We now have the tables for the YouGov Welsh Barometer poll for ITV Wales and Cardiff University. The poll’s headline numbers show Labour’s lead at 5 points with 33 per cent of the Welsh Westminster vote. That’s one of Labour’s worst ever Welsh polls, though it’s not significantly different from September, and it represents about the same swing as recent GB polls have been implying in England.
But it’s possible that in reality the swing we might expect in a real election is even bigger, because the gap between Labour and the Conservatives is almost entirely accounted for by people that don’t normally vote. Labour leads the Tories 44-19 among Remainers, trails 18-40 among Leavers, but leads 52-7 among people that didn’t vote last June, but claim that they would in the future.
While 2016 taught us to be cautious when it comes to making assumptions about turnout, it’s nevertheless still reasonable to think that people that didn’t vote in a referendum with a 72 per cent turnout probably aren’t the likeliest of voters in a general election. The risk for Labour is that their lead is soft because they’re relying in part on this type of (non-) voter and the Tories aren’t.
If we re-filter the poll to look only at those that did vote last June, the result is a 30-30 tie, which would be Labour’s second worst Welsh poll ever, including the period in which they were in government at Westminster.
Obviously this should be treated with caution – it’s a crossbreak, even if it’s a very large one – and the poll is weighted so that the full sample matches the demographic and political profile of all adults, not that the voting subset matches the profile of voters. So it could conceivably be an artefact of the weighting process. (I’ve had a look at YouGov’s recent GB polls and there’s no sign of the pattern there.)
And in any case, it’s just one poll, so a rogue reading is always possible. But it looks from this as though the balance of risks is a bit more towards things being even closer than they look.