Volatile Voters, Volatile Estimates?

Vote intentions have long been considered as stable over the course of electoral campaigns. Significant aggregate change in recent campaigns and the diffusion of panel studies have on the contrary insisted on the propensity of voters to change their minds in the short term. It is not clear however which voters are more prone to change and why. Studies also disagree on whether propensity to change at the individual level actually reflect aggregate dynamics. These issues are key in the understanding of current developments in electoral behavior. They are also key in determining the extent to which our instruments -surveys- are sufficient to capture and explain these news dynamics of change. We propose a new look in this issue of volatility from the case of the French 2017 presidential elections, characterized by precisely high volatility. We especially want to assess the weight of various sources of changes in vote choice: the very structure of preferences (change is more likely when voters do like more than one party), level of information and interest in politics as well as the relation of voters to the survey instrument. This study is based on data from the Elipss panel run in France from 2015 to 2017, gathering 13 waves of interrogations about voting behavior of the same persons. This panel is based on a probability sample of size 1000. The dataset carries most of the traditional variables of vote choice studies as well as paradata about how people do participate to such an experience. The paper can thus contribute to answer the first three questions raised by the section about the capacity to predict behavior with surveys, about the quality of online polls and new measurements of political attitudes.