We now turn to Oaxaca-Blinder style decompositions (Blinder, 1973; Oaxaca, 1973) to analyse to what extent these compositional changes in the workforce can explain the changes in the prevalence of different employment types between 2001 and 2017. This regression-based method allows decomposing the difference in the share of each employment type between two selected years into one part that is due to differences in worker and job characteristics (the ‘explained’ part) and one part that is due to differences in the impact of these worker and job characteristics on the prevalence of each employment type (the ‘unexplained’ part, or what could also be described as behavioural change).