As the EU referendum draws nearer/ promises to finally be over soon (delete as appropriate), extra attention is being paid to what opinion polls suggest the result will look like. Having enjoyed a consistent lead throughout the early part of the campaign, the Remain side has found itself falling behind Gove, Johnson, and Farage’s Leave side in several recent polls. How accurate such polling is remains to be seen but the apparent recent surge in support for Brexit could already be changing the campaign.
Much has been made of the discrepancy between phone and online polling. The BBC’s poll tracker demonstrates how phone polls consistently show Remain ahead, at times by nearly 10%, whilst online polls paint a more mixed picture. The obvious question here is ‘which method is more accurate?’, the unfortunate answer is no one really knows. Confident Remainers have argued that since phone polls seek out respondents whilst online polls are completed by participants who actively hunt them out, that the former is more reliable. The polling company Opinium has recently changed its methodology to reflect this with its latest online poll giving Remain a 2% lead; it would have been a 3% lead for Leave using the previous methodology. There is however a crumb of comfort here for Brexiteers. If Leave voters are more inclined to seek out online polls, they’re probably more likely to get out and vote too.
Taking a simple mean average of the latest 9 polls (AVE in the chart above) shows the two sides neck-and-neck at 44% each with 12% of voters undecided. Given that the margin of error in each individual poll is usually around 3% and the relatively large number of undecided voters, the actual result on 23rd June could be markedly different without the polls necessarily being wrong. Different averages, however, show different results. Weighting each poll based on how far it differs from the mean, so that outliers have less impact on the total, suggests the leave campaign has a small lead of about 2% (W.AVE in the chart). More accomplished statisticians than myself also weight polls based on date, sample size, and methodology. The consensus amongst most poll trackers, including the FT, the Telegraph, and NCPolitics is that Remain still holds a slender lead. Regardless of how polls are weighted though, everyone agrees the vote is set to be incredibly close.
One final point, not specific to EU referendum polling but relevant nonetheless, is that in attempting to gauge public opinion, polling can actually affect it. The Observer Effect refers to “changes that the act of observation will make on a phenomenon being observed”. Although usually referring to physics, the effect is equally visible in politics. Take the 2015 General Election. Polls consistently showed Labour and the Tories neck-and-neck, both likely to fall short of a majority. This arguably led to fears of a Labour government propped up by the SNP and possibly saw Lib Dem voters switch to the Conservatives to keep Ed Miliband out of Downing Street, whilst expecting a repeat coalition to be formed. In a similar way, the Remain side falling behind in recent polls is likely to spur yet more interjections from notable public figures, and possibly some sort of repeat of the Vow which swung the Scottish independence referendum.