Register-based estimates of parents' coresidence in Sweden, 1969-2007

Background: Many of the dramatic changes in family formation and dissolution observed in wealthy countries over the past 60 years are tracked through vital statistics or censuses. The signature change in family behavior -- non-marital cohabitation -- is not, however, registered in most settings. Objective: We evaluate the quality of new register-based estimates of parents' union status at birth and of separation during the childrearing years. Methods: Parents of a common child are identified through the Multi-Generation Register that links each child to each parent and therefore each parent to each other. The Total Population Register identifies the property at which each parent is registered at the end of each year. We use the five-year censuses 1960-1990 as one standard of comparison because the censuses identify the dwelling unit for each parent on the census date. Results: Property-based estimates of parents' coresidence compare very well to census reports. Register-based estimates are virtually identical with those produced from the 1992 Swedish Fertility and Family Survey; differences between register estimates and those produced from the 1991 and 2000 Level of Living Survey can be explained by differences in measurement of marriage and cohabitation. Conclusions: Estimates of parents' cohabitation based on annual, property-level registration are of sufficient quality for their use in substantive analyses of union status at birth and parents' separation in Sweden. Comments: Although register-based estimates of parents' coresidence at a child's birth or afterwards can be generated only for a select group of countries, their use can be fruitful for understanding more general processes of family change. Centralized administrative registers exist in many countries but have not been made fully available for research therefore losing much of the potential value.