Information across three generations


I’m wondering if there is a way to link individuals, or households, across three generations using the IPUMS-USA linked files. I understand that seven census years are linked as best as possible. But I wasn’t sure if I could link multiple generations in a similar way.


The Linked Samples that you are referring to us the 1880 complete count data as a starting point and then use birth year, place of birth, given name, surname, and race to link records to the previous and subsequent census samples that had this information available. The linked data is organized as seven pairs of years, with 1880 always being one of the years in the pair. This is because it is incredibly hard, even with name and birth information, to link a significant number of cases across samples, since there is no guarantee that any given person will exist in both samples. As to what this means for linking multiple generations, all three generations would need to be present in 1880 to be able to link each generation backwards and forwards in time. Furthermore, in order to establish family links the family members need to live in the same household. So, it is theoretically possible to link multiple generations, however the population would be fairly specific (3+ generation households in 1880), and if the household splits up, it would be very unlikely that you would find each generation in the Census samples.

IPUMS-USA is currently harmonizing complete count data from 1790-1940. The resulting public files will not have names which will make linking impossible, but we do have a limited number of restricted access licenses for scholarly research using internal files. These files do contain names as well as other string variables that would be useful for linking across time. For more information, send your inquiry to

I hope this helps.