Why did the polls get by-election so wrong?

Welsh identity is strong in Caerphilly, even by Welsh standards. This should have been factored into polling ahead of the Caerphilly by-election, suggests Paul Whiteley

- 

THE results of last week's Caerphilly Senedd by-election were certainly a shock to Labour and to the Conservatives, but they also cast doubt on the reliability of polling as well.

It had for some time appeared that Reform was in the running to win the seat but it ended up trailing some way behind Plaid Cymru.

A Survation telephone poll published on October 16 suggested Plaid Cymru would come second with 38%, and the election would be won by Reform with 42%. The actual results after the October 23 vote were Plaid Cymru first on 47% and Reform second on 36%.

Labour obtained 12% in the poll and 11% in the election, which is fairly accurate. The Conservatives received 4% in the polling and got half of that with 2%. Similarly, the Greens also got half of their predicted share of 3% and the Liberal Democrats got 1.5% following a prediction of 1%.

So what went wrong? The small-sample telephone poll in the constituency which Survation used does not have a good track record.

Surprisingly, national polls are more likely to be accurate than constituency polls.

There are a number of reasons why this is true, chiefly that it is harder to get an accurate sample of respondents at the constituency level than it is nationally. This is particularly true of telephone polls, where the great majority of people approached will not respond.