It is important to resist the temptation to cherry-pick polls that support one’s caseby Peter Kellner / December 18, 2017 / Leave a comment
Sorry, Remainers. The BMG poll for the Independent, which caused such a stir over the weekend, looks wrong. The Remain lead, which has hovered around 1-4 per cent in recent weeks, has NOT suddenly widened to ten points in the wake of the government’s embarrassing defeat last week in the House of Commons.
Here are four reasons to be wary of BMG’s figures.
First, they are mainly driven by a seemingly huge shift in people who did not vote in last year’s referendum. Sure, the Remain camp is being swelled by young adults who abstained last time or were not old enough to vote. But previous BMG polls included this phenomenon.
During the summer, when previous non-voters were asked how they would vote in a fresh referendum, they divided fairly steadily: around 45 per cent Remain, 25 per cent Leave. Now, suddenly, BMG say the divide is 67 per cent—16 per cent. As this group comprises more than a fifth of BMG’s total weighted sample (299 out of 1,363), this 51-point Remain lead within this group accounts the reported overall Remain lead.
However, BMG did not actually interview 299 previous non-voters. Its unweighted subsample was barely half that: 156. The margin of error on such a subsample is large—and the very fact that BMG could not track down as many non-voters as it wanted, provides a clear warning (as I know from my experience at YouGov) that the sub-sample may not be as representative as one would like.
Secondly, if it were true that, despite these sampling issues, ther…