How did the president’s deep-red critic, Sen. Jeff Flake (R-Ariz.), become the surprise roadblock to Brett M. Kavanaugh’s confirmation vote? That’s what happened last week when he voted to move Kavanaugh’s nomination out of the Senate Judiciary Committee — but only on the condition that the FBI be asked to investigate Christine Blasey Ford’s allegations that Kavanaugh sexually assaulted her when they were teens.
But votes aren’t the only way to measure politicians’ views. At the NYU Social Media and Political Participation (SMaPP) lab, we’ve developed an alternative ideological measure based on members of Congress’s social media feeds. For most members of Congress, our social media ideological ranking looks very similar to ideological measures produced from voting behavior. But that’s not true for Flake: His social media ideological rating is surprisingly close to that of his moderate Republican colleagues.
In other words, Flake votes like a staunch conservative, but he tweets like a moderate.
How we did our research
To draw this conclusion, we estimated the ideology of members of Congress using the news media that members share on their Twitter feeds. We extracted the URLs of all national online media sources (e.g., nytimes.com, washingtonpost.com, foxnews.com) from the tweets that each member of Congress has shared since 2015 (115,828 in total). We exclude URLs from “quote tweets,” which Twitter users often use when rebutting or criticizing an article. We then use a statistical model to help us infer the ideology of each member based on the news that they share.
The logic behind this approach is simple and intuitive. When users share stories from media outlets considered to the left, we rank those as liberal; when they share news stories from the right, we rank those as conservative. For instance, if you were to share stories from FoxNews.com or Breitbart.com, our rating system would conclude that you are likely to be more conservative than someone who shares stories from MSNBC.com or DemocracyNow.org.
Based on those tweets, our model then ranks each member of Congress along a left-to-right scale, assigning each legislator an ideological score. (Pew Research has recently done something similar with Facebook data.) The model also simultaneously produces an ideological score for each publication. Although we don’t discuss that in this post, the plausibility of the media scores generated by the model helps give us confidence that the model is working as intended.
Where Flake stands
Republican Sens. Susan Collins (R-Maine) and Lisa Murkowski (R-Alaska), well known as moderates, have signaled that they weren’t certain how they would vote on Kavanaugh’s nomination. But it was Flake who called for an FBI investigation before he would vote the nomination out of the Judiciary Committee and onto the Senate floor. Flake’s actions in the 11th hour and his skepticism about Kavanaugh’s nomination have led some to suggest that he is an ideological moderate. For example, on NPR’s “Morning Edition,” journalist Mara Liasson grouped Flake with Collins and Murkowski, calling them “three moderate Republicans.” Others have also referred to Flake as “a key moderate Republican.”
That’s not how Flake sees himself. As he said recently of Kavanaugh: “I’m a conservative, he’s a conservative judge. But I want a process we can be proud of.”
It’s certainly true that Flake’s voting record in the Senate is unwaveringly conservative. According to the well-known measure of political ideology based on senators’ roll call votes used by political scientists, only two senators — Rand Paul (R-Ky.) and Mike Lee (R-Utah) — are more reliably conservative in how they vote. It’s true that Flake has followed the lead of his friend and fellow Arizonan, the late senator John McCain, who was initially one of the Republican Party’s more conservative senators but who gained a reputation as a “maverick.” The label was due in part to his old-style bipartisan legislative approach and by speaking up when he thought his party was in the wrong — including casting a memorable vote last year against the bill to partially repeal the Affordable Care Act. But one can hold strongly ideological views and still be willing to buck your party or work across the aisle.
Beyond how they vote and how they talk: What do they share?
As we described earlier, we estimated the ideology of members of Congress by assessing what news sources they most regularly share on Twitter. For the most part, members’ voting ideology and Twitter feed ranking were extremely simple. But we found one important exception: Flake.
The figure below shows our scores for the senators whose voting record is the most conservative (excluding Flake); for the average Senate Republicans; and then separately for Flake, Collins and Murkowski.
If Flake shared news that was as conservative as his voting record, we’d see a score close to that of the 10 most conservative senators. Instead, he posted stories from news sources that place him to the left of the average Republican senator — closer to that of moderates Murkowski and Collins, who tweet like they vote. Flake, curiously, has tweeted links to CNN about three times more often than links to Fox News. This ratio is virtually reversed among other Republicans. In fact, where his ideology inferred from his voting record places him as the third most conservative, his ideology as inferred from his news sharing places him as the fourth most liberal Republican.
Richard M. Nixon’s attorney general, John Mitchell, said “watch what we do, not what we say.” If we judge by what Flake has done on the Senate floor, Flake will probably vote to confirm Kavanaugh. If he doesn’t, then one of the clues that this might have been coming could have been hiding in plain sight not just for the past few weeks, but also for the past few years.
Gregory Eady (@GregoryEady) is a postdoctoral fellow in the New York University department of politics at the Social Media and Political Participation Lab.
Jonathan Nagler is a professor in the politics department and a co-director of the Social Media and Political Participation Lab at NYU.