Analyses of local campaign effects are dominated by aggregate-level analyses of constituency activity. Though individual-level data are available on whether voters are (or remember being) contacted by parties during campaigns, their analysis is fraught with difficulties, not least the extent to which memory of campaign contact is itself conditioned partly on party allegiance, creating a circularity in the analysis of the impact of party contact on vote choice. To some degree, this can be (and has been) dealt with in a regression framework. However, this does not fully deal with the potential difficulties. Ideally, experimental approaches are needed to tease out definitively the effects of campaign exposure on individual’s election decisions. However, these present practical difficulties. In this paper, therefore, we utilise quasi-experimental difference-in-difference and propensity score matching methods to estimate campaign effects at the 2010 British General Election from individual-level data.
- constituency campaign effects
- quasi-experimental methods