Blame it on turnout? Citizens’ participation and polls’ accuracy

Abstract

When pre-election polls fail, citizens make choices in an environment where the information is inaccurate. This is bad for democracy. Understanding the conditions under which polls succeed or fail is thus very important for the quality of democracy. Polling firms have often blamed voter turnout when they failed to provide accurate information. There is, however, no systematic test of the impact of voter turnout on polling errors. Using data from 2104 pre-election polls in 206 elections among 33 unique countries from 1942 to 2017, I test whether polling firms have legitimate reason to blame their errors on turnout. Results systematically fail to provide evidence that the quality of pre-election forecasting is a function of voter turnout. This research entails important implications for our understanding of polls’ capacity to predict electoral outcomes and polling firms’ public reactions across time and space.