A New Strategy for Linking U.S. Historical Censuses: A Case Study for the IPUMS Multigenerational Longitudinal Panel.

Area of Science:

• Digital Humanities
• Historical Demography
• Computational Social Science

Background:

• Accurate record linkage is crucial for historical demographic research.
• Existing methods for linking digitized historical records have limitations in accuracy and coverage.
• U.S. full count censuses of 1900 and 1910 provide a rich dataset for developing and testing linkage methods.

Purpose of the Study:

• To present a novel probabilistic method for record linkage applicable to digitized historical records.
• To improve linkage rates and accuracy compared to existing methods.
• To demonstrate the method's utility using the 1900 and 1910 U.S. censuses.

Main Methods:

• A two-step probabilistic record linkage approach was developed.
• The first step links men using individual and contextual characteristics for high confidence matches.
• The second step leverages household links from the first step to include women, improving overall linkage.

Main Results:

• The proposed method achieved considerably higher linkage rates than comparable methods.
• The method demonstrated greater accuracy in linking records.
• The linkage results showed negligible differences in resembling the target population compared to other algorithms.

Conclusions:

• The developed probabilistic method offers a significant advancement in record linkage for historical data.
• This approach enhances the ability to study historical populations with greater precision and completeness.
• The method is broadly applicable to various sources of digitized historical records.