If the polls were wrong for Bernie, they can be wrong for Brexit too

Exit Sign

Here’s a line in the sand, one that stretches from Michigan in US primary season to Manchester as referendum time edges ever closer. Beware pundits bearing wonky opinions.

And beware wonky pollsters who prompt those opinions. Enter James Hohmann, national political correspondent at the Washington Post, wearing the hairiest of shirts.

“The polls were embarrassingly wrong. Hillary Clinton led by 21.4 points on the RealClearPolitics average of Michigan polls. She was up by double digits in each of the 15 surveys released so far this year, including an eye-popping 37 points in a clearly shoddy Mitchell Research survey conducted this past weekend.

“But Bernie Sanders wound up winning by 2 points, 50% to 48%. He prevailed by about 20,000 votes of more than 1.1 million cast. [Polling guru] Nate Silver, who pegged Clinton’s chances of winning Michigan at 99%, describes this as ‘one of the greatest upsets in modern political history’. He says Sanders broke Gary Hart’s 1984 record in New Hampshire for ‘greatest upset vs final polling average’.”

Hohmann adds a necessary mea culpa, though. “Just blaming the pollsters is a cop-out. I really wish I had spent a few days on the ground talking with blue-collar Democrats, instead of relying on the surveys and interviews with ‘plugged in’ operatives. This is another good reminder about the importance of getting outside the bubble and being in the fray.”

Which means, in the starkest terms, that journalists and politicians who blankly and blandly trust the polls aren’t doing a proper job. We learned that last May. We learned it again as Jeremy Corbyn overturned political wisdom. We are learning from afar, time and again, through the American election season.

Look at the recent history of EU referendum polling. Big leads for Remain on telephone polls; narrow leads for Leave (now moving to narrow leads for Remain) when polls use the online route. Can you remotely rely on results like these?

The BBC Trust’s recently revised guidelines on polls and referendums are sterling work. Don’t lead bulletins on a single poll. Look at trends, not single results. Make your own interpretation of the data. Always explain the margin for error. In short: handle with care and due scepticism.

Plus, you might add, along with James Hohmann, turn your damned screen off. Get out and talk to some people. The dirge from Detroit could become humiliation in Hereford and Hampshire.

Powered by Guardian.co.ukThis article was written by Peter Preston, for The Observer on Sunday 13th March 2016 07.00 Europe/London

guardian.co.uk © Guardian News and Media Limited 2010