Do working conditions contribute differently to gender gaps in self-rated health within different occupational classes? Evidence from the Swedish Level of Living Survey

Abstract

Socioeconomic inequality in health among women is often referred to as smaller than health inequality among men. However, we know less about differences in health between men and women within the same socioeconomic groups. In this article the lack of attention to potential socioeconomic variation in gender health inequality is argued as unfortunate, as it can obscure how mechanisms, such as e.g. working conditions, affect gendered health within specific groups. Drawing on the nationally representative Swedish Level of Living survey (LNU), class/gender interactions as well as class-separate linear probability models are estimated to explore relationships between working conditions and health among men and women with the same occupational class positions. Results show that, although class is not a large explanatory factor for general gender differences in health, there are varying within-class differences between men and women in working conditions, that can contribute to the understanding of within-class gender differences in health. This highlights that, when targeting causes of gender health inequality, it is important to consider not only what class means for women as well as for men, but also what gender means within specific classes.

Fig 1. Gender differences in physically strenuous work, by class.

Relative differences (percentage share women/percentage share men). Differences >1 indicate that the condition is more common among women than among men, differences <1 indicate that the condition is more common among men than among women.

Fig 2. Gender differences in psychosocially demanding work, by class.

Relative differences (percentage share women/percentage share men). Differences >1 indicate that the condition is more common among women than among men, differences <1 indicate that the condition is more common among men than among women.

Fig 3. Predicted probabilities of less than good SRH, by gender and class.

Marginsplot based on linear probability model with mutual adjustment for gender and class (Table C in S3 Appendix).

Fig 4. Gender differences in probability of less than good SRH, by class.

Coefficients (for woman = 1) interpreted as percentage point change in probability for less than good SRH. From class-separate linear probability models, with 95% confidence intervals (Table F in S4 Appendix). Unadjusted model regresses SRH on gender, separately by class. Stepwise adjustment for physical working conditions (Model 1) and psychosocial working conditions (Model 2). Both unadjusted and adjusted models are controlled for age and age^2.

Fig 5. Predicted probabilities of musculoskeletal pain, by gender and class.

Marginsplot based on linear probability model with mutual adjustment for gender and class (Table D in S3 Appendix).

Fig 6. Gender differences in probability of musculoskeletal pain, by class.

Coefficients from class-separate LPM. Coefficients (for woman = 1) interpreted as percentage point change in probability for musculoskeletal pain. From class-separate linear probability models, with 95% confidence intervals (Table G in S4 Appendix). Unadjusted model regresses musculoskeletal pain on gender, separately by class. Stepwise adjustment for physical working conditions (Model 1) and psychosocial working conditions (Model 2). Both unadjusted and adjusted models are controlled for age and age^2.

Fig 7. Gender differences in probability of psychiatric distress, by class.

Coefficients from class-separate LPM. Coefficients (for woman = 1) interpreted as percentage point change in probability for psychiatric distress. From class-separate linear probability models, with 95% confidence intervals (Table H in S4 Appendix). Unadjusted model regresses psychiatric distress on gender, separately by class. Stepwise adjustment for physical working conditions (Model 1) and psychosocial working conditions (Model 2). Both unadjusted and adjusted models are controlled for age and age^2.