Digitization of the 1990 U.S. Census Population Records

John W. Sullivan, US Census Bureau
Huaigu Cao, University of Southern California
Jonathan Morgan, U.S. Census Bureau
Katie Genadek, University of Colorado Boulder
Trent Alexander, University of Michigan

We describe our work digitizing 1990 U.S. Census records to support the creation of a new linked data file. While almost all variables from the 1990 Census were digitized shortly after the data was collected, respondent names were not. In order to capture respondent names, we needed to digitize the original census images, recover names, and append those names onto the original microdata containing the rest of the variables. This paper documents four key processes in that work: (1) creating high-quality digital scans of nearly 120,000 reels of microfilm, (2) carrying out Optical Mark Recognition on a set of “bubbled” items from the manuscripts, (3) carrying out automated Handwriting Recognition on 248 million individual respondent names, and (4) appending the names back onto the original 1990 data files created shortly after the census.

No extended abstract or paper available

 Presented in Session 203. Digitization and Linkage of the 1990 United States Census