Lies, Damned Lies, and (Twitter) Statistics
The Washington Post’s Wonkblog posted a story yesterday about Twitter’s presence in the halls of Congress. After scraping twitter profiles of representatives and senators and adjusting for multiple accounts, the blog’s author found fairly substantial differences between Republicans’ and Democrats’ reach on Twitter.
For instance, House Republicans had 13 percent more Twitter followers than Democrats. In the Senate, Republicans’ advantage is nearly 18 percent. Taking both houses together, Republicans have a 1.7-million-follower advantage. “In an era when razor-thin electoral victory margins are becoming increasingly common, this deficit matters,” the post concludes.
There’s one problem with this analysis: many of these “followers” are fake. A 2012 study found that nearly 40 percent of all congressional twitter account followers were either from inactive or automated accounts. Wonkblog rightly cites this study.
The real trouble is, fake accounts are not spread evenly between leaders and back-benchers. The study by Advocacy Media found that about 68 percent of Sen. John McCain’s 1.7 million Twitter followers in 2012 were fakes. The same was true for Rep. Nancy Pelosi’s 220,000 followers (she’s up to about 475,000 followers in 2014).
The former GOP presidential nominee is the real elephant in the room for Wonkblog’s analysis. McCain’s current 1.9 million followers are nearly three times the number of followers that House Speaker John Boehner has and more than those of Sens. Marco Rubio, Rand Paul, and Ted Cruz combined. McCain’s followers alone account for more than a quarter of all Congressional GOP Twitter followers. If two out of three of his account’s followers are bots and zombies, the partisan follower gap closes to a mere 42,000.
Obviously, fake accounts for prominent political figures are nonpartisan. But if seven out of ten of the most-followed congressional twitter accounts are held by Republicans, making the assertion of a yawning partisan Twitter gap based on unverified data is problematic to say the least.
That isn’t to say that Wonkblog’s post doesn’t have interesting tidbits. For instance, being in the majority apparently encourages tweeting.