Inter-racial internet dating behavior predicted a “November surprise” — Trump’s victory
Why were the Election 2016 polls wrong?
It wasn’t because of flawed models or flawed statistical analysis or Bernoulli vs Binomial (here and here) or simulations or Markov Chain Monte Carlo blah blah. It wasn’t even because of bias in the polling data.
The polls were wrong because people lie on polls. Unless they are stupid or have no filters, they know what the “right” answer is. In the case of the wrong predictions about Election 2016, specifically, it is because white people are more racist, men are more misogynist and citizens are more anti-immigrant than they are willing to admit. Not to say that there aren’t non-white racists (plenty of those here in Silicon Valley), anti-immigrant second-generation citizens or anti-feminist women. As an example — just look at the number of female late-convertors to feminism writing weird confessionals on Medium. There are also plenty of white women for whom their own racism trumps any desire for feminism or even solidarity with other women. Not to mention the“LeanIn-ists” who ignore issues of feminism and class.
A couple of weeks ago a Lyft driver (a 60-ish white male) said there might be a November surprise — that Trump had a lot more support among whites (women and men) than the polls indicated. Politically, many “middle-of-the-road” whites — those who are not explicitly feminist (“Of course I believe in ‘equality’, why does anything have to be done about it?” or “Why do feminists assert that women are superior?”) and those who are not explicitly anti-racist (“Why can’t we all just get along?” or “AllLivesMatter”) — are actually quite racist and misogynist. But they are smart enough to know that it wouldn’t do for them to admit it in public. (Or they don’t want to admit it to themselves — many candid remarks shared with me as some kind of “honorary white” have been prefaced with “Don’t get me wrong, but …”.)
In a poll (even an anonymous one) they will not voice support for Trump, but in their FB or other social media bubbles, in the privacy of their living rooms or at a tailgate party at a football game or when they are yelling at their TVs Trumpism (misogyny, racism, just general hate for others …) resonates with them. In the privacy of the voting booth, Trump will win, he has already won.
Thinking about it later, this behavior is consistent with behavior on dating sites (there is a series of three OKCupid blogs about it — just google “okcupid blog race”). There is a huge discrepancy between the proportion of people (mostly whites) who claim to be “open to dating other races” and the proportion who actually do. They seem to make that claim in order to appear cool to others of their own race (sort of a white mating call, if you will, an actual scene from Portlandia). Actual cross-racial dating rates have decreased since internet dating started.
Summary of OKC’s findings on Dating, Race and Wypipl lying:
The first blog simply looked at “openness to dating people from other races” by race and found that, overwhelmingly, wypipl were not open to dating people from other races, much more so than most other demographics. Only about 25% stated they were open to it. So they published this on their blog, it got picked up in some MSM stories about race and dating. Two years later, OKC decided to do a follow-up, and lo and behold, found that a zit-popping 75% or so were now open to dating people from other races! Wow! sudden attitude change, effects of Obama Presidency, actually acknowledging their racism and changing their minds, demographic shift to more libs on the site? Nothing of the sort, they just felt racial shame at being found out for the racists they were and went and changed their answers to the question in the intervening years. (OKC has the data.) Some time later, brilliant DS at OKC decided to check their own behavioral data by looking at likes, messaging and responding to messages by race. The numbers were consistent with the earlier survey results, only 25% or so of wypipl had response or messaging rates that were comparable for people of their own race and other races.
So screw the significance levels and the confidence intervals and the x number of simulations
Wypipl and men simply lied to the pollsters. In the privacy of the voting booth, with no one to see them or judge them, they voted for Trump.