The Record Linking Glass Ceiling. Applying Automated Methods to the Census and Women’s Marriage Records, 1881–1911

Abstract

This paper presents the results of a project creating a linked dataset of census and civil registration records from the county of Derbyshire, England. The proposed method includes women at every stage, first by comparing the performance of deterministic, probabilistic, and household-based linking methods and then expanding the linking process to capture women who have changed names between censuses due to marriage. In census-to-census linking the best results are obtained through a combination of probabilistic and household-based methods, linking between 40% and 45% of the starting population in each decade 1881–1911. The quality of these links and possible impacts of migration patterns are discussed with reference to the representativeness of the linked sample. Incorporating transcribed indexes of marriages (which are freely available online) allows women to be followed in the census across their marriages. Combined, this process reduces the gap in linking success between women and men and especially improves match rates for women in their twenties by between fifteen and twenty percentage points. These data have important potential for future record linking efforts and for research exploring women's work, marriage, and fertility in a life course perspective.