Census Bureau prison miscount skews election data
by Aleks Kajstura, June 18, 2014
How well does our nation do with voter turnout? Thanks to the Census Bureau, we may not really know. Voter registration and turnout are common metrics for gauging the health of our democracy, but the numbers are easily skewed by Census Bureau methodology that counts incarcerated people in the wrong place.
A recent report, Unequal Access: A County-by-County Analysis of Election Administration in Swing States in the 2012 Election, from the Center for American Progress Action Fund, used the Census’s population data to rank counties by the percentage of the county’s population that was registered to vote, and the percentage that actually turned out to vote.
The report based its numbers on CVAP (Census Voting Age Population); CVAP gives you a rough estimate of the number of folks eligible to vote. It seems like just the sort of number you’d want to judge your turnout against. But since the Census counts incarcerated people as if they are residents of the county where the correctional facility is located, things can go terribly wrong with the math.
DeKalb County, Missouri, for example is ranked as second-worst in the state for both voter registration and turnout. But that’s because roughly a quarter of the people the Census counted there were not actually residents of the county, but rather folks incarcerated at two state prisons that happened to be located there.
The people incarcerated in those prisons are very likely not residents of DeKalb, and therefore not eligible to vote in the county (whether or not they are disenfranchised by the state). Taking the prison populations into account when looking at the voter registration and turnout, the County’s numbers look much better (middle to top of the rankings).
The simplest solution for getting accurate election data would be for the Census Bureau to just count incarcerated people at home, where they reside. In the meantime, the best practice for researches is to use the Bureau’s group quarters data to adjust Census-reported population when doing voting analyses.