Abstract
This paper uses data on 11 EU countries to explain cross-national differences in internal migration rates. It is found that 89% of the variance in gross migration in current member states can be explained by variations in employment protection, international migration, the share of ownership occupied housing and the average region size of a country. Results concerning net migration rates suggest that although disparities in unemployment are important determinants of net migration, so are employment protection, long-term unemployment and the share of owner occupied housing.
Acknowledgements
This research was supported by the European Commission 5th Framework Programme – ACCESSLAB project, the author thanks the participants of the ACCESSLAB workshops in Budapest and Bonn for comments on earlier versions of the paper.
Notes
See an Appendix available from the author for a detailed data description.
The share of migration among contingent regions may be influenced by a country's geography. Thus we measure the share of migration among contingent regions relative to the share of contingency relationships in a country. In a country with n regions there are n(n–1) pairs of sending and receiving regions (migration within the region is not measured). If m of these region pairs are contingent the share of contingency relationships (s) is given by s = m/n(n−1).
The gross migration rate is the share of population of a country changing place of residence within a year, the net migration rate is the sum (across all regions) of the absolute values of the difference between in and out migration in a region (Fredriksson, Citation1999). Differences between these two measures are due to churning.
This can be explained by risk-averse workers preferring security in existing employment to migration with uncertain prospects when unemployment is high everywhere, and losses in search effectiveness of long-term unemployed (due to human capital losses or discouragement effects).
This last effect is ambiguous, however, since more regions close to each other could also decrease migration if commuting is a substitute to migration.
This is defined as in Footnote 3 above.
Doppelhofer et al. (Citation2000) advise equal priors for all models when the number of potential regressors is small, as in our case.