Abstract
Objective
To study predictors and consequences of unemployment.
Design
Prospective cohort study.
Setting
11 construction companies in southern Finland.
Participants
586 male employees, aged 40-59 years at baseline in 1991 and not retired during a 4 year follow up.
Main outcome measures
Long term unemployment, stress symptoms, disease, alcohol consumption, exercise activity, and body mass index.
Results
In a multiple logistic regression model, long term unemployment (>24 months v ⩽24 months) was predicted by age 50-54 years v 40-44 years (odds ratio 2.0, 95% confidence interval 1.1 to 3.7), ⩽3 years’ employment in the present job (3.1, 1.9 to 5.1), previous unemployment (2.1, 1.2 to 3.8), being single (1.8, 1.1 to 3.1), current smoking (2.6, 1.4 to 4.7), high alcohol consumption (2.1, 1.1 to 4.3), body mass index <23 kg/m2 v 23-29 kg/m2 (2.4, 1.3 to 4.4), frequent stress symptoms (2.0, 1.2 to 3.2), mental disorders (7.8, 1.5 to 40.0), and skin diseases (2.0, 1.0 to 3.9). Workers who were unemployed long term reported increased stress (2.1, 1.2 to 3.5) but fewer incident diseases (0.6, 0.4 to 0.9), decreased alcohol consumption (2.9, 1.6 to 5.2), increased exercise (1.9, 1.2 to 3.0), and increased body mass index (2.3, 1.3 to 4.0) compared with the rest of the cohort.
Conclusions
The workers’ perceptions of work did not predict unemployment. Health based selection to long term unemployment was shown. Smoking and high alcohol consumption predated unemployment, but favourable lifestyle changes were observed among the unemployed workers. Stress predicted unemployment, which further increased the stress.
Key messages
Among middle aged construction workers, long term unemployment was predicted by an unstable occupational history, health problems, smoking, high alcohol consumption, low body mass index, and being single
Stress predicted long term unemployment, which further increased the level of stress
Men unemployed long term reported increased exercise and decreased alcohol consumption, but increased body mass index, and fewer incident diseases than the rest of the cohort
Changes in personal economic situation with unemployment mediated the effect of unemployment on stress but not on alcohol consumption
Introduction
A relation between unemployment and ill health seems established: several studies reported mental distress,1–3 frequency of use of health care,4–6 and mortality5–8 to be high among unemployed individuals. The mechanisms of this relation are not, however, obvious. It is possible that (a) poor health leads to a low status on the labour market and increases the risk of unemployment, (b) job loss and unemployment have adverse effects on health, or (c) both mechanisms are active.9
The ideal way to investigate predictors and consequences of unemployment would be to conduct a prospective study of a cohort of employed people at baseline, with a subgroup becoming unemployed, and who were followed up over a long period. Studies on the effects of factory closure provide an option, but a suitable comparison group is difficult to find and potential participants are usually alerted to the possibility of closure around the same time as the researchers.1,10 Socioeconomically disadvantaged people are at a high risk of both health problems and unemployment; it has not always been possible to control for this source of confounding. It seems plausible that characteristics apart from health—for example, lifestyle or attitudes towards work and family—would play a role in unemployment or re-employment.
We studied some predictors of unemployment in a cohort of construction workers during Finland’s recession in the early 1990s. We also studied changes in health and lifestyle with duration of unemployment and whether some changes resulted from increased economic strain while jobless.
Participants and methods
Data collection
We compiled our baseline data by interviewing 961 men employed in the construction industry in Finland in 1991.11 We approached all blue collar workers, aged 40-64 years, of 11 construction companies with occupational health departments in the Helsinki metropolitan region or in Häme, southern Finland; 947 men (98.5%) agreed to participate. We were given staff lists for each worksite, but a few men could not be traced owing to change of worksite. Interviews at baseline and follow up lasted 45 minutes. Baseline interviews were conducted at the worksites by nurses trained for the purpose. Responses were entered into a computer during the interview. Four years later, 741 men (77.1%) were available for interview conducted by telephone by the Kuopio Regional Institute of Occupational Health.
We restricted our analyses to men aged 40-59 years at baseline (707 participants). We excluded 104 participants who retired during follow up and 17 participants with missing data on duration of unemployment, leaving 586 men.
Unemployment
We dichotomised the number of times workers were unemployed during the 5 years preceding the baseline interview (0 or ⩾1). Participants were asked how many months they had been unemployed or temporarily laid off for each year since the beginning of 1991. Unemployment was measured as the total number of months. Owing to the skewed distribution of this variable, we defined duration of unemployment as long term (more than 24 months’ redundancy during follow up) and short term (joblessness for 1-24 months during follow up).
Sociodemographic variables and economic situation
Being single was defined as unmarried, separated, or widowed (v married or cohabiting); vocational training (yes or no) included training at work and at school; and occupational skills (skilled or semiskilled) were on the basis of job titles. The men were asked if their economic situation was fair, average, or poor.
Lifestyle and health
We ascertained how much physical exercise the men took during leisure time by asking how often they undertook such exercise for at least half an hour so that they got at least slightly out of breath and started to sweat (classified as never to 2-3 times a month v once a week to daily).
Smoking habits, inquired of at baseline only, were classified as never smoker, stopped smoking, and current smoker. We calculated an alcohol consumption index by converting separately assessed amounts of beer, wine, and spirits into absolute alcohol (g/day). An alcohol free period of 1 year was recorded as 0 g alcohol. The rank correlation coefficient between the alcohol consumption indices of 1991 and 1995 was 0.71. The index was used both as a continuous variable (range 0-6.57)and as a dichotomised variable (index class represented upper 10% of the variable distribution). Body mass index (kg/m2) assessed the extent of overweightness.
The participants were asked about current diseases diagnosed by a physician. We considered disorders that were musculoskeletal, cardiovascular, respiratory, mental, neurological and sensory, gastrointestinal, skin, neoplastic, haematological, and urogenital. We also calculated the number of diseases (0, 1, 2, or ⩾3).
Musculoskeletal symptoms (Cronbach’s α=0.85) were scored by asking participants if they had had aches, stiffness, tenderness on movement, numbness, or pain at particular sites during the past year (14 items: unilateral or bilateral neck, occipital region; radiation of neck pain to the arm; shoulder, upper arm; elbow, lower arm; wrist, hand, fingers; thoracic spine; low back; radiation of low back pain to lower limb; hip joint; thigh; knee; calf; ankle, foot; sole, toes): never or rarely (score 1), occasionally (2), rather often (3), and often or continuously (4).
Stress symptoms (α=0.75) were scored as for musculoskeletal symptoms by asking participants whether they had had particular symptoms recently (16 items: headache; irritability or fits of anger; tension or nervousness; loss of energy; dyspnoea without physical exertion; excessive perspiration without physical effort; tremor of hands; rapid or irregular heart beats; dizziness; reduced libido; sleeping difficulties; nightmares; diarrhoea or irregular bowel function; abdominal pain or acid troubles; nausea or vomiting; loss of appetite.)
Work related variables
We asked participants how many years they had been in their present employment, and whether they had been unemployed or temporarily laid off during the past 5 years (no=1, at least once=2). Physical strain was assessed by asking whether the work required too much in relation to the participants’ ability (sum score (α=0.80) on basis of five items—physical work and use of muscle power, lifting and carrying, abrupt efforts, repetitive work movements, and stooped and twisted work postures). Participants were asked whether job satisfaction was impaired by rushed and tight schedules, forced work pace, lack of influence, lack of recognition, or interference from supervisors. Scores were summed to give a psychosocial strain score (α=0.55). Participants were also asked to describe their work ability using a visual analogue scale from 0 (worst) to 10 (best).
Attitude to work was assessed by asking participants to select the statement most relevant to their job: (a) my job is like any other—I do my job but the only thing that matters is the money; and (b) it is a job that means something to me—in addition to the income, I get personal satisfaction out of it. Participants were asked to describe the significance of their home and family and pastime activities (classified as not significant v fairly significant, or very significant).
Statistical methods
We examined separately age adjusted odds ratios of the predictor variables for both short term (1-24 v 0 months) and long term (>24 v ⩽24 months) unemployment. Age was classified as 40-44, 45-49, 50-54, and 55-59 years. The associations between determinants and unemployment were studied by using smoothers (cubic B splines in S-Plus software).12 We especially studied the functional form of the relation, linearity, and optimal cut off points. The classification for body mass index was on the basis of smoothing, but most classifications were made according to the frequency distribution of the variable (mostly about one third put in the index class).
Multiple logistic regression analyses included independent variables that were significantly associated with unemployment when allowing for age (we did not include the variable of home and family owing to a high covariance with marital status). The final models included predictors significantly associated with unemployment.
Duration of unemployment was classed into four equally sized groups (0, 1-8, 9-20, 21-47 months); long term unemployment versus other was also used. Allowing for age, we performed logistic regression analyses of the relation between duration of unemployment and changes to health and lifestyle. Disease variables were dichotomised (0 v ⩾1 diseases) when necessary. The index category of the other variables represented the upper or lower decile of the variable difference.
To account for the possibility that a change in the employee’s economic situation would have mediated some effects of unemployment, we compared models with and without the trichotomised difference of the variable describing the participant’s economic situation.
Results
At follow up, 241 men (41.1%) were employed, 293 (50.0%) unemployed, 21 (3.6%) laid off, 18 (3.1%) on sick leave, and 13 (2.2%) reported other situations. Four out of five participants had been unemployed during follow up—mean total duration 12 months (table 1).
Table 1.
Distribution of number of months men were unemployed during follow up
| Months unemployed | No (%) of men (n=586) |
|---|---|
| 0 | 125 (21.3) |
| 1-6 | 124 (21.2) |
| 7-12 | 100 (17.1) |
| 13-24 | 122 (20.8) |
| 25-47 | 115 (19.6) |
Predictors of unemployment
The highest odds ratios of unemployment were in the 50-54 year old age group (table 2). The age adjusted odds ratios of factors measured at baseline, with long term unemployment during follow up, are presented in table 2, and the multivariate model of the predictors is shown in table 3. Subsequent unemployment was associated with previous unemployment, duration of present employment contract, being single, heavy alcohol consumption, and current smoking, but not physical or psychosocial strain at work, work ability, or valuation of pastime activities.
Table 2.
Association between baseline measurements and long term unemployment (>24 months) during follow up (odds ratios adjusted for age group)
| Variable | % | Odds ratio (95% CI) |
|---|---|---|
| Age group (years) | ||
| 40-44 | 31.4 | 1 |
| 45-49 | 32.8 | 1.10 (0.65 to 1.87) |
| 50-54 | 27.0 | 1.67 (0.99 to 2.83) |
| 55-59 | 8.9 | 1.03 (0.46 to 2.34) |
| Occupational history | ||
| Previous unemployment | 18.4 | 2.57 (1.59 to 4.15) |
| ⩽3 years in present job | 34.3 | 3.50 (2.27 to 5.40) |
| Sociodemographic | ||
| No vocational school | 51.9 | 1.39 (0.91 to 2.12) |
| Semiskilled worker | 30.4 | 1.91 (1.25 to 2.93) |
| Single | 22.6 | 2.37 (1.51 to 3.74) |
| Lifestyle | ||
| Smoking: | ||
| Never smoker | 31.1 | 1 |
| Stopped smoking | 25.6 | 1.55 (0.81 to 2.97) |
| Current smoker | 42.5 | 3.31 (1.91 to 5.73) |
| High alcohol consumption | 9.1 | 2.89 (1.58 to 5.31) |
| Alcohol consumption (continuous variable) | 1.76 (1.31 to 2.38) | |
| Infrequent physical exercise | 32.0 | 1.37 (0.89 to 2.11) |
| Body mass index (kg/m2) | ||
| <23 | 15.0 | 2.15 (1.25 to 3.68) |
| 23-29 | 62.6 | 1 |
| >29 | 22.4 | 1.25 (0.76 to 2.08) |
| Symptom scores | ||
| Frequent musculoskeletal symptoms | 32.3 | 1.22 (0.79 to 1.89) |
| Frequent stress symptoms | 34.3 | 2.06 (1.36 to 3.14) |
| Diseases | ||
| Cardiovascular | 18.1 | 1.39 (0.84 to 2.30) |
| Musculoskeletal | 32.3 | 1.22 (0.79 to 1.89) |
| Respiratory | 13.7 | 1.25 (0.71 to 2.21) |
| Gastrointestinal | 11.9 | 2.11 (1.21 to 3.68) |
| Urogenital | 2.6 | 0.64 (0.14 to 2.88) |
| Skin | 10.4 | 1.88 (1.03 to 3.41) |
| Neoplastic | 2.7 | 1.42 (0.45 to 4.53) |
| Haematological | 2.0 | 0.84 (0.18 to 3.93) |
| Neurological and sensory | 14.5 | 1.75 (1.03 to 2.97) |
| Mental | 1.4 | 7.28 (1.70 to 31.18) |
| No of diseases: | ||
| 0 | 20.8 | 1 |
| 1 | 23.9 | 1.27 (0.67 to 2.39) |
| 2 | 21.7 | 0.99 (0.50 to 1.95) |
| ⩾3 | 33.6 | 1.52 (0.84 to 2.73) |
| Working conditions and work ability | ||
| High physical strain | 37.9 | 1.24 (0.81 to 1.88) |
| High psychosocial strain | 31.5 | 1.15 (0.75 to 1.78) |
| Low work ability | 23.7 | 0.94 (0.58 to 1.50) |
| Work related factors | ||
| Attitude to work | 24.5 | 1.46 (0.92 to 2.30) |
| Low valuation of home and family | 30.9 | 1.94 (1.27 to 2.97) |
| High valuation of pastime activities | 33.4 | 1.53 (0.97 to 2.42) |
Table 3.
Multivariate logistic regression model of predictors of long term unemployment (>24 months)
| Variable | Odds ratio (95% CI) |
|---|---|
| Age (years): | |
| 40-44 | 1 |
| 45-49 | 1.12 (0.62 to 2.03) |
| 50-54 | 2.00 (1.08 to 3.71) |
| 55-59 | 1.39 (0.55 to 3.52) |
| Previous unemployment | 2.13 (1.20 to 3.80) |
| ⩽3 years in present job | 3.09 (1.88 to 5.08) |
| Single | 1.84 (1.09 to 3.12) |
| Smoking: | |
| Never smoker | 1 |
| Stopped smoking | 1.98 (0.99 to 3.97) |
| Current smoker | 2.58 (1.42 to 4.69) |
| High alcohol consumption | 2.11 (1.05 to 4.25) |
| Body mass index (kg/m2): | |
| <23 | 2.35 (1.26 to 4.36) |
| 23-29 | 1 |
| >29 | 1.71 (0.96 to 3.02) |
| Frequent stress symptoms | 1.97 (1.22 to 3.21) |
| Mental disorder | 7.75 (1.51 to 39.93) |
| Skin disease | 1.99 (1.01 to 3.93) |
Frequent symptoms of stress was predictive of long term unemployment (table 2). Participants with diagnosed mental, gastrointestinal, skin, or neurological or sensory disorders had an increased risk of long term unemployment. The mental disorders were almost all at the level of neurosis (7 cases; age adjusted odds ratio 10.6, 95% confidence interval 2.0 to 55.5). The most prominent disorders predictive of unemployment were gastric or duodenal ulcers (27 cases; 3.5, 1.6 to 7.7), allergic eczema (25 cases; 2.4, 1.03 to 5.6), and ocular diseases when diffraction abnormalities were excluded (15 cases; 3.7, 1.3 to 10.4).
Table 3 shows the final model of long term unemployment. At a probability of 0.22, the model (sensitivity 58% and specificity 78%) correctly classified 74% of the cases.
The final multivariate model of short term unemployment included: age group (only men aged 55-59 years differed significantly from those aged 40-44 years; 3.56, 1.29 to 9.83), having been in the present job for no more than 3 years (1.92, 1.16 to 3.17), and alcohol consumption (continuous variable: 2.41, 1.27 to 4.60). The predictive performance of the model at the cut off point of 0.74 was sensitivity 50%, specificity 61%, and overall correct percentage 53.
Consequences of unemployment
Participants unemployed long term reported an increased frequency of exercise,decreased alcohol consumption, and increased body mass index compared with the rest of the cohort (table 4); the dependencies of frequency of exercise and alcohol consumption on duration of unemployment were monotonous (table 5).
Table 4.
Long term unemployment (>24 months) during follow up in relation to changes in lifestyle variables, body mass index, symptom scores, and incident diseases (comparison group was all other participants). Odds ratios adjusted for age group
| Variable | Odds ratio (95% CI) |
|---|---|
| Increase in exercise frequency* | 1.91 (1.24 to 2.96) |
| Decrease in alcohol consumption† | 2.88 (1.59 to 5.22) |
| Increase in body mass index* | 2.28 (1.30 to 4.00) |
| Increase in stress symptoms* | 2.06 (1.23 to 3.47) |
| Decrease in musculoskeletal symptoms† | 1.22 (0.67 to 2.22) |
| Incident diseases | 0.61 (0.40 to 0.94) |
Index category is upper tenth of score difference.
Index category is lower tenth of score difference.
Table 5.
Multivariate logistic regression analyses (adjusted for age group) for duration of unemployment and change of participants’ economic situation during follow up in relation to changes in stress symptoms, disease morbidity, body mass index, and alcohol consumption. Values are odds ratios (95% confidence intervals)
| Variable | Increase in stress symptoms* | Incident diseases | Increase in body mass index* | Decrease in alcohol consumption† |
|---|---|---|---|---|
| No of months unemployed | ||||
| 0 | 1 | 1 | 1 | 1 |
| 1-8 | 1.39 (0.61 to 3.15) | 1.16 (0.71 to 1.88) | 0.59 (0.25 to 1.40) | 3.57 (0.99 to 12.93) |
| 9-20 | 1.70 (0.76 to 3.79) | 0.81 (0.49 to 1.34) | 1.05 (0.47 to 2.34) | 5.11 (1.41 to 18.56) |
| 21-47 | 2.09 (0.95 to 4.60) | 0.67 (0.40 to 1.13) | 1.89 (0.88 to 4.06) | 8.04 (2.28 to 28.38) |
| Economic situation | ||||
| No change | 1 | 1 | 1 | 1 |
| Better | 0.50 (0.06 to 3.90) | 0.82 (0.33 to 2.05) | 1.02 (0.22 to 4.66) | 2.00 (0.42 to 9.60) |
| Worse | 2.13 (1.30 to 3.50) | 1.05 (0.73 to 1.52) | 0.89 (0.50 to 1.57) | 1.33 (0.74 to 2.42) |
Index category is upper tenth of score difference.
Index category is lower tenth of score difference.
Participants employed long term reported increased stress but fewer incident diseases compared with the rest of the cohort (tables 4 and 5); scores for musculoskeletal symptoms did not change significantly with unemployment. We observed no associations between specific diseases and duration of unemployment.
In a logistic regression model including the length of unemployment, change in the participant’s economic situation during follow up, and age group as independent variables, increased economic hardship was associated with increased stress (table 5). The association between duration of unemployment and increased stress was reduced to a statistically non-significant level. In similar models, the change in a participant’s economic situation was not associated with decreased alcohol consumption (table 5), incident diseases, or increased body mass index.
Discussion
Economic trends cause wide fluctuations in the demand for labour in the construction industry. During the Finnish recession of the early 1990s, nearly 50% of construction workers were jobless at one point. In our study, the participants were employed at the time of the first interview, but only 20% remained employed during the entire 4 year follow up.
It has been argued that when the unemployment rate is high, individual characteristics are of less importance in relation to job loss than when under more favourable conditions; mortality statistics give credence to this.8 We identified some predictors of unemployment even during mass redundancy. However, we could predict unemployment, particularly short term, only to a minor extent.13 Thus, unemployment is mostly determined by other factors such as economic hardship, bankruptcy, and perhaps personal factors that were not addressed.
Health and lifestyle factors as predictors of unemployment
The stability of the participants’ employment history, marital status, and some lifestyle and health characteristics independently predicted long term unemployment. It was not possible to estimate the extent to which these increased the probability of job loss or decreased the chance of re-employment. However, as most of the variables had predictive value only in relation to long term unemployment, it seems that the latter was mostly the case.
It is possible that employers consider marital status when deciding about employment. It is also possible that individuals with a family want to maximise their income and actively seek re-employment, but as the association was not confined to married or cohabiting workers it may reflect a more general factor. Excessive alcohol consumption poses potential problems for work output, occupational safety, and cooperation. Also a worker consuming excess alcohol may be less motivated to seek another job. Smoking could indicate risk taking and even susceptibility to accidents.14 It seems more probable that smoking is intricately entwined with an individual’s life history.15
Our results support the hypothesis of unemployment being selective according to an individual’s health: when other factors were accounted for, mental disorders, skin disorders, and stress symptoms were predictive of long term unemployment. Gastrointestinal disorders lost significance as predictors when smoking and alcohol consumption were accounted for; these may have contributed to the diseases.
Unemployment was not associated with educational level, but this could have been due to the low variation in education among the participants. That work related strain, both physical and mental, and work ability were not associated with unemployment is interesting considering the physical strenuousness of construction work: how work was perceived by the employee was not of importance in relation to unemployment.
Diseases and stress
Participants who were unemployed long term reported less incident diseases than the rest of the cohort. The finding is at variance with earlier reports.4–6 Finnish people have universal access to health care, and it is unlikely that the reduced economic resources of unemployed workers would have greatly affected the probability of them consulting a physician; the empirical evidence supported this. As occupational health services were no longer available to the unemployed worker, perhaps an unwillingness to seek care elsewhere was a factor here. Another possibility is that the reduction in work load and other work exposures influenced seeking care or even morbidity.
It has been assumed that a person finds losing a job less stressful when the rate of general unemployment is high.16 In our study, stress increased among the unemployed workers during a period of mass redundancy and did not level off when duration of unemployment increased. A chronic state of stress may lead to physical and mental morbidity. Thus a longer follow up of our cohort might have shown other health consequences.
Lifestyle changes
Increased stress among the unemployed workers did not lead to increased alcohol consumption.1,17 As with previous findings, alcohol consumption decreased with increasing duration of unemployment, perhaps due to the change in work and leisure routine: diminished economic resources were not of importance. Becoming unemployed may have different effects on health behaviour depending on the individual’s age. Reports on the adverse effects of unemployment on lifestyle have been in young people.18,19
It was not possible for us to study changes in smoking habits, but in another prospective series of Finnish construction and forest workers around the same period, the proportion of regular smokers decreased with unemployment.20 The increase in exercise frequency observed by us may be because the construction workers are accustomed to high physical activity. Health campaigns directed at unemployed people have also promoted exercise as a means of preserving work ability. Despite increased physical activity during leisure time and decreased alcohol consumption, the unemployed workers tended to become more obese than the rest of the cohort. A similar finding has been reported in middle aged British men.17
Generalisability of results
Our findings may not be generalisable to other types of employment. In construction work, repeat short periods of unemployment are common owing to seasonal and other variations in construction activity, and unemployment may not be as much a threat to occupational identity21 as it would be to workers in more stable employment (a more profound increase in stress might be expected if such workers are made redundant). In terms of the comparative increase in economic hardship, one mechanism of increased stress and ill health,6,22 the consequences are comparable.
One limitation of our study was that all information was self reported, although the data were obtained by trained interviewers. The baseline interview was conducted at the workplace, but follow up information was obtained by telephone. Any bias concerns the follow up data and should not affect comparisons within it. To increase the reliability of information on the duration of unemployment, this was assessed separately for each year of follow up and then summed.
In conclusion, selection to long term unemployment according to poor health, particularly mental health, was evident. The association of stress with unemployment was reciprocal: stress predicted unemployment, which further increased the level of stress. High alcohol consumption and smoking predated becoming jobless, but favourable lifestyle changes were observed among unemployed workers.
Acknowledgments
We thank Timo Aitta, and Risto Hagman for scoring the responses for alcohol consumption.
Footnotes
Funding: Finnish Work Environment Fund.
Competing interests: None declared.
References
- 1.Morris JK, Cook DG. A critical review of the effect of factory closures on health. Br J Ind Med. 1991;48:1–8. doi: 10.1136/oem.48.1.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hallsten L. Mental health and unemployment. On health-based selection to work. Arbete och Hälsa 1998:7. (Dissertation in Swedish with English abstract.)
- 3.Weich S, Lewis G. Poverty, unemployment, and common mental disorders: population based cohort study. BMJ. 1998;317:115–119. doi: 10.1136/bmj.317.7151.115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Elkeles T, Seifert W. Unemployment and health impairments. Eur J Public Health. 1993;3:28–37. [Google Scholar]
- 5.Jin RL, Shah CH, Svoboda TJ. The impact of unemployment on health: a review of the evidence. Can Med Assoc J. 1995;153:529–540. [PMC free article] [PubMed] [Google Scholar]
- 6.Mathers CD, Schofield DJ. The health consequences of unemployment: the evidence. Med J Aust. 1998;168:178–182. doi: 10.5694/j.1326-5377.1998.tb126776.x. [DOI] [PubMed] [Google Scholar]
- 7.Bethune A. Economic activity and mortality of the 1981 census cohort in the OPCS longitudinal study. Popul Trends. 1996;83:37–42. [PubMed] [Google Scholar]
- 8.Martikainen P, Valkonen T. Excess mortality of unemployed men and women during a period of rapidly increasing unemployment. Lancet. 1996;348:909–912. doi: 10.1016/S0140-6736(96)03291-6. [DOI] [PubMed] [Google Scholar]
- 9.Bartley M. Unemployment and ill health: understanding the relationship. J Epidemiol Community Health. 1994;48:333–337. doi: 10.1136/jech.48.4.333. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Avery AJ, Betts DS, Whittington A, Heron TB, Wilson SH, Reeves JP. The mental and physical health of miners following the 1992 national pit closure programme: a cross sectional survey using general health questionnaire GHQ-12 and short form SF-36. Public Health. 1998;112:169–173. [PubMed] [Google Scholar]
- 11.Matikainen E, Liira J, Rytkönen H. Health, work ability and working conditions of construction workers. Study design and follow-up in 1991-1995. Työ ja ihminen. 1998;12:85–93. , 152. (In Finnish with English abstract.) [Google Scholar]
- 12.S-PLUS 4 guide to statistics. Seattle: Data Analysis Products, Math Soft; 1997. [Google Scholar]
- 13.Leino-Arjas P, Liira J, Mutanen P, Malmivaara A, Matikainen E. Unemployment, health and health behaviour of construction workers. Työ ja ihminen. 1998;12:139–148. , 155. (In Finnish with English abstract.) [Google Scholar]
- 14.Ryan J, Zwerling C, Jones M. Cigarette smoking at hire as a predictor of employment outcome. J Occup Environ Med. 1996;38:928–933. doi: 10.1097/00043764-199609000-00017. [DOI] [PubMed] [Google Scholar]
- 15.Lee AJ, Crombie IK, Smith WCS, Tunstall-Pedoe HD. Cigarette smoking and employment status. Soc Sci Med. 1991;33:1309–1312. doi: 10.1016/0277-9536(91)90080-v. [DOI] [PubMed] [Google Scholar]
- 16.Dooley D, Catalano R. Unemployment as a stressor: findings and implications of a recent study. In: Badura B, Kickbusch I, editors. Health promotion research. Towards a new social epidemiology. Copenhagen: WHO Regional Publications; 1991. pp. 313–339. . (European series No 37.) [PubMed] [Google Scholar]
- 17.Morris JK, Cook DG, Shaper AG. Non-employment and changes in smoking, drinking, and body weight. BMJ. 1992;304:536–541. doi: 10.1136/bmj.304.6826.536. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Janlert U, Hammarström A. Alcohol consumption among unemployed youths: results from a prospective study. Br J Addict. 1992;87:703–714. doi: 10.1111/j.1360-0443.1992.tb02716.x. [DOI] [PubMed] [Google Scholar]
- 19.Montgomery SM, Cook DG, Bartley MJ, Wadsworth M. Unemployment, cigarette smoking, alcohol consumption and body weight in young British men. Eur J Publ Health. 1998;8:21–27. [Google Scholar]
- 20.Liira J, Leino-Arjas P. Unemployment in construction and forest work. Predictors and consequences in a 5-year follow-up. Scand J Work Environ Health. 1999;25:42–49. doi: 10.5271/sjweh.382. [DOI] [PubMed] [Google Scholar]
- 21.Ezzy D. Unemployment and mental health: a critical review. Soc Sci Med. 1993;37:41–52. doi: 10.1016/0277-9536(93)90316-v. [DOI] [PubMed] [Google Scholar]
- 22.Lynch JW, Kaplan GA, Shema SJ. Cumulative impact of sustained economic hardship on physical, cognitive, psychological, and social functioning. New Engl J Med. 1997;337:1889–1895. doi: 10.1056/NEJM199712253372606. [DOI] [PubMed] [Google Scholar]