30 November 2018

What would happen if Britain held a second EU referendum?


As Westminster continues to be consumed by Brexit and how Britain should proceed, one topic that has begun to be discussed more seriously than it had previously is the possibility of a second EU referendum. The change in rhetoric, if not policy, from senior Labour figures (John McDonnell this week used the word “inevitable”) will have done nothing to dampen this talk.

It should be reiterated that there are many, many hurdles in the path of another plebiscite actually being held, in terms of Parliament, politics and, perhaps most importantly, time. But since there is at least a theoretical possibility of such a vote taking place, it’s worth considering how it might play out.

The most obvious variable is what the question would be. It seems reasonable to expect that it would not simply be a Remain-Leave question, since that would not resolve the source of the current impasse, namely the precise meaning of “Leave”, if that option were to win once again.

On the basis that the plausible three policy options appear to be leaving with a deal, leaving without a deal, or not leaving, the question could in theory be a choice between any two of those three, or a three-way choice using second preferences.

Whichever of these it is, what would matter is how each outcome is preferred relative to the others. Some of the polling on the draft withdrawal agreement has shown more encouraging signs for Theresa May than a fortnight ago, but where does it sit in comparison to Remain and no deal?

Firstly, on the three-way, Deltapoll (after squeezing undecideds) found 33 per cent supporting the May deal, 28 per cent backing no deal and 39 per cent backing Remain. Among those stating a first preference, Survation found 23 per cent backing the deal, 30 per cent preferring no deal, and 47 per cent for Remain.

The questions were quite different, and the results may reflect this. On the head-to-head comparisons, Deltapoll and Survation asked preferences directly, while YouGov asked respondents whether MPs should accept or reject the deal given a stated likely alternative.

May’s deal beats no deal 58-42 with Deltapoll, loses 54-46 with Survation and wins 53-47 with YouGov’s “what should MPs do” question. May’s deal beats Remain 56-44 with Deltapoll, but loses 55-45 with Survation and 56-44 with YouGov. And leaving without a deal beats Remain 52-48 with Deltapoll but loses 56-44 with Survation.

It seems, then, that pollsters cannot agree on where opinion stands now, which may in a large part reflect the difference in questions – and hence what they are measuring.

But what might happen in the event of an actual referendum? Well, one thing that could happen, given that few voters will be expecting one, is that polls react to a referendum being called. This could take the form of switching between options, switching in and out of the ‘don’t know’ column, or becoming more or less likely to vote.

One of the biggest questions is how a campaign would play out. First time around, Leave gained during the short campaign, and it’s possible that the same thing happens again. But equally, there would have been many “soft” voters first time around, with no “default” choice of how they voted last time.

The other question is around the accuracy of polls. In 2016, every poll in the final week overestimated Remain. But that is not necessarily to say that we should assume the same error in a future referendum.

Most polls are weighted to the 2016 result, which in theory means they are measuring the change in opinion rather than the level, which ought to improve accuracy. This also reduces the probability that the high turnout in 2016, if repeated, catches pollsters out again.

At the same time, the strength of feeling is still something polls find hard to pick up. This is clearly still strong among many Leavers, and could potentially strengthen further in the event of a second referendum. But it might also have strengthened among Remainers, and indeed it may be significant that voters in the 2017 election were slightly more Remain than those in the 2015 election.

Ultimately, polling referendums is always difficult and slip-ups shouldn’t be a shock, but it’s unclear who might benefit. It clearly adds yet another layer of uncertainty though.

The data, combined with the uncertainties surrounding how it is measured and how it might evolve, paints a very unclear picture of what a second referendum might produce. One of the two pollsters that asked has May’s deal ultimately winning after second preferences, another has Britain voting to remain in the EU. And while neither has ‘no deal’ winning, one has it reaching a runoff against Remain, and the other has it winning the same runoff.

As such, given a second referendum, none of the three outcomes – deal, no deal, or no Brexit ­– can be completely ruled out.

Matt Singh is the founder of Number Cruncher Analytics.