Should ACS person weights sum exactly (within state) to census state population estimates?


I’m analyzing various years of the single-year ACS for california. I’ve noticed that the ACS weights when summed over all observation come very close to the official California population estimates, but they are not exact. Would you know if weights are revised for intercensus ACS samples after the 2010 census? Have others encountered this problem, and is there a simple fix (perhaps rescaing the weights so that they add to census popuation estimates for the state)?


Steve Raphael



The Census Bureau’s Population Estimates Program (PEP) generates the official population estimates for states and other geographies. These estimates are updated annually for all years after the most recent decennial census. In contrast, the ACS PUMS files are static and do not reflect these revisions. Since the PUMS data are intended to be representative at the state level, you might consider calculating the weighted percentage frequency for your variables of interest using the IPUMS microdata. Then, apply this percentage to the total population reported by the PEP.

Hope this helps.