Why the UK tells us little about the US election

Brexit was a shock, the Tory election victory was a shock, hence the US election will be a shock. Not so fast, this is based on misinterpretations of both the Brexit and general election results and also misses crucial differences between the situations.

Firstly let’s deal with Brexit. People were angry with loss of national power and immigration and voted to rip things up in the hope that things would get better. This would point towards parallels with Trump’s support, there are some, but it’s not totally the same. Compare the detailed demographics of Trump’s support with polls of who voted for Brexit. Older voters support Trump and Brexit as do white voters. On the other hand Trump’s misogyny has led to a big gender gap, one which didn’t exist in the Brexit vote. There’s also the narrative that Trump gives a voice to the ignored white working class. The evidence for this is patchy, the voters earning the least break heavily for Clinton but that will include a larger number of African-American and Latino/a voters due to racial wage disparity. Trump does best amongst middle class voters, however when the vote of the white cohort is broken down by education level those without a tertiary education skew strongly to Trump. This is very similar to what was seen in the Brexit vote. The Brexit Leave vote however had a significant advantage amongst those disengaged with politics. This narrative is constantly applied to Trump by self-flagginating liberal journalists. However most of his support in the primaries came from the Republican base, not new voters.

There is one important thing to say about the Brexit surprise, it wasn’t a surprise, not for the polls. In the two weeks before the referendum equal numbers of polls had Leave and Remain ahead. Given that so many votes were made by post (partly due to the timing around summer holidays) it’s important to take into account the polling over an extended period leading up to the vote. This presents another interesting, yet overlooked parallel, between Brexit and the US election. While some US states like North Carolina have attempted to curtail early voting others such as Colorado have made it much easier. This year Colorado, the most likely pivot state for a Clinton victory, has given almost all voters a postal vote. By the 1st of November 1 million people in Colorado had voted already with plurality being registered Democrats. That’s 40% of the number of votes cast at the last presidential election. Hence for ~40% of votes cast in Colorado the relevant polling data will be from last week, not the final election poll.

It shouldn’t be a big surprise that referendum with no real polling precedent and with polling showing roughly a 50/50 split ended up with a 52/48 result. What was a surprise was that it was assumed that there would be a late swing to Remain, there probably was but it wasn’t enough and a large enough number of postal votes had been cast that the effect was diluted. Brexit was a coin toss, it landed on tails, Leave.

Next there’s the matter of the polling disaster in the 2015 UK general election. Going into polling day the expected result was a 35%/35% tie between Labour and the Conservatives. It ended up with the Conservatives winning 38%/31%. After this the UK pollsters got together and held an inquiry to find out why they hadn’t got the result right. They dismissed the “late swing”, the “shy Tories” and the idea that better favourability ratings for Cameron over Miliband suggested people were going to vote Tory in the end. The issue was that the samples the pollsters were taking had too many Labour voters. One big reason for this was the way the polls were done, by sample quotas. The pollsters would poll a bunch of people and once they had responses from enough people in a particular demographic group they stopped looking for people from that group in their surveys. This meant that demographic cells in a pollster’s table would fill up with the easiest voters to contact. The away from the main polling companies the academic British Election Survey and the British Social Attitudes Survey both attempt to survey the UK population and their views using repeated calls. This means that they should correctly sample harder to reach voters. These surveys also keep track of how easy each voter was to contact. They found that Labour voters were much more likely to respond to the first or second call than Conservative voters. This meant that quota-based systems were intrinsically biased towards easier to reach Labour voters.

So how does this link in with the American election. Well there is the possibility that the polls are similarly biased. However why would Trump’s white, less likely to have tertiary education base be less likely to respond to polls? Even if they were the US elections provide a much richer data environment than UK elections. That’s because there are party primaries which have both opinion polls and published results. Turns out once you take into account undecided voters in the polls Trump didn’t outperform his polls during the primaries. Now could the general election polls be biased in other ways? Yes, but it would have to be some sort of effect not present in the primary polling (such as suppression of the African-American vote for Clinton).

So what does this mean? Well firstly stop saying the shock of Brexit means Trump will win, because it doesn’t. There are some parallels but there’s no evidence this leads to an anti-Trump polling bias. There’s also no evidence that the bias in UK polls in 2015 will be replicated in the US and it didn’t appear during the primary polling. There is however a parallel in early and postal voting meaning that late swings in the polls are less significant (and this is probably positive for Clinton). If she can win all the states that have voted Democratic in the last 6 elections, plus New Mexico (where she’s 12 points up), plus her running mate’s state of Virginia and two of New Hampshire, Colorado and Nevada (the latter two where she is doing well in early voting numbers) then she wins. Florida and Ohio are big prizes but in the end they are unlikely to be anything but a cherry on top.

Some sort of polling surprise may happen and Trump may win, but it’s unlikely to be due similar effects to Brexit.

Why the UK tells us little about the US election

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 )

Google+ photo

You are commenting using your Google+ 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 )


Connecting to %s