The European Journal of Public Health Advance Access originally published online on August 9, 2006
The European Journal of Public Health 2007 17(2):139-144; doi:10.1093/eurpub/ckl119
Health inequalities |
Are manual workers at higher risk of death than non-manual employees when living in Swedish municipalities with higher income inequality?
Göran Henriksson1, Peter Allebeck2, Gunilla Ringbäck Weitoft3 and Dag Thelle4
1 Department of Social Medicine, Göteborg University SE 405 30 Göteborg, Sweden
2 Department of Social Medicine, Karolinska Institute SE 171 76 Stockholm, Sweden
3 Centre for Epidemiology, National Board of Health and Welfare SE 10630 Stockholm, Sweden
4 The Cardiovascular Institute, Sahlgrenska universitetssjukhuset Göteborg University, SE 413 45 Göteborg , Sweden
Correspondence: Göran Henriksson, Department of Social Medicine, Göteborg University, Box 453, SE 405 30 Göteborg, Sweden, tel: +46 31 773 68 61, fax: +46 31 16 28 47, e-mail: goran.henriksson{at}socmed.gu.se
Received November 23, 2005, accepted May 2, 2006
 |
Abstract |
|---|
 Top Abstract MethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
Objectives: To test the hypothesis that manual workers are at higher risk of death than are non-manual employees when living in municipalities with higher income inequality. Design: Hierarchical regression was used for the analysis were individuals were nested within municipalities according to the 1990 Swedish census. The outcome was all-cause mortality 1992–1998. The income measure at the individual level was disposable family income weighted against composition of family; the income inequality measure used at the municipality level was the Gini coefficient. Participants: The study population consisted of 1 578 186 people aged 40–64 years in the 1990 Swedish census, who were being reported as unskilled or skilled manual workers, lower-, intermediate-, or high-level non-manual employees. Results: There was no significant association between income inequality at the municipality level and risk of death, but an expected gradient with unskilled manual workers having the highest risk and high-level non-manual employees having the lowest. However, in the interaction models the relative risk (RR) of death for high-level non-manual employees was decreasing with increasing income inequality (RR = 0.77; 95% CI, 0.63–0.93), whereas the corresponding risk for unskilled manual workers increased with increasing income inequality (RR = 1.24; 95% CI, 1.06–1.46). The RRs for skilled manual, low- and medium- level non-manual employees were not significant. Controlling for income at the individual level did not substantially alter these findings, neither did potential confounders at the municipality level. Conclusions: The findings suggest that there could be a differential impact from income inequality on risk of death, dependent on individuals' social position.
Keywords: attempted suicide, income inequality, myocardial infarction, Sweden, socio-economic position
There seems to be no strong evidence for an association between income inequality and health outside the United States and to some extent the United Kingdom.1
In a systematic review of 98 aggregate and multilevel studies examining the association between income inequality and health the authors examined the evidence for different hypotheses in this research area and after an extensive analysis they concluded that ‘there seems to be little support for the idea that income inequality is a major, generalizable determinant of population health differences within or between rich countries’.2
However, there is no reason to conclude a priori that the health threatening effect from income inequality should affect all people in a similar way. It could well be that disadvantaged groups of individuals suffer more from income inequality than do affluent groups. Few studies have explored the possibility of such a differential impact depending on the people's social position, a possible cross-level interaction. Subramanian and Kawachi3 conclude that the results from the relatively few studies exploring a possible cross-level interaction are diverging. There are some evidences suggesting that affluent people experience health benefits from living in areas with higher income inequality,4 whereas others suggest that income inequality is more detrimental for health of poor or near-poor people,5–7 and yet another suggesting that middle income non-elderly experience higher mortality risk when living in high income inequality states.6,7
There are comparatively fewer studies from Scandinavian countries on income inequality and health. We found two Danish studies,8,9 one Swedish10 and one Finnish study11 addressing this issue. Only one of the Danish studies8 approached the possibility of a differential impact but found no evidence supporting this hypothesis. Osler et al.8 explored the effect on mortality from income inequality at the parish-level within Copenhagen.
However, according to both the psychosocial12 and neo-material13 explanations for how income inequality might affect health we may expect that the effect from income inequality varies between different social positions
We wanted to test the hypothesis that manual workers are at higher risk of death than are non-manual employees when living in municipalities with higher income inequality. We explored the issue by using a large sample of employed people in Sweden.
|
Methods |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
The core dataset is based on the 1990 Swedish census, which is linked to earlier censuses thus providing data on socio-economic circumstances at the individual level on basically all Swedish citizens 1990. Data on all-cause mortality from the national cause-of-death register were linked to census data through the unique personal identification number. All-cause mortality data 1992–1998 was used.
The study population consists of all people in the 40–64 years of age in the 1990 Swedish census with the same code for parishes in the 1985 and 1990 censuses, thus presumed as having lived in the same area for at least 5 years until 1990, and who were reported as being employed in the 1990 census. Only those who were reported as being unskilled or skilled manual workers, lower-, intermediate-, or high-level non-manual employees were included in the study population. Thus, the study population comprised 1 578 186 individuals (table 1).
|
Data on socio-economic status were gathered from the 1990 census for individuals aged 40–64 years in 1990. Socio-economic groups were defined according to a classification used by Statistics Sweden, which is primarily based on occupation, but also takes the educational level of occupation, industrial sector, and position at work into account.
The income measure used at the individual level was disposable family income per unit of consumption. Even though the possible effect from income at the individual level is not of the primary interest of the study, it is nevertheless an important variable to include in the analysis in order to check for a possible ‘artefactual association’.14–16
The Gini-coefficient was used as a measure of income distribution at the municipality level. The income distribution variable was treated as a continuous variable, centred on the mean.
The size of population in each municipality, proportion of manual and non-manual workers, mean income in each municipality were used in models as possible level-2 confounders (table 2).
|
The outcome variable was death from all causes during 1992–1998.
|
Statistics |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
Individual-level data were organized in categories according to gender, 5-years age groups, income quintiles, and socio-economic status. Person-years were calculated for each individual. Follow-up time was 7 years and started in 1992; end points were death or end of follow-up time.
A two-level Poisson model was then fitted to data with counts of death as the outcome variable and log(person-years) as offset.
All analyses were carried out using MlwiN 1.10 software package.17 We allowed for extra Poisson variation (EPV) in the full model.
A random intercept model was used where individuals (first level) were nested within municipalities (second level). The ‘crude’ model consisted of income distribution, gender, and age group as explanatory variables (Model A in Appendix). In Model B socio-economic position was introduced with high-level non-manual employees as a reference group. We tried to fit a random slope model allowing slopes for the socio-economic variables to vary between municipalities. However, these models did not converge. We added interaction terms for the different socio-economic positions at the individual level and income inequality at the municipality level. Thus the full model (Model C) was
 |
This model was used to predict the incidence rates of death from all causes for each of the five socio-economic groups controlling for age and sex.
Disposable family income was also added to the full model to see whether the effect from income inequality could be explained by the level of individual income. The size of population in each municipality, proportion of manual and non-manual workers, mean income, and proportion of poor people were also added to the full model to control for confounding from these variables.
|
Results |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
There was no significant association between income inequality and mortality in the crude age and sex-adjusted model (Beta-estimates for all models is provided in the table in Appendix)
When socio-economic position was added to the model there was an expected gradient in mortality risk with unskilled manual workers having the highest risk and high-level non-manual employees the lowest. The relative risk (RR) for all-cause mortality among manual workers was 1.6 as compared with the reference group, high-level non-manual employees. The estimated RR for income inequality was not significant.
When interaction terms between socio-economic position and Gini-measure were introduced in the model the association was inverse for the reference group (male high-level non-manual employees 40–44 years). The risk of death decreased with increasing income inequality. The corresponding RR for death of a reference person living in the municipality with the highest income inequality was 0.77 compared with the reference person living in the municipality with the lowest income inequality.
The association between income inequality and mortality was the opposite for unskilled manual workers. For individuals in this group the risk of death was increasing with increasing income inequality. The RR for death among unskilled manual workers was 1.24 in the municipality with the highest income inequality compared with the unskilled manual worker living in the municipality with the lowest income inequality.
The RR for skilled manual, lower and medium non-manual employees were not significant.
The estimates from the interaction model were used to calculate the predicted death rates for the different socio-economic positions. The rates are plotted in figure 1.
|
The scatter graph describes a pattern of increasing differences in the incidence of death with increasing income inequality among the five socio-economic groups. However, the predicted slopes for skilled manual workers, lower and medium level non-manual employees are not statistically significant from zero.
The death rates for different socio-economic positions show the expected pattern with the highest incidences among unskilled manual workers and the lowest among the high-level non-manual employees.
The RR for unskilled manual workers living in the municipality with the highest inequality compared with unskilled manual workers in the municipality with the lowest income inequality was 1.2 (table 3).
|
The corresponding RR for high-level non-manual employees was 0.8. As is clear from figure 1 the RR for death of unskilled manual workers compared with high-level non-manual employees living in the municipality with the lowest income inequality was
1.4 whereas the corresponding RR for the same groups living in the municipality with the highest income inequality was 1.9.
As is seen in figure 1 values on income inequality are unevenly distributed with an outlier corresponding to the municipality with the highest income inequality (Gini = 0.32) and all others with a range from 0.18 to 0.25. However, leaving this municipality out of the analysis did not alter the estimates appreciably.
Introducing individual-level family income into the full model reduced the beta-estimates of income inequality and socio-economic position as well as the interaction terms. As an illustration the beta estimate for income inequality was reduced from –2.002 (SE 0.746) to –1.674 (SE 0.755). However, the estimates were still statistically significant and for the sake of clarity this model is not shown in tables or figure.
None of the possible level-2 confounders was found to affect the estimates of income distribution or the interaction terms in any significant way.
|
Discussion |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
We found an interaction between individual-level social position and income inequality in Sweden. We think that from a theoretical point of view as well as from a policy purpose the findings are interesting.
Daly et al.7 investigated the effect from income inequality on mortality in a sample of 6500 persons living in the United States. They elaborated with inequality measures indicating different aspects of income inequality. Percentile quotients were used to investigate effects from overall dispersion, dispersion at the lower end, and dispersion at the higher end of income spectrum between 1980 and 1990. They used a single-level logistic regression model. They found a negative effect of income inequality in the US states on mortality only for individuals with middle range incomes.
When stratifying on income levels Kennedy et al. found that individuals with incomes <USD 20 000 and living in the US states with high income inequality had a higher odds ratio (OR) for reporting fair or poor self-rated health as compared with high-income individuals (>USD 35 000) living in states with high versus low income inequality.6
Kahn et al.5 studied maternal depressive symptoms among 8060 women in the US states. Gini-coefficient was used as a measure of income inequality at the state level. Single-level regression models were used to estimate the effects on depressive symptoms and self rated health. They found that low-income women in states with high-level income inequality had a higher risk of depressive symptoms as compared with low-income women living in states with the lowest income inequality. OR was 1.7 (95% CI, 1.0–2.6).
Subramanian et al.4 did not find any support for an association between income inequality and low income at the individual level. However, they found some evidence that high-income groups had a lower probability for reporting poor health when living in a high-level income inequality state.
In the Scandinavian countries Osler et al.8 explored the relation between income inequality at the level of small residential areas in Copenhagen and mortality after adjusting for individual income and other potentially important characteristics such as lifestyles, housing structures, and education, were controlled for. The study comprised of nearly 26 000 men and women who were followed up for a mean of 12.8 years. When analysis was stratified for individual-level income quartiles there was no difference between high- and low-income inequality within residential areas.
Regarding our study, it could be discussed whether it would be more appropriate to choose individual income as the individual-level measure to prevent biased estimates on income inequality measure. However, our main interest was in exploring the possibility of a differential impact from income inequality on mortality depending on people's social position. We think that the Swedish socio-economic index is a better proxy for the social position than individual income, at least among those being employed. This is also the reason why the study population is restricted to those reporting as being employed.
However, it would be a fallacy to conclude that our results support the hypothesis that there is, after all, a causal relation between income inequality and mortality even in the Swedish society. The results are merely statistical associations. We would suggest that our results perhaps challenge the general impression that there is no evidence for an association between income inequality and health outside the United States and the United Kingdom.
Income inequality in Sweden is principally made up by high incomes. The 90th income percentiles in the municipalities with high income inequality are much higher than the 90th percentiles in the municipalities with the lowest income inequality. In 1990 the 90th income percentile (measured as disposable family income) was about SEK 140 000 in the municipalities with lowest income inequality whereas in the municipalities with the highest income inequality it was about SEK 240 000. The corresponding 10th income percentiles were SEK 60 000 and 70 000, respectively. This means that richer people tend to live in high-level income inequality municipalities. If there is a clustering of rich people in certain municipalities there could also be different qualities and/or quantities of welfare institutions. It has been hypothesized that affluent people when living in the same residential areas tend to attract resources to their areas.18
An important question is whether the choice of municipalities in Sweden is the appropriate level for measuring income inequality. A recent review supports the suggestion that income inequality measured at larger areas serves better as a measure of social stratification than measured at smaller areas.19 This is a relevant remark since many municipalities in Sweden are small. Our results seem to support Wilkinsons and Picketts suggestion. The smaller municipalities tend to be more equal regarding income, even if the variation within the Gini quintiles is large (table 2).
These considerations indicate that it is important to perform studies on the different municipalities and the way people organize in different parts of municipalities, the way welfare institutions are organized but also stratification processes and policy development concerning welfare investments within municipalities.
|
Conclusion |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
We could find no provable association between income inequality as a general detrimental factor and health in Sweden but some indication that income inequality could be a detrimental factor for disadvantaged groups. However, these findings suggest that mechanisms and distinctive features such as stratification processes within municipalities are important to consider to understand if and how income inequality links to health.
Conflict of Interest: None declared.
| Key points
|
|
Appendix |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
The ß-estimate for income distribution, Gini, was –0.8842. The ß-estimate in the Poisson model corresponds to the antilog of RR when the exposure variable is increased with 1 unit.
In this case we also have to consider that the Gini-coefficient varies between municipalities with the lowest value = 0.182 and the highest = 0.311, thus the range equals 0.129. From this we can calculate the RR of mortality for the reference person (male 40–44 years) living in the municipality with the highest income inequality as compared with a reference person living in the municipality with the lowest as
 | (A1) |
|
|
References |
|---|
|
TopAbstractMethodsStatisticsResultsDiscussionConclusionAppendixReferences |
|---|
1 Mackenbach JP. (2002) Income inequality and population health. BMJ 324:1–2.
2 Lynch J, Smith GD, Harper S, Hillemeier M, Ross N, Kaplan GA, et al. (2004) Is income inequality a determinant of population health? Part 1. A systematic review. Milbank Q 82:5–99.[CrossRef][Web of Science][Medline]
3 Subramanian SV and Kawachi I. (2004) Income inequality and health: what have we learned so far? Epidemiol Rev 26:78–91.
4 Subramanian SV, Kawachi I, Kennedy BP. (2001) Does the state you live in make a difference? Multilevel analysis of self-rated health in the US. Soc Sci Med 53:9–19.[CrossRef][Web of Science][Medline]
5 Kahn RS, Wise PH, Kennedy BP, Kawachi I. (2000) State income inequality, household income, and maternal mental and physical health: cross sectional national survey. BMJ 321:1311–5.
6 Kennedy BP, Kawachi I, Glass R, Prothrow-Stith D. (1998) Income distribution, socioeconomic status, and self rated health in the United States: multilevel analysis. BMJ 317:917–21.
7 Daly MC, Duncan GJ, Kaplan GA, Lynch JW. (1998) Macro-to-micro links in the relation between income inequality and mortality. The Milbank Quarterly 76:315–39.[CrossRef][Web of Science][Medline]
8 Osler M, Prescott E, Gronbaek M, Christensen U, Due P, Engholm G. (2002) Income inequality, individual income, and mortality in Danish adults: analysis of pooled data from two cohort studies. BMJ 324:1–4.
9 Osler M, Christensen U, Due P, Lund R, Andersen I, Diderichsen F, et al. (2003) Income inequality and ischaemic heart disease in Danish men and women. Int J Epidemiol 32:375–80.
10 Gerdtham U-G and Johannesson M. (2004) Absolute income, relative income, income inequality, and mortality. J Hum Res 39:228–47.
11 Blomgren J, Martikainen P, Makela P, Valkonen T. (2004) The effects of regional characteristics on alcohol-related mortality—a register-based multilevel analysis of 1.1 million men. Soc Sci Med 58:2523–35.[CrossRef][Web of Science][Medline]
12 Wilkinson RG. (2000) Inequality and the social environment: a reply to Lynch et al [comment]. J Epidemiol Community Health 54:411–3.
13 Lynch JW, Smith GD, Kaplan GA, House JS. (2000) Income inequality and mortality: importance to health of individual income, psychosocial environment, or material conditions. BMJ 320:1200–4.
14 Gravelle H. (1998) How much of the relation between population mortality and unequal distribution of income is a statistical artefact? BMJ 316:382–5.
15 Judge K, Mulligan JA, Benzeval M. (1998) Income inequality and population health. Soc Sci Med 46:567–79.[CrossRef][Web of Science][Medline]
16 Rodgers GB. (2002) Income and inequality as determinants of mortality: an international cross-section analysis. 1979. Int J Epidemiol 31:533–8.
17 Rasbash J, Browne W, Goldstein H, Yang M, Plewis I, Healy M, et al. A user's guide to MLwiN. In. 1.10 ed. London: Multilevels Models Project, Institute of Education, University of London; 2000.
18 Massey DS. (1996) The age of extremes: concentrated affluence and poverty in the twenty-first century. Demography 33:395–412 discussion 413-6.[Web of Science][Medline]
19 Wilkinson RG and Pickett KE. (2006) Income inequality and population health: A review and explanation of the evidence. Social Science & Medicine 62:71768–1784.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||