Winner of the New Statesman SPERI Prize in Political Economy 2016


Saturday 2 November 2019

Tactical voting advice for December 2019


According to a survey commissioned by the Electoral Reform Society, around a quarter of voters plan to vote tactically in the forthcoming General Election. That is a lot of voters, although I would argue the number should be higher. This raises the obvious question, which is who should I vote for in my constituency?

For many it is a good question, because the current polls suggest that the result is not obvious. Take, for example, the seat of Kensington in London. If you look at the last election the answer seems obvious. In 2017 Labour just won with just over 42.2% of the vote, with the Conservatives on just under 42.2%. The LibDems got just over 12% and the Greens 2%. So if you are inclined to vote LibDem or Green it seems obvious to vote Labour.

There is a website that effectively tells you that. Tactical vote 2019 tells you who won your seat last time, and then tells you who to vote for to ensure the Tories do not win. This website is slightly more sophisticated in allowing you to enter your party preferences in rank order, but for Kensington, when I put in LibDem as first choice, it also tells you to vote Labour. There may be other sites I’m not aware of.

But there is a problem. The current polls are very different from 2017. Taking the latest poll tracker from the Guardian Labour now have 25% of the vote, down from 40% in 2017. The Liberal Democrats are now on 17%, which is over double their 2017 performance of 7.4%. Apply those differences to the Kensington vote and the LibDems have a very small lead over Labour, with the Tories winning comfortably.

Best for Britain (B4B) has just launched its tactical voting site that tries to take into account this change in the polls. (I’m not sure if this is instead of or in addition to the People’s Vote’s tactical voting exercise.***) It is more sophisticated than that, because it also applies multilevel regression and post stratification to map national or regional polling into seats. Either because of different timing or different polls, or because of this additional analysis, it suggests that current polling would imply an even bigger LibDem lead in Kensington. As a result, B4B gives a tactical voting recommendation to vote LibDem.

As you can see from the link I wasn’t too happy when I saw this. I am still not happy, but I think as a result of conversations and more thought I’m clearer about what the problem is with what B4B are doing. Needless to say there was a lot of negative reaction to their launch: see here, and here are just some examples. It is completely wrong to think such a reaction is predictable so B4B shouldn’t worry about it. To get people to vote tactically you have to persuade them it is the right thing to do, when they might naturally vote for a different party. (I did hope to do a more systematic analysis of the match between analysis and recommendations, but B4B did not respond to my request for a full list of these.)

This was the major failing of B4B. It looked too much like “the computer says”, with no explanation of why the numbers they came up with are what they are, and why B4B were or were not just following them. There were a number of obvious things they could have done better. First, they could have given the national percentages for the poll on which the analysis was based, so we could see how that related to current polling. This is particularly important given the wide and systematic (i.e. persistent) spread between polling companies right now, due to how they treat their data.

Next, they could have given some indication of why any constituency swing was bigger or smaller than the average swing implied by the numbers they were using. Put the two together and I might have some idea why their predicted swing, towards the LibDems and away from Labour, was greater than the one I calculated earlier. Let me stress that I am not doubting the methodology or implying bias. It is simply that if you are going to persuade people to vote in a way they would rather not, you have to do more than just say "the computers says" and it has worked in the past. You have to give people some understandable reasons.

However, even if it had done all this, I still think its recommendation is wrong. We saw in 2017 that the polls can change dramatically. You can certainly give plausible reasons why that could happen again. Given that uncertainty, the call I would have made for Kensington is ‘too close to call’ at this stage. Anything else will only lead to trouble when you are recommending voting against a sitting Labour MP this distance from the election. (For some of the factors peculiar to Kensington, see this from Ailbhe Rea.)

It would have been much better if B4B had kept its options open in places like Kensington, and instead focused on seats where who to vote for tactically was much clearer (like Canterbury or St. Ives), and in particular on the subset that could make a difference to the result. That would have been far more useful, and less contentious, and actually helped the cause of tactical voting at this early stage. If Labour’s position in the polls improved compared to the LibDems, as it could well do, they will avoid looking foolish. If the polls don’t move or move the other way, they will have broken Labour voters in to the idea of voting LibDem gently.

Does the criticism of B4B invalidate the idea of tactical voting? Of course not. Tactical voting is probably the only thing that can save us from five years of a Johnson government and a very hard or no trade deal Brexit. What it shows is that in a few places, who to vote for at this very early stage in the campaign is unclear. If you happen to be in those constituencies, all I can suggest is watch this space. 

***Postscript (03/11/19) This article describes what happened to the independent People's Vote tactical voting campaign. 

No comments:

Post a Comment

Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time. In addition, I cannot publish comments with links to websites because it takes too much time to check whether these sites are legitimate.