The Impact of Survey Mode on the Measurement of Polarization in the Public: 2016–2020 U.S. Panel Evidence

Over the last 40 years, affective polarization has noticeably increased, with recent elections showing especially large gains. Moreover, these increases have surpassed other indicators of polarization, such as issue attitudes and partisan-ideological sorting. Could the recent surge in affective polarization be influenced by the growing use of internet-based, self-administered surveys? I argue that measures of partisan affect are more vulnerable to mode effects than other indicators of polarization. In this paper, I employ American National Election Studies panel data from 2016 to 2020. Crucially, in 2016 the survey was conducted with a dual mode in which a portion of respondents were interviewed in the traditional face-to-face format while others completed a self-administered online survey. Both sets of respondents were re-interviewed in 2020 where all respondents completed the survey online. I find that differences in party affective polarization and candidate trait ratings between the two modes, substantial in 2016, largely disappear when social desirability effects are removed. In contrast, issue extremity and partisan-ideological sorting show no noticeable mode effects.