Why referendum polls are biased towards Leave

It’s now confirmed that the EU referendum will be on June 23rd 2016. No doubt there will be a huge number of polls before then adding to an already substantial list. Given the failure of polling at the last election and the currently huge differences between polls (particularly between online and phone samples) no-one would suggest they are a solid guide to the eventual outcome. Nonetheless, polls make headlines, shape the debate and affect morale and the surrounding hysteria is only likely to grow as the day of the vote approaches.

It’s odd therefore that a small, deliberate but significant error in these polls generally goes unnoticed. That is the exclusion of Northern Ireland.

Despite Northern Ireland being the one part of the UK to border another EU country, most polls only contact people in England, Wales and Scotland. In a general election, where the parties of Northern Ireland are completely different, this is understandable. Ahead of the referendum, where all votes are equal, it’s harder to forgive.

What’s more, Gibraltar, while not part of the UK, also has a vote (or 20,000) in the referendum, yet is also excluded from the polls. The same is true of Brits living overseas (but not for more than 15 years) who number several million. Only around 100,000 of these were on the electoral register at the last election, but even that is electorally significant when they like Gibraltarians will surely overwhelmingly vote for ‘remain’.

So there are three groups of voters who can walk to another EU state those in Northern Ireland, those in Gibraltar and those living abroad (especially in Spain, Ireland and France) and none are included in the polls. But the exclusion of Northern Ireland is probably the most significant and least defensible. What kind of difference does its exclusion make?

Well, Northern Ireland is relatively small but still significant in any vote, with around 3% of the UK’s population (2 million of 63 million). Omitting it wouldn’t make too much difference if it had similar views to the rest of the UK, of course. But we’ve got good reason to believe it doesn’t. A poll last year found a massive 58-16 split in favour of remaining in the EU, and a survey of businesspeople this week found an 81-11 split. Another (though perhaps less robust) showed that Unionist voters were against but still found an overall 56-28 split.

We don’t know how turnout will differ across the country but let’s assume that Northern Ireland (perhaps together with those expats and Gibraltarians) will provide 3% of voters and that they’ll split 2 to 1 in favour of remaining in the EU. With those assumptions, in the rest of the UK the part covered by the polls remain only needs 49.5% of the vote. Leave would need 50.5%. With the extreme assumption that those 3% of voters outside Great Britain all vote to remain, the ‘in’ campaign would only need 48.5% of the mainland vote.

So the exclusion of Northern Ireland doesn’t completely undermine the polls: other sources of error are likely much bigger. But we should still remember that systematic bias towards leave, especially when the polling is close (and that’s on top of the fact that ‘don’t knows’ tend to disproportionately support the status quo in any referendum). It would be nice, however, if we didn’t need to guess at how the views of Northern Ireland would change the polls.

Leave a Reply

Your email address will not be published. Required fields are marked *