Found: The Stupidest Political Science Study Ever

You know how the left is obsessed with proving that Trump’s election is entirely owing to Russian Kallusion. Now we have empirical social science to prove it! At last!

From the “peer-reviewed” internet journal First Monday (I’ve never heard of it either) comes this new study from several academics at the University of Tennessee: “Internet Research Agency Twitter Activity Predicted 2016 Election Polls.” The Internet Research Agency (IRA) is the Russian bot factory that unleashed lots of Facebook posts and Tweets during the 2016 election campaign. Here’s the central empirical claim of this new study:

We find that the release of these tweets parallel significant political events of 2016 and that approximately every 25,000 additional IRA re-tweets predicted a one percent increase in election opinion polls for one candidate.

So let’s see if I have this straight: Hillary’s campaign spent about $1 billion, but a couple hundred thousand bucks and some re-tweets was all it took to swing the voters massively to Trump? Seriously? Who knew winning elections was so simple? And how did we ever win the Cold War against these guys?

Question: Is this purported relationship linear? If so, then why didn’t the Russians gin up 250 million more retweets, and ensure that Trump got 65 percent of the popular vote?

If you click on the link, you’ll see this “study” is very short. It could be printed in about six pages. The four authors acknowledge in their concluding discussion that “Any correlation established by an observational study could be spurious.” Ya think? There’s also this curious caveat: “Here we have tested prediction, not causality. It seems unlikely that 25,000 re-tweets could influence one percent of the electorate in isolation.” And also this one: “using aggregated data means we cannot know the extent to which the participants in election polls were exposed to IRA disinformation.” In other words: we ain’t got squat.

This study is a statistical exercise in torturing the data till it confesses what you want, and only within some carefully defined parameters that bias the results in a particular direction. But it was enough for NBC News to run with, in a story today headlined “New study shows that Russian propaganda may really have helped Trump,” even though the study—and NBC’s own reporting in the body of the story—doesn’t conclude that at all:

In an interview with NBC News, Ruck said the research suggests that Russian trolls helped shift U.S public opinion in Trump’s favor. As to whether it affected the outcome of the election: “The answer is that we still don’t know, but we can’t rule it out.”

The fact that NBC News would jump on this frail statistical exercise tells you all you need to know. This really does fall under the literal understanding of “fake news.”

UPDATE: Philip Bump in the Washington Post is not impressed either. Also Eric Boehm at Reason.

Responses