Today brings an updated paper on the impact of Fox News on American elections. Here's the basic chart for presidential elections:
The methodology is the same one the authors have used before, which relies on the channel position of FNC in local markets. It turns out that the lower the channel position the more a channel is watched, but the channel position is basically random. This provides a way to answer the question: "If someone is randomly exposed to more Fox News, what happens?"
The answers are (a) they become more conservative, (b) they become more Republican and (c) they vote more for Republicans. The results are substantial:
A one-standard-deviation decrease in FNC’s channel position increases the share of Republican voters by about .75 percentage points....The effect sizes are substantively quite large. Indeed, the effect of just a one standard deviation shift in FNC’s channel position (roughly 29 positions), which induces about 7 minutes of additional viewership per week, is larger than the effect of a one standard deviation change in local real wages.
If I'm reading this right, it means that seven minutes of Fox News per week is likely to increase the Republican share of the vote by 0.75 percentage points. The authors find similar effects for House, Senate, and Governor's races.
As you can see from the chart, the effect of Fox News increased steadily early on, as Fox became more conservative at the same time that it broadened its reach on cable systems. It appears to have stabilized more recently, which is unsurprising since Fox is on pretty much every cable system now and can hardly crank up the conservo-meter any further.
This is the kind of research I'd like to see for social media. It wouldn't be easy since you need to find some kind of endogenous instrument (like channel position) to base your comparison on, and it's not clear what that might be. And the results are likely to be fuzzy since social media, unlike Fox, is not 100% ideological in a specific direction. Still, someone clever needs to take a crack at it. Based on the evidence we have right now, I suspect that the impact of social media on elections is fairly small, but it would be nice to see something more rigorous on the subject.