Alcohol

Socio-environmental and stress-based theories of alcoholism propose that alcohol abuse and addiction develop as coping reactions to stressful socio-environmental conditions such as unemployment[17]. However, it is also known that risky behaviours, such as problem drinking, result in increased unemployment[18]. Thus, it is likely that both causation and selection effects are involved in the relationship between alcohol consumption and unemployment. Cross-sectional studies in Scotland[19] and France[20] have found an increased frequency of alcohol abuse among the unemployed compared to the employed. Longitudinal studies have yielded mixed results; studies in the US[21] and Sweden[22] suggest that unemployment may increase alcohol consumption, but British and Norwegian studies demonstrate no such effect [23–25]. There is some evidence that the relationship between unemployment and alcohol use may be time-dependent; a recent investigation suggested that short-term unemployment decreases alcohol use while longer unemployment increases it[26].

Smoking

Cigarette smoking, like alcohol use, has been found to be associated with stressful socioeconomic conditions such as unemployment[27]. Cross-sectional studies generally show that unemployed people are more likely to smoke than employed people[28], although unemployed smokers sometimes report smoking fewer cigarettes a day than their employed counterparts[29]. Longitudinal studies on unemployment and smoking have presented mixed findings[24] and are not consistent across countries[30]. Yet, there is some evidence that job loss is a risk factor for increasing smoking and that these effects may be long lasting[31].

Body Weight

There is evidence from longitudinal studies that unemployment may impact body weight; data from the British Regional Heart Study[24] showed that men who had experienced some unemployment during the study period were more likely than continuously employed men to either lose or gain more than 10% of their body weight. In the study of a factory closure in Michigan, job loss was shown to have an impact on body weight. Subjects who lost their job showed greater instability in their weight over the two years of observation, even after re-employment[32]. In a nationwide representative sample of Finnish subjects, overweight in women was associated with long-term unemployment[33]. Yet, the association between food insecurity and body mass index (BMI) has been found to be curvilinear (i.e., thin and overweight people display different behaviours from normal weight subjects while facing periods of economic and food insecurity)[34].

Objectives

As is the case with any epidemiological research short of a clinical trial, it is virtually impossible to control for all of the variables that could determine both employment instability and health behaviours simultaneously. However, when facing contradictory evidence on the relationship between unemployment and health behaviours, it would be precipitous to conclude that differences are due simply to a better or worse adjustment for the effects of selection. A key factor in understanding these mixed findings could be the social context in which the unemployment occurs.

Systems of formal support and social benefits are known to help maintain the health status of individuals exposed to unemployment. Perceived health status has been shown to differ between groups of unemployed people defined by the types of benefits they receive[30, 35] Mental health is also affected by social assistance programs with the receipt of benefits being associated with a reduction in depression symptoms[35].

Our aim is to examine the impact of unemployment on smoking, drinking and body weight changes among re-employed individuals. This study expands upon previous work by investigating whether unemployment compensation influences these behaviours. Receiving unemployment compensation may help alleviate the stress of unemployment and thereby reduce the negative behavioural effects for two reasons. One reason is because receiving benefits reduces the psychological stress associated with a sharp decline in income. The second, in line with research on the study of income inequality, social capital and health[36], is because unemployment compensation might act as a form of social support for the unemployed. While the benefits conferred in unemployment compensation are primarily financial, previous research has shown that the individual's perception of support, in addition to the extent of actual supportive behaviours, mediates its health protective effects[37]. Thus, by reducing the financial, psychological and sociological strains of unemployment, the receipt of unemployment compensation may help reduce the negative behavioural changes associated with job-instability.

Participants

Data for this study were provided from the Panel Study of Income Dynamics (PSID) conducted by the University of Michigan, Ann Arbor. The PSID is a longitudinal survey of a representative sample of US individuals and the families with whom they reside. The study emphasizes the dynamic aspects of economic and demographic behaviour but also contains information on mental and physical health. Beginning with a national sample of 4,800 families in 1968, the PSID has traced individuals from those original households and their offspring since that time. Because the original focus of the study was the dynamics of poverty, the 1968 sample included a disproportionately large number of low income households. To make the sample more representative of the population, additional families were added in 1995 and 1999 resulting in a sample size of 6,997 households in 1999. The PSID has a very high response rate ranging between 96.9% and 98.5%. Approximately 92% of the sample is interviewed by telephone each year; the remaining 8% of the sample is interviewed in person[38]. The household head, defined as the husband in a husband-wife pair or the primary wage earner, is most often the respondent.

Measures

Statistical Analysis

A logistic regression analysis was performed to investigate whether a period of unemployment in 2000 resulted in a change in smoking, drinking and body weight among re-employed people in 2001 in comparison with baseline levels recorded in 1999. All of our models adjust for potential confounders of the relationship between the likelihood of a change in smoking, drinking and body weight and unemployment status. These include: age, sex, race, education, health status, income, number of household members and marital status of the respondents. In addition, we created different variables to control for prior smoking, drinking and weight for each of the outcome variables as described above. The income variable used was the total household post-government income created for use in the Cross-National Equivalent File[38]. This represents the combined income after taxes and government transfers. Likelihood ratio tests comparing models assuming the exposure showed a log-linear effect with a more general model including the exposure as a categorical variable were used to determine the most appropriate means for modelling, income, number of household members and, BMI in 1999. None of the likelihood ratio test statistics for these variables were significant for any of the outcome variables and in order to increase our power to detect an association, income, number of household members and BMI in 1999 were modelled as continuous variables.

We calculated the mean increase in the number of cigarettes smoked per day between 1999 and 2001. We also calculated the mean increase and decrease in body weight (measured in both pounds and BMI) between 1999 and 2001. Because alcohol consumption was measured as a categorical variable, we were only able to determine if consumption had increased by 1, 2 or 3–4 categories (equivalent to increasing consumption by 1–2 drinks per day, 3–4 drinks per day or 5 or more drinks per day). For all three outcome variables, we investigated any differences in the magnitude of changes by employment status in 2000.

The statistical analysis was performed using binary logistic regression analysis with SPSS 11.5 V and SPSS 13.0. Correlation and colinearity analyses were performed to confirm the appropriateness of the models.

Descriptive

We investigated other possible differences among the three groups at baseline. Although the number of weeks and the hours per week worked in 1999 were similar, annual earnings were substantially different. The continuously employed in 2000 reported an average individual labour earning of $45,316 in 1999 (for a total of 2,115 hours of work), those who experienced unemployment in 2000 but received unemployment compensation reported an average earning of $37,589 in 1999 (for 2,259 hours of work) and those who experienced a period of unemployment in 2000 but did not receive unemployment compensation reported an average earning of $27,980 in 1999 (for 2,139 hours of work). The income difference between the groups was not statistically significant in a GLM analysis when controlling for the variables included in our analytical models.

The subjects who experienced unemployment in 2000 reported slightly higher levels of smoking, drinking and BMI in 1999 than those who were continuously employed in 2000. However, the difference was not significant when controlling for age, gender and ethnicity in a GLM analysis.

Alcohol

Among the 604 participants who reported an increase in alcohol consumption, 88.6% increased by 1–2 drinks per day, 8.9% increased by 3–4 drinks per day and 2.5% increased by at least 5 drinks per day. The proportion of individuals increasing their drinking by more than 3 drinks per day was slightly greater among those unemployed in 2000; among those employed, 11.2% increased by 3 or more drinks per day compared to 12.5% and 16.7% of individuals who were unemployed with and without compensation, respectively. The magnitude of the increase in alcohol consumption was not seen to differ greatly between those who were light and moderate/heavy drinkers at baseline in 1999.

Smoking

Among those who increased smoking, the average increase was by 8.5 cigarettes per day (95% CI 7.7–9.4). The average increase was slightly higher among those who were unemployed in 2000 and did not receive compensation. Individuals who experienced unemployment without compensation had an average increase of 9.5 cigarettes per day (95% CI 5.9–13.0). Those who were employed had an average increase of 8.4 (95% CI 7.4–9.3) and those who were unemployed with compensation had an average increase of 5.1 (95% CI 2.3–8.0).

Body Weight

Employment status in the year 2000 did not significantly affect the risk of increasing body weight. Reporting fair or poor health significantly increased the risk of reporting body weight increases. African Americans were also at a significantly greater risk of increasing their body weight than were whites.

Among those whose weight decreased, the mean weight-loss was 7.1 pounds (95% CI 6.6–7.6) or 0.9 BMI units (95% CI 0.8–1.0). The mean decrease in body weight did not greatly differ according to employment status in 2000; in the employed group, the mean decrease was 6.9 pounds (95% CI 6.4–7.4) while in the unemployed groups the mean decrease was 8.0 (95% CI 4.2–11.7) and 8.3 (95% CI 5.2–11.5) for those with and without benefits, respectively. Among those who increased weight, the mean increase was by 12.8 pounds (95% CI 12.2–13.4) or 1.8 BMI units (95% CI 1.74–1.94). For those employed in 2000, the mean increase was 12.6 pounds (95% CI 11.9–13.2). For those unemployed with compensation the mean increase was 11.5 pounds (95% CI 8.5–14.6) and among those without compensation the mean increase was 14.8 (95% CI 8.2–21.6). The mean increase and decrease in BMI when stratified by employment status showed a similar pattern to weight change in pounds.

We used multiple regression models to investigate possible differences among the employment groups of interest and did not find statistically significant differences in the percentage of body weight change.