Changes in U.S. Gubernatorial Social Media Campaigning from 2014 to 2018

The next speaker in this AoIR 2019 session is the fabulous Jenny Stromer-Galley, who shifts our focus to 2014 and 2018 gubernatorial campaigns in the United States. She begins by noting the significant growth in negative advertising in U.S. elections, and this increase may also have led to a gradual decline in voter turnout as well as a general mistrust of political and democratic institutions.

Research into the uses of social media in political campaigning should aim to generate similar longitudinal datasets, to compare campaigning strategies over multiple cycles. This would also enable us to identify the rhythms of individual campaigns, and the way that such rhythms are evolving from campaign to campaign. Such rhythms are also influenced by the broader electoral context, such as which parties are in power and whether there are predictions of a change in government, as well as the gender and other identity aspects of the candidates. These might also influence campaigns’ likelihood to run attack ads.

And of course much of the research remains focussed on platforms like Twitter, where large-scale data access remains somewhat more possible than on Facebook and elsewhere.

The present study is interested in changes between the 2014 and 2018 gubernatorial campaigns (and in between, Donald Trump was elected as U.S. President, and this also affected campaigning styles and the prevalence of negativity in campaigning); it works with data from Facebook and Twitter for the last eight weeks of the 2014 and 2018 campaigns, yet because of the changes in Facebook API access there are also some gaps in the 2018 datasets.

The content was then classified by human coders and supervised machine learning algorithms for the presence of attacking content, which provides solidly but not perfectly accurate classification of the content; it also attempted to test for the presence of incivility but this has proven too unreliable for use at this stage.

Candidates seemed to attack less on Facebook than Twitter in 2014, but attacks on Facebook increased in prevalence in 2018. Female candidates were much less likely to attack than male candidates, and Republicans also seemed somewhat more likely to run attacks. In competitive races, attacks are more likely, and incumbents are less likely to attack. There are few significant changes from 2014 to 2018 – at present, there is little evidence that campaigns have become more negative from 2014 to 2018.