“The Brexit Effect” – political incorrectness, polling, and a Trump win.

Some time before the end of the primary season, I was convinced that Donald Trump would be the next President, and indeed have now won a few small bets on the matter. [1] His positions, while completely untenable and offensive in the eyes of “the establishment”, seemed tailor-made to appeal to those who are fed up with that establishment, and given his expertise in reality television and mass publicity – key to charisma in the modern day – my sense was that Trump would be able to rally a good chunk this group and lead them to victory. I further thought this was especially likely against Clinton, who in many respects embodies “the establishment”.

I’m not sure if that story ended up being true or not, so insofar as I was right it may have been for the wrong reasons. Hillary did win the popular vote, after all – but it’s clear that Trump was indeed a better position than the polls and pundits would have it.

I submit, however, that the surprise result we encountered should perhaps be considered another instance of what I’ve heard being described as the “Brexit effect.” This phenomenon has gone by other names throughout history – the Bradley effect, the Wilder effect, Shy Tory Syndrome – and I’m sure there are other phrases that have been used for it was well.

In brief, this effect is one in which social desirability pressure causes people to be more reserved about their support for controversial political causes, especially those which are considered politically incorrect. As a result, polls understate support for such causes, and in-person polls especially so – but the privacy of the voting booth tells another story.

This effect has now happened – or is argued to have happened – many times in recent history. I am not an expert statistician and do not have the technical chops to determine whether or not similar effects were in play in all such cases. But what I do know is that I personally know several people who have at least some private sociopolitical opinions that they don’t share in public, and in almost all cases those opinions are those that are conservative or “politically incorrect.” [2]

I wasn’t hearing many people talking about this consideration prior to the US elections, except in the context of “Hey Hillary fans, people thought Remain would win too, be sure to go out there and vote!” Some did bring the effect up – Thomas Edsall discussed this in a column in May, and the (in)famous Scott Adams maintained that Trump supporters were being bullied into silence for some time – but insofar as these concerns did come up they were often dismissed. [3]

I suspect it’s time to take them more seriously.

 

 

[1] Before you give me points for accurate forecasting, I lost confidence in my original prediction following the Access Hollywood reveal, which seemed too big a scandal for Trump to live down, and was assuming afterwards that Hillary would end up taking it – so I can’t really claim credit for being “right all along”.

[2] This isn’t to say there can’t be secret liberal dissenters too – but the general narrative in America and much of the West these days has been a liberal/progressive one, so one would expect more secret conservative dissenters than the reverse.

[3] Indeed, many commentators immediately pre-Election Day were attacking Nate Silver and FiveThirtyEight for “skewing the data” in claiming that Trump had even a 30% chance to win. It’s possible, of course, that these commentators were right and we’re seeing a lucky hit for Trump – but at least to me this win isn’t what a lucky 30% chance feels like.

Advertisements

2 thoughts on ““The Brexit Effect” – political incorrectness, polling, and a Trump win.

    1. He had a big buffer in the battleground states. He went well beyond the minimum electoral vote.

      The basic thing is, Silver had meta-uncertainty across multiple models. Most other people used a model and gave its chances as the actual chances. This is not good Bayesian reasoning, which prescribes a distribution across models and greater uncertainty. Maybe in a normal election you could pull back on the meta-uncertainty. In this election, as bizarre as it was, the most honest thing to do was to admit you only had a clue. 60% was a good estimate. 85-98% like other people said? Those were not good estimates.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s