Belgium witnessed, as other European countries, profound socio-economic changes during the past forty years, including the de-standardisation of the Fordist employment model. This transition has been accompanied with rising unemployment levels and stronger links between unemployment rates and economic conjuncture. During the latest economic crisis, unemployment has increased considerably. In 2012, there were 10.7% (18.7 million) unemployed persons in the Eurozone and 7.5% unemployed persons in Belgium. Considerable regional differences exist in this respect. In 2011, unemployment rises to 16.9% in the Brussels-Capital Region compared to 9.5% in the Walloon Region and 4.3% in the Flemish Region. Long-term unemployment has become more common and now also affects the higher educated and the higher socio-professional groups. Research into the mortality effects of joblessness is thus timelier than ever.
At the individual level, employment status is a key factor determining the financial and psychological well-being of individuals and their families. Despite the availability of data in Belgium, there has been very limited research in this area. To fill this gap, WP2 wishes to probe into the association between unemployment and mortality considering the following research questions:
Long-term trends in mortality differentials by employment status from 1970 to 2006
How have inequalities by employment status evolved since the 1970s in Belgium? Can the observed relationship be related to the cyclical evolution of unemployment in Belgium? Stated otherwise, can inequalities be related to the share of the unemployed in the total population? What is the pattern for men and women, did inequalities evolve in the same way? This analysis will use census data from 1970 to 2001, each census being linked to mortality data during a five-year follow-up.
In-depth analysis of inequalities in cause-specific mortality by employment status and employment contract, 1991-2010
The association between unemployment status and cause-specific mortality will be investigated through rate differences and rate ratios. The population will be stratified for several ascribed characteristics such as gender, region, nationality and age, distinguishing four age categories: (1) young adults aged 20-29 (2) adults aged 30-39 (3) middle-aged adults aged 40-54 and (4) middle-aged adults aged 55-64. Rate differences will be calculated and absolute differences in overall mortality will be decomposed, calculating the share of each cause (group) of death to inequalities in overall mortality. Rate ratios will be investigated through survival analysis (Poisson regressions). Selection bias may exist due to socio-economic factors – educational level, housing conditions and living arrangements. Adjustments will be made for such characteristics in regression models. In addition, variables will be cross-classified to probe into interaction effects and as such into the mechanisms of inequalities. Do unemployed persons with higher education or good quality housing have better life chances than those with primary or secondary education or those living in bad quality housing? If yes, for which causes and what does this tell us regarding the mechanisms of inequalities by unemployment?
Getting people into work is of critical importance for reducing health inequalities. However jobs need to be sustainable as well. Besides employment status, the 2001 census provides information on the type of contract (full time, part time, seasonal work, interim work…). Including this variable will provide additional insights into the role of (non-standard) employment conditions.
Longitudinal analysis of mortality differentials according to employment trajectories
This part of WP2 focuses on the impact of employment duration. The data allow a comparison of employment status in the 1991 and the 2001 census. The survival analyses will be repeated for specific subgroups – for those unemployed at both censuses, for those employed in 1991 but not in 2001, for those unemployed in 1991 and employed in 2001 (effect of re-employment) and for those employed at both censuses.
WP2 will also use longitudinal data originating from the Crossroads Bank for Social Security (CBSS). This database allows a reconstruction of employment trajectories and gives the opportunity to fine tune the indicators of employment in the censuses in terms of duration of unemployment. Mortality can be related in survival regressions to these and other indicators that are unavailable in the census data (such as income class). The effect of re-employment can be integrated too. Risk models such as the Poisson model will be used including duration of unemployment. These survival regression models do not take the life course into consideration. Sequence analysis offers an alternative here. This strategy will help identifying typical trajectories of employment careers, which can then be related to mortality. It will thus be possible to evaluate how the instability of employment trajectories is related to differentials in mortality.