As a use to journalists, Facebook sent around information on how a 2016 possibilities have finished during “engagement” over a final month — how many millions of people have finished an rivet to a politician’s Facebook, and how many engages those people finished did.
Here are a numbers.
“Oh ho!” we say. “Look during all that engagement!” And maybe: “Clinton is series one, and she is winning and therefore there is a association here! Also, Trump!”
Or maybe you’re like, “OK, what does this tell us?”
That is what we am like (as a kids say).
We can review a information from Facebook to how a possibilities have fared in polling over a same time period, that will maybe give us a improved clarity of a application of these numbers. We can tract how most a check numbers have altered contra how most improved or worse any candidates’ engages are than everybody else’s.
Giving us this.
There appears to be a attribute between a series of people articulate about a claimant over a final month and how that claimant has fared! Maybe, then, articulate about a claimant on Facebook is related to polling increase! Now, this creates sense, generally, as people get vehement about possibilities and wish to worry their friends and family with their newfound crushes. (“Sam Johnson is in a attribute with Ben Carson.”)
If we demeanour during a association between a candidates’s Facebook rendezvous and their polling change, it’s indeed sincerely strong. There’s a 0.61 r-squared between a series of interactions and a change in a polls, suggesting a couple between a two. (Editor’s note: Philip is a nerd.) Maybe Facebook can be predictive!
But something else happened over a final month: Joe Biden announced that he wouldn’t run for president. That meant that his decent-sized bottom of support went elsewhere, mostly to Hillary Clinton and Bernie Sanders who, as a usually dual big-name Democrats in a race, also got a lot of interactions.
If we take a Democrats out of a mix, a association between interactions and check numbers collapses. That decent 0.61 r-squared is now a 0.22 — and a closer to 1 we get, a stronger a correlation. You can see a problem above. Jeb Bush and Carly Fiorina saw large polling drops though didn’t have terrible communication numbers. So maybe Facebook isn’t predictive after all.
We can even loop in another metric: Interactions per person. Maybe if people are prone to correlate over and over, a claimant has some-more support.
Nope. Pre- and post-excluding Democrats, a numbers here are about a same as for interactions on a whole. Ted Cruz and Ben Carson have some-more intent fans, though that’s all we learn.
So because does Facebook send these numbers around, if they don’t unequivocally tell us anything about a candidates? Because Facebook is in a business of removing us to use Facebook, not in a business of presaging domestic campaigns. They only wish media outlets to cover a Facebook numbers, essay about how …