Photo: Liam McBurney/PA Wire/PA Images
How would our politics be without opinion polls? Well, pretty certainly pretty different. I go back with polls a long time (having been polls adviser to the Labour Prime Minister Jim Callaghan) and can remember both the struggle to establish them as accurate data and then the pluses and pitfalls of their existence.
The struggle first: the left of the Labour Party did not like opinion polls because they tended to show their shibboleths—nationalisation, nuclear disarmament—were not popular. The right liked them because they showed that moderate policies were what the people wanted. Many National Executive Committee meetings were taken up by futile debates on the subject. But successive leaders of the party wanted polls, and commissioned them.
This was not always to their advantage. Callaghan’s decision not to hold a general election in autumn 1978—as the report of the recently published Lords Committee on Political Polling and Digital Media, which I chaired, recalls—owed much to polling in marginal constituencies which showed that Labour was unlikely to gain the seats we needed for a majority. Unusually Labour’s excellent pollsters, Mori, failed to point out that the sample sizes in their survey were too small adequately to support their conclusions. Unfortunately Callaghan, a great man but no statistician, was moved by the crude findings. The rest is history.
If the polls are accurate, they of course add to knowledge of what is going on in politics. Arguably this bolsters democracy. However in Britain the polls have managed three catastrophes in succession: the 2015 general election where they predicted a hung parliament, the European Union referendum where they expected “Remain” to win; and the 2017 general election, which predicted a shoo-in for Theresa May.
It will be interesting to see if the polls currently being conducted for next week’s local elections continue that worrying trend or provide a return to form for the pollsters. To misquote Oscar Wilde,” our report says “to get one election wrong may be regarded as a misfortune, to get two wrong looks like carelessness, and to get three wrong suggests something somewhere has gone horribly amiss.”
There’s no way to tell if these are blips on an honourable record, or if rather polling is now dead as the dodo. There are however reasons to think polling is getting more difficult. One is the increased reluctance of people to answer pollsters’ questions: even when conducted online, samples tend increasingly to be skewed towards those with a particular interest in politics who may or may not have similar views to those who have no interest. Secondly, once if you made sure that your sample reflected the class structure of the population you couldn’t go far wrong. The working class voted Labour and the middle class voted Tory. Today more middle class voters vote Labour than working class voters.
Then there is the pollsters’ dirty little secret: the margin of error. Without much scientific basis, pollsters assume a conventional margin of error of +/- 3 per cent. But even if this is right, it is much larger than most people realise. Take a poll reporting Labour 40 per cent / Tory 40 per cent. Allowing for the margin of error this could be Labour 43 per cent / Tory 37 per cent, meaning a 6 point Labour lead, or Labour 37 per cent / Tory 43 per cent, a 6 point Tory lead. So how much real information is the public getting from such a poll?
So polls should be treated with a pinch of Saxa. Some would go further. We could ban polls in the run-up to general elections, as do 16 of 28 EU countries. This would stop all focus being on the “horse-race” and allow voters to concentrate on policy issues. Or we could state-regulate polls more strictly, as to an extent does France with its Commission des Sondages. In principle the Commission can condemn polls that do not meet its standards, though to be fair many in France question its effectiveness.
My Select Committee drew a deep breath and resisted going this far. We did not want to interfere with the electorate’s right to information; and were concerned that polling might move offshore even further beyond our control.
However we do support two important sets of changes. The first is a strengthening of the industry’s own British Polling Council, so for example it holds open public inquiries into poll performance after every general election. The second is more powers for the official Electoral Commission, so as to identify who commissions what polls during elections. In addition we want better media training in the reporting of polls. We support the government’s Digital Charter and would like to see a Lords’ select committee especially focussed on digital and social media in politics, especially in the wake of the Cambridge Analytica affair.
If these things do not happen, it seems inevitable that politicians will again look at regulation and bans after the 2022 election, particularly if it produces another polling fiasco.
