The EU Referendum: One Week Away. Time to Freak Out.

The numbers don’t look terribly promising for the remain camp.  There has been a significant shift in the polling data to leave across every polling house in the past two to three weeks.

Clearly, leave are winning the argument. How the trio of clowns leading this argument are winning it is mildly shocking, but winning it they are. It’s always been an easier argument to make: sovereignty good, immigrants not. It’s an argument, if crafted well, plays on raw emotions. This has been crafted well. Remain have to talk up the status quo, make a strong economic case (which isn’t exactly sexy) and say the occasional nice thing about Brussels, while simultaneously dodging the question of the free movement of labor.  I’ll have more thoughts on these as the week progresses, but today I want to discuss what we should be looking for in the week ahead.

One week out of the Scottish independence referendum, I was very confident in a victory for the union. While the pro-union camp freaked out over one poll that showed independence ahead, I wasn’t freaking out. Indeed, in media work in advance of the referendum, I came within half a point of predicting the outcome, an assessment based on extant polling data and political science. In short, I nailed it (and did better than the polls). How I know I nailed it is that after the referendum, two different radio interviews played back my original predictions. A moment of sheer terror quickly turned into relief in happily accepting the congratulations of the presenter.

I’m not going to nail this one. Up until about three weeks ago, I was consistently predicting a 52-48 remain vote. Now, I honestly don’t know, but if I were to place a bet, it would be on Brexit (and maybe I should be a betting man: as of this morning the bookies were still giving remain a 60% probability of winning). Unlike in Scotland, where one pro-independence poll freaked people out (and there were only ever two polls that showed independence winning) the past couple of weeks of polling have shown not only consistent movement in the direction of leave, but also most polls show leave with either a narrow or significant lead. The second problem we face in trying to forecast the results of the referendum is that the various British polling houses have all been continuously tinkering with their methodologies (some sampling, but mostly likely voter models) such that the numbers are all over the map. Granted, they failed dramatically in advance of last year’s general election, but they were quite reliable in predicting the result and vote share of the Scotland referendum. So why change what worked well in the binary choice environment of Scotland? Additionally, there has been a pretty consistent (but not absolute) split in the mode of a survey. On-line surveys have estimated larger levels of support for leave than phone surveys, and I have a pretty simple guess as to why this is (such that when the first phone surveys also began to suggest a leave victory, that’s when I started freaking out).

With those caveats, disclaimers, and a general lack of a clue as an introduction, this is what I’m looking at in the next week and on the day:

First, the undecideds. In polls that report undecideds, the higher that number, the better for remain. In referendum voting, the status quo has the power of incumbency, and the closer we get to polling day, the larger the probability than any given undecided voter will vote for the status quo. Thus, stories like these, about an LSE study suggesting that up to 30% of voters will not decide until the final week (and half of those on election day itself) should give supporters of remain some hope.

Second, turnout. Bluntly, if turnout is higher than expected, the odds are better than remain will win. This is an easy one. As I mentioned in my last Brexit post, there is a significant and substantive relationship between age and support for staying in the EU. The young would much rather stay in the EU. This is rather unfortunate for the remain camp, given the young would also much rather be doing virtually anything else than, you know, voting.[*] Increases in turnout do not have a straight linear effect on all subcategories of the overall population; a one-percent increase in overall turnout has a stronger effect on those categories voting at lower rates. More voters means an asymmetrically larger share of young voters.

I usually know in advance how an election will turn out (or at least in the case of the general election last year, at least I thought I knew).  This time I’m in the uncomfortable position of not knowing.

[*] Note, this is explicitly not a “kids these days” argument. The youngest cohorts have traditionally been the age group that turns out the least.