Aggregate Turnout Is Mismeasured

Abstract

We reveal errors in aggregate-level turnout measures in the US, UK and Sweden. In many cases, errors in measurement exceed substantive effects for institutional and event-based explanations of voter turnout: We find: 1) Errors in the vote numerator in the IDEA database sufficient to reverse the 1980-2016 trend in US VAP turnout; 2) VAP and VEP are increasingly diverging because of higher rates of non-eligible immigrants; 3) errors in the registration denominator in the IDEA database large enough to create an erroneous 18 percentage point fall in US RV turnout between 2004 and 2008, and 4) underestimation of RV turnout due to inaccuracies in the voter register up to 11.5 percentage points in the UK and 20 percentage points in the US. In sum, aggregate-level turnout measurement is so flawed that current claims about levels and explanations of turnout using them are practically meaningless.

Keywords: turnout, vep, vap, idea, aggregate, comparative politics, error, measurement, bias