A visual overview of 2018’s election Twitter

Summary

Election activity on Twitter in the final days of the 2018 midterms comes with bewildering speed. However, a look at the networks formed by the highest-frequency posters in all competitive races show that Democrats and Republicans have different patterns of interaction.

Methodology

The graph below is a slightly stylized look at the most active posters over the last week in the Pollchatter.org database. To create the nodes of the graphs, we gathered the top 100 most active posters (original tweets or retweets). Then, for each account, we added the top 10 accounts they retweeted most often in that period, and then the top 10 accounts that each of those accounts retweeted most frequently.

Each individual account is a dot (node) in the network diagram. Nodes that have more connections (degrees) – thus, have been retweeted more often – are relatively larger. Connections (edges) between the nodes indicate a frequent-retweet relationship. However, these are not weighted edges, so the connections say nothing about the relative frequency of such retweets.

This is very simple network graph, able only to show clusters of association among highly active users and the accounts they are most interested in. With this caveat, several observations can be made.

Network graph for high-frequency election-Twitter posters, 10/28 - 11/3

Network graph for high-frequency election-Twitter posters, 10/28 – 11/3 (click for larger version)

First, there is an obvious, predictable Democrat/Republican split, as might be predicted. The left side of the network is all Democrat clusters, while the right is Republican.

Both parties exhibit a densely interlinked center. However, the character of this center differs between the two.

The GOP side

The GOP cluster is dominated by a few big-name celebrity accounts – Donald Trump and Donald Trump Jr. being the most obvious ones – along with key Senate candidates such as Ted Cruz (Texas), John James (Michigan), Mike Braun (Indiana), and Patrick Morrisey (West Virginia). The networks around each of these candidates are fairly closely interlinked, meaning that numerous accounts are tweeting and retweeting within multiple districts.

The lavender component centers on New Hampshire politics, and particularly GOP candidate Eddie Edwards in NH-01, while the slightly removed green component is driven by the AZ-01 congressional race, with looser connections to the main cluster.

The yellow GOP cluster is an interesting anomaly. This group seems to be highly active, very conservative accounts with only loose connections to the GOP mainstream, but strong connections to each other. The character of their posts and account names raises some suspicion that they might fall into the category of propaganda troll, but more specific research is needed to say anything more definite.

The blue group that connects the two sides is somewhat heterogeneous, but prominent here are obvious media, election-analyst and polling-service accounts. This is probably a good sign, in the sense that the left and right are sharing some (if minimal) sources of information about the political environment.

The Democrat side

The Democrat side too has a few obvious candidate hubs – particularly Abigail Spanberger in VA-07, Carolyn Bourdeaux in GA-07, and Joe Cunningham in SC-01. But it is more clearly held together by closely linked networks of activist accounts, such as the dense network of large nodes in the lower-left quadrant (StormResist, CaptainsLogAz, Prometheus2018, etc).

This group’s less-connected components are also built around specific districts or candidates, including (moving counter-clockwise from the top) Jessica Morse in CA-04 (tan), Nate McMurray in NY-27 (gray), Antonio Delgado in NY-19 (red), Bob Menendez in the New Jersey Senate race (light green), Tedra Cobb in NY21 (green), Lauren Underwood in IL-14, and Scott Wallace in PA-01 (although here, the candidate is less well-connected than the activist network around him).

What does this tell us?

The Republicans certainly seem to be more densely connected around a handful of their Senate races. Trump’s presence helps tie this together, but even in his absence, high-activity activist accounts are focusing on a few of the big Senate campaigns.

Similarly, it seems that there is a) considerably more activity within the Democratic House campaigns than on the GOP side, and that b) the tightly linked group of high-frequency activist accounts plays a bigger role than “celebrities” in holding the Democrat cluster together.