Since the Census Bureau first collected data on voting in the Current Population Survey (CPS) in 1964, the estimates of voter turnout have almost always been higher than those reported by election officials. In that year, the Census Bureau estimate exceeded reported turnout by 6 million out of 70 million votes cast (Powers and Dodge 1965). The Census Bureau estimates have continued to be higher than official tallies since that time, though they vary from year to year and state to state (Bauman and Julian 2010). CPS is not alone in overestimating voter participation. Researchers have found the problem to universally affect survey-based estimates of voting in the United States and around the world (Clausen 1968, Selb and Munzert 2013).