My Lords, as the noble Lord, Lord Foulkes of Cumnock, noted, following the recent retirement from your Lordships’ House of
my noble friend Lord Ashcroft, I think that I am the only Member of this House who is a pollster by trade. Therefore, I declare my obvious interest and draw the attention of the House to my entry in the register. I am the co-founder of a research company. For well over a decade I have earned my living by conducting research. A very small part of that research is polling, a very small part of which is political polling. The Market Research Society says that less than 1% of all the market research conducted in this country is polling and a tiny fraction of that is political polling. The noble Lord, Lord Foulkes, exaggerates when he describes it as a multi-million pound political polling industry.
I put firmly on the record that I, of course, accept that the recent general election was a serious failure for those of us who produce opinion polls that try to capture accurately the proportions in which people support the different political parties. The polls did not get it all wrong; they were pretty accurate in describing and predicting the scale of the landslide that took place in Scotland. They got the vote share for the Liberal Democrats and UKIP about right, but they got wrong the single most important thing—the proportions of Labour and Conservative votes in England. That was a serious error.
Voting intention polls are meant to scatter either side of a mean—in other words, when you look back after the election, there should be about as many polls a bit over what each party got as there are a bit under. However, that did not happen with the recent general election. All the final polls overstated the support for the Labour Party and understated the support for the Conservatives. In fact, if we look back further, there were more than 1,000 polls in the second half of the last Parliament, and, far from scattering either side of a mean, only one of those polls put the Conservative share of the vote higher than the 38% that they eventually got.
We know that at some point between 2010, when the polls were pretty accurate, and 2015, when they were wrong, tried and tested methods suddenly failed to capture accurately a snapshot of how voters were going to vote, and in what proportions they were going to support the Labour and Conservative Parties. One of the flaws of the regulatory body proposed by the noble Lord, Lord Foulkes, is that, had it existed during the last Parliament, it would have had no point of reference and no way of detecting during that five-year period that suddenly, poll methods which had been accurate had become inaccurate. It is only after they err at an election that we can see that clearly, and at that point, obviously, they must be addressed.
There was a serious failure and the polling industry takes it seriously. Before breakfast on the morning after the election, the polling organisations had all agreed without reservation that a full and open inquiry had to be held. That inquiry was established within 24 hours under the joint auspices of the British Polling Council and the Market Research Society and under the independent chairmanship of Professor Patrick Sturgis, a highly respected academic and director of the ESRC’s National Centre for Research Methods. The inquiry holds its first public evidence-gathering session this afternoon.
Since the general election, I have encountered an amazingly large number of people who are very keen to tell me that they knew all along that the polls were wrong, and they had always foreseen a Conservative victory. I struggle to recall many people who said that before the fact, but I note for the record that the Minister, my noble friend Lord Bridges of Headley, is one of the few who certainly did. As noted in yesterday’s Question for Short Debate, he not only placed a bet a year ago that the Conservatives would win a majority but even correctly predicted the exact size of that majority. As the noble Baroness, Lady Hayter, said, perhaps we should all have saved ourselves the trouble and just polled my noble friend Lord Bridges. It would have been quicker, more accurate and cheaper than polling the 4,000 others whom we polled in our final pre-election poll.
However, it is important to remember that we have been here before, and more important still to remember the lessons of that history. One of the reasons why almost everybody assumed the polls were right was that they had been right for the previous four general elections. However, those four consecutive successes for the polls came after another humiliating failure. As many noble Lords will remember, in 1992 the polls were also wrong—in fact, they were even more wrong in 1992 than they were in 2015. After that failure there was a full inquiry, conducted publicly and transparently, just as there is now. Its conclusions led to a series of changes in the way that voting polls were conducted. These changes by and large fixed the problem. As I noted, at the next four elections, the polling organisations that used those post-1992 methods got the result right. They scattered either side of party vote shares and remained within their margin of error. That is about as accurate as we can expect polls to be, as the noble Lord, Lord Foulkes, noted earlier. Most if not all the changes that were adopted by the industry to fix the failure in 1992, which resulted in the polls being more accurate subsequently, would have been less likely if the noble Lord’s Bill had applied and the regulatory body he proposes had existed.
The noble Lord has expressed a nostalgic attachment to face-to-face polling, as if that is the immutable gold standard of doing a poll, but the switch away from face-to-face polls was one of the central recommendations and conclusions of the 1992 inquiry. The inquiry concluded that for many different reasons it had become too difficult to get a representative sample of the whole population—of all different types of voters—by doing a face-to-face poll, so one of the recommendations was to switch to random digit-dial telephone polling. It was very controversial at the time. Many people opposed it and felt that switching from face-to-face to telephone at a time when only about 90% of households had a fixed-line telephone was a dubious step to take. After the inquiry, some pollsters switched to the phone method; some did not. Learning from the inquiry, some adopted new measures of weighting polls; others did not. At the next election in 1997, the pollsters which had made those switches were accurate and the ones that had not got the result wrong again.
The lesson from the last time the polls were wrong is that we need to define the problem openly, frankly and fully, and then innovate to solve it. The flaw at the
heart of the Bill, in my opinion, is that it would obstruct this process, not help it. The noble Lord’s Bill would give a new regulatory authority responsibility for, “specifying approved sampling methods”. As I say, it is highly likely that in 1992 such a body would have judged telephone polling to be too risky and would have probably stopped the change that made the single biggest step towards fixing the problem and restoring accuracy to the polls. The noble Lord also wants “the wording of questions” to be governed by the new authority that the Bill would create. That seems to be verging on the Orwellian—the idea that we have a state-established body that will decide what you can and cannot ask, and in what terms, seems extraordinary to me.