Abstract. Opinion polls are a well-established part of political news coverage, especially during election campaigns. At the same time, there has been controversial debate over the possible influences of such polls on voters’ electoral choices. The most prominent influence discussed is the bandwagon effect: It states that voters tend to support the expected winner of an upcoming election, and use polls to determine who the likely winner will be. This study investigated the mechanisms underlying the effect. In addition, we inquired into the role of past electoral performances of a candidate and analyzed how these (as well as polls) are used as heuristic cues for the assessment of a candidate’s personal characteristics. Using an experimental design, we found that both polls and past election results influence participants’ expectations regarding which candidate will succeed. Moreover, higher competence was attributed to a candidate, if recipients believe that the majority of voters favor that candidate. Through this attribution of competence, both information about prior elections and current polls shaped voters’ electoral preferences.
Deep learning-based election results prediction using Twitter activity
