Newspaper coverage of female candidates during election cam paigns: evidence from a structural topic model

This paper studies how female candidates are covered by newspapers during election campaigns. Previous studies have generally found that the media portray men and women quite similarly, but they have tended to use relatively selective sources and to focus on the amount of coverage and its tone, but not on on its content. We aim to gain a more nuanced understanding of this phenomenon by focusing on the Swiss national elections 2015, relying on an almost comprehensive sample of news items covering most of the duration of the campaign. Our dataset includes about 205,000 documents from 70 sources, covering over 3,900 male and female politicians. We analyze these texts first with a focus on the amount of coverage and second using structural topic models, a natural language processing technique that allows us to identify inductively the themes (topics) of newspaper coverage and, importantly, how the candidates’ gender is linked to both the topics and the language used to discuss them. Results reveal a gender bias in media attention only for non-incumbent candidates. Moreover, some topics, such as the one referring to the election campaign, is discussed more frequently in connection with female candidates, while others, such as topics focusing on personality and political profiles, are more prevalent in articles mentioning men. We uncover no major gender bias in the language used for these topics