Pekka Ilmakunnas
International Journal of Manpower, Vol. ahead-of-print, No. ahead-of-print, pp.-
This study examines the relationships between age, education or gender dissimilarity and movement from one workplace to another, examining different dissimilarity measures and asymmetries in these relationships.
Large-scale employer–employee register data from Finland were used to estimate discrete time duration models for the probability of job-to-job exits from plants. The alternative dissimilarity measures were the Euclidean distances for age and education and the shares of opposite gender, age and education groups.
When the Euclidean distance is used as the dissimilarity measure, age dissimilarity is negatively related to workplace exits; however, age dissimilarity is positively related to exits for young women. Educational dissimilarity, meanwhile, is positively related to exits. When the share of opposite groups is used, the results for age and educational dissimilarity depend on how the opposite age and educational groups are defined. The share of women is positively related to the probability of job change among men, but for women, the share of men negatively affects exits.
Identification relied on the assumption that unobservable individual characteristics can be sufficiently approximated using within-individual averages of the variables.
Researchers should conduct extensive sensitivity analyses and allow for asymmetries in workplace relational demography research.
Only a few previous studies used large-scale datasets to estimate the effects of dissimilarities on turnover, and those studies did not systematically compare different methods of measuring dissimilarities.