Did Gore's Kiss Work?
Campaign Effects in Election 2000

D. Sunshine Hillygus
Department of Political Science
Stanford University
December 2001

My research examines how campaign events affect voters during presidential elections. Previous campaign research has been conceptually constrained by insufficient data that has, in turn, impaired our understanding of the relationship between campaign activities and voter preferences. Using extensive panel data of the American electorate during the 2000 Presidential Election, I am able to explore the complex ways that key campaign events lead to changes in people's vote preferences. I find that there is not one campaign effect as suggested by previous research; rather, the effect of a particular campaign event depends on an individual's previous vote intention and demographics, the performance of the candidates, and the timing of the event. In the 2000 election, I find that Gore converted more supporters following the conventions (and the famous smooch!), while Bush solidified more voters following the debates. This research is important because it addresses previously unexplored nuances about the role of campaigns in American elections, and will help candidates, media professionals and citizens to better understand the factors that explain vote intent.

Political science research has long concluded that American voters maintain stable preferences throughout presidential electoral campaigns. Decades of research have argued that voters select a candidate before the campaign has even begun, so the campaign can have only "minimal effects" on the attitudes and behaviors of voters. Non-campaign characteristics such as partisan affiliations, demographic attributes, and assessments of the incumbent administration or political party are thought to be the primary determinants of vote intention. Yet recent elections have witnessed declining levels of party identification in the electorate and simultaneous increasing levels of campaign efforts by the candidates and parties, reopening the debate surrounding campaign effects. Unfortunately, recent research has stalled on the vague and overly simplistic question of "do campaigns matter?". I argue that limits in methodologies and data have prevented campaign research from advancing our understanding of when and how campaigns influence voter preferences. The way campaign data have been collected and analyzed has largely shaped the way voting behavior has been conceived. For instance, the "campaign" is often treated as a singular, time invariant event (at most, varying by state), a conceptualization that ignores the dynamics of political campaigns and masks the possibility that different campaign events may have varying effects and on different portions of the population. Furthermore, campaign research is often limited to cross-sectional data or pre- and post- election panels that cannot document individual-level movements in vote preferences. Thus, cumulative and aggregated data are deficient.

With the 2000 election, detailed panel data of the American electorate were collected, finally allowing a critical analysis of vote choice dynamics. The Knowledge Networks election dataset includes 29,000 respondents, 102,000 responses to the vote intention question, and a rich set of relevant demographic, political, and attitudinal variables. These data indicate that not only is vote intention much more variable than previously believed, but that inconsistency is not limited to just the less partisan or less politically interested. It also appears that this vote choice inconsistency may be attributable at least in part to presidential campaign efforts.

Using a transition model to compare individual candidate preferences before and after major campaign events (party conventions and presidential debates), I find that the effects of various respondent characteristics and opinions differ depending on the respondent's previous vote preference. For instance, Gore supporters reacted differently to the debates than did Bush supporters and Independents. Moreover, it appears that the effects of campaign events may fluctuate based on the performance of the candidates. Offering empirical support to the conclusions of pundits, consultants, and polls that Gore's affectionate kiss with his wife before his convention speech strengthened his support and his stiff performance during the debates weakened his support, I find that Gore was able to increase support among Democrats, Independents, and women following the conventions, whereas Bush was able to activate Republicans and solidify support among politically interested Independents following the presidential debates.

These findings suggest that current campaign theory is woefully underdeveloped. To understand the impact of campaign events on electoral outcomes, political science research must move beyond the overly simplistic debate as to whether campaigns have an effect or not--campaign events affect vote intention in complex ways.