![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
![Publish Open Access Finland]({"type":"image" , "src" : "/sda/110396/JournalAd-OA-Agreements.gif"})
Abstract
The U.S. Census Bureau has created a set of linkable census, survey, and administrative records that provides longitudinal data on the American population across the past eight decades. While these files include modern decennial censuses, Census Bureau surveys, and administrative records files from other federal agencies, the long time span is only possible with the addition of the complete count 1940 Census microdata. In this paper, we discuss the development of this linked data infrastructure and provide an overview of the record linkage techniques used. We primarily focus on the techniques used to produce a beta version of a linkable 1940 Census microdata file and discuss the potential to further document and extend the infrastructure.
Acknowledgements
The authors gratefully acknowledge support from Constructing Life Course and Intergenerational Data Through Censes Data Linkages (R21 HD087914) and the Minnesota Population Center (P2C HD041023), both funded from the Eunice Kennedy Shriver National Institute for Child Health and Human Development (NICHD). We also thank the participants in the 2015 Minnesota Population Center Record Linkage workshop and the 2015 Population Association of America Annual Meetings for their comments and suggestions. This research was conducted as a part of the Census Longitudinal Infrastructure Project (CLIP). Trent Alexander and Catherine Massey conducted this work while employees of the U.S. Census Bureau. Any opinions and conclusions expressed herein are those of the authors and do not necessarily represent the views of the U.S. Census Bureau.
Notes
1 Transactions occur when corrections or name changes are made on a record of a particular SSN. This allows us to observe both married and maiden names for women. The average number of transactions per SSN is 2.1 (Harris 2014).
2 The PVS string comparator was developed by Winkler (1995) and measures the distance between two strings on a scale from 0 to 900, where a distance score of 0 is given if there is no similarity between two text strings and a score of 900 is given for an exact match. The cutoff value for the string distance is set to 750 in the Name Search module.
3 In modern linking conducted at the Census Bureau, linkages to the corresponding tax year provide addresses in the reference file (e.g. 1999 tax year returns are used to PIK the 2000 Census). We are unaware of any digitized historical administrative data that would provide a reference point for addresses in the 1940 Census.
4 We use Black et al. (2015) GNIS codes for city of birth listed in the Numident. We then aggregate city up to the county level for the match.
5 Although address stings are available for many records in the 1940 Census, we do not observe addresses for observations in the reference file. Consequently, 1940 respondent address is not useful for PIK assignment.
6 Outside of the Census Bureau, many scholars link person records across public (pre-1950) decennial census data (for example, see Abramitzky et al. 2014; Collins and Wanamaker 2014; Ferrie 1996; Long and Ferrie 2013). At the forefront, the IPUMS project produced the most comprehensive linkages of individuals across the 1850–1930 U.S. censuses with their IPUMS Linked Representative Samples (Goeken et al. 2011; Ruggles, 2011). These linkages began with individuals in the 1880 complete-count census file linked to the other earlier censuses to produce seven pairs of linked samples. The IPUMS is expanding their historical linkages to the full-count historical population censuses to construct the Multigenerational Longitudinal Panel (IPUMS-MLP). A collaborator of IPUMS-MLP, the Longitudinal, Intergenerational Family Electronic Micro-database (LIFE-M) project, uses historical vital records to improve the linkages of women and minorities across the complete-count 1880–1940 censuses (Bailey et al. 2017). Similarly, the Early Indicator’s project links Union Army enlistees and their families across numerous decennial censuses (Costa et al. 2017).