We reported our findings in a CSMaP Data Report produced by the NYU Center for Social Media and Politics. In short, when candidates mentioned a policy topic in the debates, Twitter users generally tweeted about that topic at roughly the same rate. No single candidate brought up a policy issue that suddenly spiked in the Twitter discussion. What’s more, in the swing states, Twitter users generally tweeted about the same policies as did users in the rest of the country.
How we did our research
We collected tweets that contained hashtags related to the debates or the primary season, as well as tweets that mentioned any of the candidates. We used the debate transcripts to note which issues candidates discussed during the debate and how much time each candidate spent discussing each issue. We classified the tweets and speeches into 22 policy topic categories based on the main issue discussed in the tweet or the debate response. We narrowed our analysis to the eight most popular policy topics across the debates: civil rights, the economy, education, the environment, health care, immigration, international affairs, and law and crime. Finally, we used geolocation data provided by Twitter to estimate variation in discussion across states.
Did any of the candidates find an issue that resonated with the public?
Candidates generally try to find issues on which their stances resonate with the public. Did they succeed?
To answer this question, we looked at the issues users tweeted about when discussing the debates in general, as well the topics they mentioned when discussing specific candidates.
First, we compared the number of tweets on each of these policy issues with the time candidates spent discussing them during the debates. Those two measures corresponded remarkably well. However, Twitter users mention topics related to civil rights and immigration more than the candidates talked about them.
Interestingly, while the debate moderators (and candidates) have been criticized for failing to discuss climate and environmental issues in the past, we find that the environment did not get as much attention as other issues among the tweeting public. As can seen in the figure above, Twitter users discussing the debates devoted even less attention to the environment than the candidates did.
To see if topics “overproduced” for a given candidate, we compared tweets about that topic that mentioned the candidate of interest to the amount of time that the candidate spoke about that issue during the debates. For instance, when Andrew Yang spoke about immigration, it generated about twice as many tweets as his other topics. When he talked about the environment it generated barely a third as many tweets as other topics. We found that the response to each candidate was remarkably uniform across topics — none of the major candidates found an issue that especially resonated with the tweeting public. When Twitter users mentioned candidates and issues, they did so in the approximately same proportion as the time the candidates spoke about the issue.
We found remarkably little variation in the discussion of issues across the “battleground” states.
In the six swing states — Arizona, Florida, Michigan, North Carolina, Pennsylvania and Wisconsin — Twitter users responded much as did Twitter users elsewhere. The exceptions were civil rights and immigration, which were in fact discussed a bit differently. For example, tweets concerned with civil rights made up 22 percent of tweets from North Carolina, while only 19 percent of tweets on civil rights were from Florida.
Arizona and Florida had the highest proportions of immigration-focused tweets are Arizona and Florida — which have higher shares of immigrants than the other swing states.
Why is this important?
Our research suggests that the Democratic debates affected the online conversation. At the very least, this implies that debate observers were paying attention to what was being discussed — and, thus, that the debates do give candidates an opportunity to shape public conversation, however briefly. But apparently the tweeting public couldn’t wrest control of the policy conversation away from the politicians and focus it anywhere in particular. This may be because the candidates were already discussing issues that interested Twitter users; our lab has found elsewhere that politicians tend to follow social media discussions of public issues. Or it may be because the debates made for compelling political theater.