The unanticipated success of the Conservatives in winning a narrow overall majority in the UK general election has raised questions about the performance of the polls, and the British Polling Council (BPC) has announced that an independent inquiry will be conducted into what happened. On average the GB-wide polls pointed to a dead heat between the Conservatives and Labour while none put the Conservatives more than one point ahead. In the event the Conservatives enjoyed a seven point lead.
But how did the polls perform in Scotland? In the following table we show the results of the three polls that were conducted in the final days of the campaign up to and including the day before polling day, and which thus can reasonably be regarded as attempts at estimating the eventual outcome.
There was evidently nothing like an error on the scale recorded by the Britain-wide polls. In particular there is no evidence that the polls were systematically underestimating Conservative support. While two polls slightly underestimated the party’s support, another slightly overestimated it, while the average of all three was almost spot on. That kind of pattern is just what would one expect to see if the errors in the polls were simply the consequence of the chance variation to which all polls are subject rather than systematic bias.
On Labour and the SNP, however, the record is not quite so good. All three polls overestimated Labour support, on average by 2.0 points, only slightly less than the average 2.4 point discrepancy in the estimate of Labour support in the GB-wide polls. At the same time all three slightly underestimated SNP support.
Of course the polls pointed quite firmly to the fact that the SNP were heading for a landslide, while the slight underestimate of SNP support contrasts with the slight overestimate of the Yes vote in the final polls in last September’s referendum. And we have to remember that that there is a greater probability that three polls could all err in the same direction by chance than there is that eight (the number of final GB-wide polls that were conducted) could do so. But perhaps we should not assume that polling in Scotland will not have anything to learn from the BPC’s inquiry.