Most studies on political campaigns aim to capture attitude change as a result of the political messages disseminated by the political actors and the media before an election. Selection problems, inherent in media consumption, inhibit a causal interpretation of these findings. Survey, laboratory or even field experiments solve this problem but only at the expense of creating a largely artificial setting, thereby ignoring real-world campaign dynamics. We combine the best aspects of both approaches by using the case of the murder of Pim Fortuyn as a natural experiment, i.e. an observational study characterized by an exogenous process assigning units to different types of treatments in a haphazard fashIon that is as good as random. The founder of the homonymous party (LPF) was murdered in the front yard of his house nine days before the election day and amidst the fieldwork for the preelection wave of the 2002 Dutch Parliamentary Election Study. This event turned the campaign one-sided due to the other parties being silenced by a campaign stop agreed upon. 75 per cent of the 1949 respondents were interviewed before the murder and 25 per cent were interviewed after the murder. Using genetic matching to achieve balance in a long series of pretreatment covariates, we treat the first group as the counterfactual of the second. By using the same indicators as given by the post-election wave of the survey - all respondents had received the treatment by then - we have a placebo test for almost each treatment test. When taken as a whole, the findings offer overwhelming evidence in favor of campaign effects. People do not simply appear more sympathetic to LPF but they are also more likely to regard it as the best party representing their ideas and also to classify it as the most preferred party in the forthcoming coalition. Moreover, they seem more likely to both vote for the LPF and declare themselves as adherents of the party. More importantly, attitudinal change is not only limited to the case of LPF. Rather, the event seems to have had an impact also on the parties of the left, as one would expect if the effects were to be attributed to the unidirectional flow of information that took place after the assassination. By the same token, when exploring the treatment group in more detail, we see that it is mainly those people who are interviewed in the last days of the campaign, only few days before poll day who appear to change their attitudes on the one had towards LPF and on the other hand towards the two parties perceived as the primary representatives of multiculturalism, namely PvdA and the Greens. This finding ensures that the mechanism driving this change is the campaign setting that was fundamentally transformed after the murder of Pim Fortuyn. With this paper, we also provide an important methodological suggestion which refers to the quasi-experimental use of existing observational data in order to identify the causal effects of real-word political events.
