Introduction The contribution of work-related factors – such as employment sector and occupational exposures – to sickness absence is known to be strong. In addition to the potential causal mechanisms, the associations may also be affected by selection, i.e. individuals with characteristics associated with a high likelihood of sickness absences ending up in particular types of jobs.

Objectives We aimed to investigate the influence of unobserved individual characteristics in explaining the effect of work-related factors on full (fSA) and part-time sickness absence (pSA).

Methods We used register-based panel data for the period 2005–2016 on a 70% random sample of the Finnish working-age population. The relationships between employment sector and occupational exposures (% exposed to physically heavy work and job control score based on job exposure matrices) and the annual onset of fSA and pSA were investigated among men and women. First random effects (RE) models were applied controlling for observed sociodemographic factors and then fixed effects (FE) models, that examine within-individual changes over time and thereby further account for unobserved time-invariant individual characteristics.

Results In the RE analyses, public employment sector, physically heavy work and lower job control each increased the use of fSA and pSA among both genders. When unobserved individual characteristics were controlled for with the FE models, the effects on fSA attenuated. With pSA the effects of employment sector and physical heaviness of work among women even became reverse. The effect of lower job control on pSA remained specially among women.

Conclusions The role of individuals’ unobserved characteristics in explaining the effect of work-related factors on sickness absence should not be neglected. The effects of work-related factors are likely to be overestimated when using traditional approaches that do not account for unobserved confounding, i.e. selection of individuals with a high likelihood of sickness absence into particular work environments.