Given the huge scrutiny faced by the pollsters since the General Election, last week’s referendum was a key test for their revised methods. We already know that most of them failed to cover themselves in glory, but for the record here are the final polls from each of the main companies:

  • ComRes (phone): Leave 44, Remain 56.
  • Populus (online): Leave 45, Remain 55.
  • IpsosMORI (phone): Leave 48, Remain 52.
  • YouGov (online): Leave 49, Remain 51.
  • TNS (online): Leave 43, Remain 41 (excluding undecideds: Leave 51, Remain 49).
  • Opinium (online): Leave 51, Remain 49.

Evidently there’s further to go in either finding representative panels or improving their weighting and turnout models.

This is bad news for the polling industry as a whole, but it’s also a difficulty for politicians and campaigners. The second most inaccurate result on there is from Populus, whose boss, Andrew Cooper, was the preferred pollster of both Cameron and the Stronger In campaign. We don’t know if numbers they were getting from him privately were similar to those Populus published, but reports that Downing Street staff were reportedly sending gleeful texts to their opponents on the evening of referendum day suggest that was indeed the case.

A false sense of security is a dangerous thing; instead of realising they had a problem, which they could then have striven to fix, erroneous polling seems to have led them to believe they had a fairly comfortable lead. Of course, those same polls could have had a demoralising effect on Leavers, too, but that’s why Vote Leave strove to convert the figures into the message that it was too close to call, and therefore that their grassroots had to redouble their targeting and Get Out The Vote operations. With the wrong numbers to rely on, the pro-EU campaign steamed to disaster.

36 comments for: The pollsters still have a long way to go to fix their models and restore their reputations

Leave a Reply

You must be logged in to post a comment.