Does music influence Citizens’ Evaluations of a Candidate?

on mai 17, 2017 | 0 Comment


Emmanuel Heisbourg
Département de Science Politique, Université de Montreal
Superviseur: André Blais




I will conduct a study investigating whether and to what extent music impact citizen’s evaluations of a candidate. The debate concerning the political influence of music is not a new one. Devlin (1986, p. 29-30) argues, for example, that music “is often an integral part of the mood-setting devices of political ads”. Likewise, Thorson et al. (1991) show that music has an influence on how individuals react to political ads. It remains unknown, however, whether and to what extent music influences the evaluation of a candidate sponsoring an ad. This is an important research question, since individuals’ evaluation of a candidate influences their likelihood of supporting this candidate in a given election (Hayes, 2005).

For measuring the impact of music on evaluations of a candidate, I will conduct an experiment in which I expose participants to a real campaign ad from a local (unknown) candidate. This candidate is in favor of more public spending in health and education. The experiment will be composed of two treatment and a control group: In the first treatment group, the candidate’s proposals are presented in the form of a campaign ad containing voice, images, and a melancholic song. In the second treatment group, his proposals are presented without this song. Finally, in the control group, the message is presented without this song and the images.

Following their exposure to one of these ads, participants will be invited to answer a survey containing a battery of questions on how they evaluate the candidate. This includes whether the candidate is compassionate, honest, etc. (Hayes, 2005). I will also ask participants how convincing the campaign ad is (for checking the reliability of the treatment), their overall like or dislike of the candidate, and whether they would vote for the candidate in an election, as well as their partisanship, political ideology, and gender. These last measures will be used as control variables in the analysis. I will conduct this experiment with 192 students from McGill (64 in each group) at the Center’s Media Lab. As such, I will be capable of capturing a mean difference of 0.5 in the main dependent variables (regarding candidate evaluations). 

I hypothesize that:

H1: Participants in the voice-images-music condition will evaluate the candidate as being more compassionate and honest (for instance), than participants in the other treatment (voice-images) and in the control group.

H2: Participants in the voice-images condition will evaluate the candidate as being more compassionate and honest (for instance), than participants in the control group, but less than participants in the other treatment (voice-images-music) group.