Let’s talk about polls

The Brexit is over. Well, the voting is anyway, and not the total mess as a result of it. But I want to make a simple point about the polls leading up to it.

My argument is that all the polls (except for a few) were absolutely spot on. And when people complain about how inaccurate and wrong they were, they are really talking about the inherent problem of a 50/50 model of democracy.

First of all, here is what the polls said between 2010 and until election day. As you can see, throughout this entire period, there were a few ups and downs in the overall trend, but all the polls accurately predicted that it would be a very close race (again, except for a very few outliers).

Source: Financial Times Brexit Poll Tracker

Now, we look at this and then we look at the final result which was 51.9% to “Leave” and only 48.1% to “Stay”, and you might claim that these polls failed to predict that “Leave” would win.


But think about it this way. Imagine we had another election, this time deciding whether ‘pancakes should become the national breakfast’ of the country. And then you did a number of polls where the result was an overwhelming majority (around 75%) to “No”, while only a tiny group of people thought “Yes”.

Then came election time and the final results turned out to be 77% “No” and 23% “Yes”.

If this was the result, would you claim that the polls had been wrong? No, of course not. The polls showed that about 75% +/- would vote no to this idea, and 77% ended up doing just that.

The polls were spot on all the time, just as the Brexit polls were spot on.

The problem we have today is that our political system (in every western country) has turned into this completely polarized system of two equal sized groups voting against each other. And you don’t need polls to measure that.

We know that every single election is going to be 51 percent / 49 percent to any side. And we know that the outcome of each side is going to be so close as to not be statistically predictable.

You cannot do a poll of 2,000 people out of 64 million, or 0.003% of the population, and completely accurately predict whether it will be 51% to one side or 51% to the other side.

For one thing, just the difference between what people say they might do before the election and what they actually end up voting might vary more than that.

In fact, here is a poll done by the same company on the same day (I assume for different media companies). One says ‘Stay’ would win, while the other says ‘Leave’ would win… with the same percentage.

(BTW: I would be embarrassed if I was working at this company having released these two polls on the same day.)

And yet, despite that we know this, we in the media are obsessed about the specific numbers. Not a day went by without the media reporting that the latest polls showed it was now 46% instead of 45%, even though that variance was well below the threshold of uncertainty.

What we should be doing is to look at the numbers as a pattern. And as a pattern, the polls were completely spot on. Both in showing what would likely happen on the day of the actual election, but also the gradually changing populace during the five years leading up to it.

Sure, you can look at each individual poll, and complain that it wasn’t exactly spot on. Like how the latest YouGov poll before the election showed “Remain” to be 51% with “Leave” to be 49%, but that poll only sampled 0.006% of the population.

I will argue that this poll was spot on too. It very clearly showed that this would be incredibly close.

Why is this important, you ask? Well, because it makes a difference in how people behave. One example is this.

Source: FT

What you see here is the turnout by median age, and not surprisingly young people turned up the least. Please note the deceptive y-axis which starts at 55% and ends at 80%.

We know that young people don’t care about politics, because it’s practically meaningless. During a normal election in the UK, the winner will either be the Labour Party or the Conservative Party, and it doesn’t really matter which one it is. The changes they make are insignificant in relation to people’s daily lives, and the other side will take the leadership anyway in a few more years.

So when young people heard that the Brexit vote would likely be 51% to “Remain”, and thinking that this was just another pointless election like every other they had ever experienced, they decided they didn’t really have to act on it.

Of course, this time it did make a big difference. This election was massively disruptive, but nobody had really told them that. The Leave side had narrowed this whole thing down to the question of a few immigrants, while the Remain side was fumbling about with conjectures that something bad might happen in terms that no normal person could relate to.

So, everything about this failed. We failed to look at the polls as patterns and instead focused on the misleading details. We lulled young people into thinking that their vote wasn’t really needed, and we simplified a hugely complex and wide-ranging question into something that many young people considered to be a pointless argument, in the usual political bickering type of way.

So the real problem here wasn’t the polls. They were absolutely spot on. The real problem is that we have created a political system where not working together is the only way things are done, which is insane. And a system where each election, regardless of what it is about, always ends up 50/50.

We are basically running the world by flipping a coin.