Do Campaign Events Matter? New Evidence from Voting Advice Applications

Abstract

THE EFFECTS ON ELECTORAL OUTCOMES attributable to specific events that take place during an election campaign have been difficult to observe using conventional survey methods. Campaign events such as television interviews, poll releases, and newspaper endorsements have generally been seen as negligible when compared with ostensible determinants of vote choice such as partisanship.1 However, there is limited empirical evidence to support or refute the effects of campaign events, given that efforts to infer their size and direction are thwarted by sample limitations that inhibit the measurement of public opinion directly prior to and in the immediate aftermath of a particular event. A unique opportunity to examine the effect that campaign events have on elections can be found in the sample collected by Voting Advice Applications (VAAs), which are novel sources of public opinion data in terms of both their extraordinarily large sample sizes and their continuous accrual of observations over the duration of a campaign. As the name implies, VAAs are interactive, online instruments that survey users as to their political preferences and, based on the responses provided, estimate the user’s alignment with the various political parties or candidates running for office.2 A number of VAAs across the world have proven to be wildly popular, drawing millions of users and thus offering the closest thing that political scientists presently have to a source of “big (survey) data” on the dynamics of public opinion during elections. Because they continuously sample public opinion throughout an election campaign, VAAs allow for time-series cross-sectional estimates that illuminate the dynamics of public opinion in ways that are not possible using conventional polling data.