limitations that merit recognition. Although health status measures were ascertained prior to job loss, it is possible that respondents' health status may have changed after the 1992 interview and prior to job loss. However, it is unlikely that such changes in health would be systematically related to plant closings in which all workers lose their jobs. Similarly, unmeasured factors potentially associated with health, such as intervening life events, may have been omitted from the model. Such omitted variables would not bias our results unless they were also correlated with involuntary job loss. We believe that this is unlikely, given the substantial range of covariates in the HRS and included in our models.
Another limitation relates to the classification of retirement. Some respondents may have chosen to report that they had retired when, in fact, they had been laid off. Such respondents would have been excluded from our analysis. The HRS explicitly asked retirees whether they would have been laid off and whether they were encouraged to leave their jobs. We found only four of the respondents reported this occurrence, but there may be others who did not report this way due to the social stigma that may accompany layoff.
In addition, there remains the potential for response rate bias, although HRS response rates are quite high for a national survey. It has been found that initial HRS nonrespondents (who were later induced to respond) were more likely to be non-White and less likely to report poor physical health than were traditional respondents (Juster and Suzman 1995). In this study, we have controlled for both race and health status. However, the potential response bias implied by Juster and Suzman 1995 remains a concern for the generalizability of the reported results.
Finally, data were available for only the 2-year follow-up date; we could therefore assess only the short-term consequences of involuntary job loss. Future studies will be able to estimate the longer run effects of involuntary job loss on health, as additional waves of the HRS become available.
Industry leaders, policy makers, and clinicians have reason to be concerned with the health effects of job loss and related unemployment. Others have demonstrated that unemployment negatively affects future income and earning potential (Jacobson et al. 1993; Kletzer 1989; Neal 1995; Ong and Mar 1992; Ruhm 1990, Ruhm 1991), expected consequences of losing one's major source of income. This study shows that, in addition to the economic consequences of worker displacement, there may be important health consequences of job loss, especially among older workers.
|
Acknowledgments |
|---|
We thank the participants in the Health Economics and Policy Workshop as well as the Aging Workshop at Yale University. We are particularly grateful for suggestions by William White, Jody Sindelar, John Rizzo, Ann Huff-Stevens, and Thomas Gill. We also thank Dennis Heffley for providing encouragement to pursue this study.
Received for publication March 31, 1999. Accepted for publication November 3, 1999.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  W. T. Gallo, J. E. Brand, H.-M. Teng, L. Leo-Summers, and A. L. Byers Differential Impact of Involuntary Job Loss on Physical Disability Among Older Workers: Does Predisposition Matter? Research on Aging, May 1, 2009; 31(3): 345 - 360. [Abstract] [PDF]
|
|
|
|
|
|
R. W. Johnson Employment Opportunities at Older Ages: Introduction to the Special Issue Research on Aging, January 1, 2009; 31(1): 3 - 16. [PDF]
|
|
|
|
 |
|
 Y. Hao Productive Activities and Psychological Well-Being Among Older Adults J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2008; 63(2): S64 - S72. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. A. Amirkhanyan and D. A. Wolf Parent care and the stress process: findings from panel data. J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2006; 61(5): S248 - S255. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. E. Locke Presidential address: psychosomatic medicine and biodefense preparedness--a new role for the american psychosomatic society. Psychosom Med, September 1, 2006; 68(5): 698 - 705. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. T. Gallo, E. H. Bradley, J. A. Dubin, R. N. Jones, T. A. Falba, H.-M. Teng, and S. V. Kasl The persistence of depressive symptoms in older workers who experience involuntary job loss: results from the health and retirement survey. J. Gerontol. B. Psychol. Sci. Soc. Sci., July 1, 2006; 61(4): S221 - S228. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Siegel, E. H. Bradley, W. T. Gallo, and S. V. Kasl Impact of Husbands' Involuntary Job Loss on Wives' Mental Health, Among Older Adults J. Gerontol. B. Psychol. Sci. Soc. Sci., January 1, 2003; 58(1): S30 - 37. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. E. Kim and P. Moen Retirement Transitions, Gender, and Psychological Well-Being: A Life-Course, Ecological Model J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2002; 57(3): P212 - 222. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. W. Marshall, P. J. Clarke, and P. J. Ballantyne Instability in the Retirement Transition: Effects on Health and Well-Being in a Canadian Study Research on Aging, July 1, 2001; 23(4): 379 - 409. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
