IPUMS Multigenerational Longitudinal Panel: Version 2.0

The IPUMS Multigenerational Longitudinal Panel project links individuals' records between full count censuses from 1850 to 1950. IPUMS disseminates full count census enumerations for nine census years from 1850 to 1950. These full count data, covering over 800 million individual records, are the fruit of collaboration between IPUMS and the world's two largest genealogical organizations --Ancestry.com and FamilySearch-- to leverage genealogical data for scientific purposes. Full count data have opened the possibility of automated record linkages across census years to construct millions of individual life histories and trace millions of families over multiple generations. This version of IPUMS MLP links individuals across 10, 20, and 30 year census intervals and chains them together to create life histories as long as 80 years. The 1880-1950 census data are also linked to the Social Security NUMIDENT files containing records for persons who had died or would have reached 110 years of age by the end of 2007.

The overall linking strategy was similar to version 1.2, but incorporates a number of improvements to the training data and the replacement of logistic regression with the XGBoost linking algorithm. More importantly, women are added to the first stage of the linking process, so they are no longer dependent on the second (household) linking stage. This results in many more links for women. A new third stage was also added to leverage the Social Security information, which supplies women�s birth and married surnames. This is used to link women across decades in which they become married and their last name changes.

The data are disseminated through a new interface in the IPUMS USA system that allows users to select any combination of census pairs and receive an extract containing those records. The system is a significant advance over the one introduced with version 1.1.