Why don't the American Community Survey (ACS) Public Use Microdata Sample (PUMS) file estimates match American FactFinder (AFF) estimates?
The American Community Survey (ACS) Public Use Microdata Sample (PUMS) files are intended to allow data users to conduct custom analyses when the desired estimates are not already available in American FactFinder (AFF).
If users attempt to recreate AFF tables with PUMS files, the estimates generated will be slightly different from the pretabulated estimates. These differences are due to the fact that the PUMS files include only about two-thirds of the cases that were used to produce estimates on AFF.
The Census Bureau also makes additional modifications to the PUMS files to protect the confidentiality of respondents. These modifications can include, for example, the "top-coding" of continuous variables with outlying values (such as income or transportation time) and the reduction of category availability for variables with hundreds of categories (such as ancestry or birthplace). These changes can also cause PUMS estimates to differ slightly from those in AFF.
When data users have doubts about whether they are correctly using the weights to compute estimates, they should attempt to reproduce the estimates that are provided in the PUMS Estimates for User Verification. (The standard errors provided were computed using the replicate weight method.)