Magnus Bygren 2013
SAGE journals Acta Sociologica 2013 56: 3
The way employee flows generate ethnic and gender segregation across workplaces is investigated using a population sample of 80,139 workplaces with 977,978 employees in the Stockholm area. Comparisons of actual stocks and flows of employees across workplaces to counterfactual simulations of these reveal that segregation clearly has a random component to it: Even with random allocation of employees to workplaces, segregation would still be substantial. Systematic (non-random) segregation appears to be upheld primarily because employees recruited to workplaces are similar to those already employed there, not because underrepresented groups within workplaces are systematically screened out. This tendency appears to be less connected to between-group differences in education, occupation or industry, but instead largely sustained by the tendency of employers to select new employees from a pool of workplaces where their employees have been employed previously. Network recruiting might generate this pattern, but unobserved individual and workplace factors cannot be ruled out as potential confounders. The results speak to theories of homosociality applied to segregation processes: If homosocial biases affect segregation, they apparently do so mostly in the recruitment process to workplaces, but less so through processes of exclusion of minorities from workplaces.