Long-Term Illness as an Occupational-Career Interruption: Gender Differences in the Determinants and Outcomes in Sweden

Abstract

This paper examines one particular type of occupational-career interruption; long-term illness. Event history analysis is used to uncover gender differences in the factors that influence illness work absenteeism and regression for the consequences of this type of interruption in Sweden, a country with a generous welfare state and high level of gender equality. Analysis of a panel survey (HUS) attests to gender differences in both the determinants and the effects of illness interruptions. For example, having a child in the household and the absence of a spouse or cohabiting partner significantly raise men's likelihood of experiencing an illness interruption, but is non-significant for women. With regards to the consequences of an illness interruption, women suffer significant losses in wages, while men experience no significant effect. For women, working full-time and working in a blue-collar job influence wages at a significant level. The paper concludes with a discussion of the implications of such findings and ends with policy implications.

Acknowledgments

The author is indebted to Kazimierz M. Slomczynski for guidance and expert advice in the formulation and revision of this topic and on previous versions of this paper to the Scandinavian Journal of Disability Research guest editor, and to the three anonymous reviewers for their helpful advice and suggestions.

Notes

1. It is important here to note that I am not measuring whether a person will become ill; instead, I focus on whether they are removed from the labor force due to their own illness.

2. The exact penalty for an illness varies depending on the type of health condition and the type of health indices used (Chirikos 1993); however, the type of illness is not distinguished in this paper because all illnesses can be construed as long-term, hence the type of illness matters less.

3. These waves were selected due to the similarity in the questions asked.

4. 1993 was chosen for all of the variables (except log of wages in 1998 – the dependent variable in Table 4) because this signifies the starting point in which an interruption could occur.

5. An interaction for private sector job in 1993 and child in the household in 1993 was also included in the analysis, but was removed due its non-significant effect.

6. To make the interpretation more meaningful, I calculate the (e^beta−1)*100 value for the hazard ratios.

7. Age and age squared cannot be included in the model due to multi-collinearity problems with the duration variable.

8. Interactions for illness and a college degree, and illness and working full-time, were also included in previous models but were removed because of non-significance.