Well, the results from the big debate are in, and the loser is clear: Twitter. Nate Cohn at the New York Times called it, noting “the ‘winners’ and ‘losers’ of the debates can end up looking a lot different than what you see on Twitter.” But the real hammer was dropped by our own Monkey Cage Twitter(!) feed, which declared:
Has to be said: we really don't know whether Google searches or tweets during/after a debate are a measure of anything meaningful.
— The Monkey Cage (@monkeycageblog) August 7, 2015
As the parent of a soon to be teenager, I know that nothing makes things more alluring that being told you shouldn’t look at it, so with that in mind, here’s the taboo Twitter data (which we are collecting at the NYU Social Media and Political Participation (SMaPP) lab) that you absolutely should not be paying any attention to today (click on figure for bigger version):
And for good measure, here are three questions that you should not be asking yourself about this data.
First, is there a logical fallacy in saying “the party elites will decide the nominee” — the argument invoked by Cohn in the piece quoted above on the basis of excellent political science research — then not wondering if party elites are interested in which candidates are exciting voters, and, if so, how those party elites come to the decision of which candidates are exciting voters? The same question holds for the argument that the media reaction to the debate may be more important than the debate itself. How exactly does the media determine who won the debate? Can the media ignore real time data on how hundreds of thousands of Americans are reacting to the debate? Especially if we at the Monkey Cage are happy to give the rest of the media that data?
Second, is there any value in noting that right around 10 p.m., Marco Rubio seemed to jump out of the pack and dent the attention being paid to Trump, whose Twitter mentions then began to decline for the remainder of the debate? I mean, it is possible that these are bots (accounts run by computer programs), but that wouldn’t explain why Rubio’s bots suddenly swung into actions and Trump’s grew quieter. (As an aside, I’m betting this is the only time you will read “Trump grew quieter” in any post-debate analysis.) It’s also possible that all of these comments are from Democrats mocking the candidates; we just don’t know from looking at counts like the ones in the figure. But, given time, we can figure this out (and we’ll be working on this throughout the campaign at the SMaPP lab) using a wide range of tools, including examining random samples of the Tweets by hand (which could quickly put to rest whether the response was overwhelmingly negative or not). I wonder if the Rubio campaign is going to be doing that today? I wonder if what they find might be interesting to “party elites”?
Third, if Bush and Trump are the leaders of the pack, does anyone care if Trump continuously, dare I say, trumps Bush in Twitter attention? (Sorry – couldn’t resist…) Does it matter that it went on during the debate last night? Does it matter, as the figure below shows, that it’s been going on the past couple of months? Does it matter that the gap seems to be declining, but that this is primarily because Trump is getting less attention not that Bush is getting more attention?
As a final note, it is worth pausing once again to note what this Twitter data is. It is not a bunch of carefully screened people sitting in the extremely artificial environment of a television studio discussing what they thought of the debate. It is not a bunch of people being given dials by a television studio and asked to record their impressions as positive or negative as the debate progresses. It is people in the course of their own lives, without prompting from scholars, journalists or public opinion firms; without agreeing to participate in a study or focus group; responding to the debate in real time sharing their opinions about the debate, the candidates, the election, the issues, etc. Even if the predictive value for the ultimate winner of the primary of these data are low (and even this is a subject for future research, as these data have not been around as long as polling data or even prediction markets), maybe it is still worth a quick peek this morning.
[Huge h/t to Alexandra Siegel, an New York University candidate in political science and a graduate research associate of the SMaPP lab, who stayed up well after the debate ended preparing these figures.]