Does IPUMS-International add value to the data?
The process of integration itself adds value to the data by fully documenting all codes and compiling all variable documentation in a hyperlinked web format. But we do many other things as well:
Most census samples have a certain number of structural errors: the roster for a household might be incomplete, members from different households might be merged together, person records might not have corresponding household records, and so forth. All IPUMS samples are processed using a consistent set of diagnostic tools to uncover such problems. Even most samples previously in public circulation through other outlets have a small number of such errors. We fix them, either by sampling around bad cases, by performing logical edits, or by using whole-household substitution.
IPUMS-International creates a consistent set of constructed variables for all samples. Most important are the family interrelationship “pointer” variables that indicate the location within the household of every person’s mother, father, and spouse.
In the future, IPUMS will carry out missing data allocation and consistency edits on important variables. This kind of data editing performs logical fixes when possible, or it involves finding a donor record that shares key characteristics with the person in question and substituting their response for the variable. Allocation is a more statistically sound way to deal with missing data than simply excluding such cases from analyses. Missing data allocation is technically demanding, and we do not expect to carry it out in the near future.