Following a couple of elections in which the pollsters have come in for considerable flak for mis-calling the final outcome, last Thursday proved what must surely have been a welcome corrective.
As Sir John Curtice points out in today’s Daily Telegraph, several companies got the result almost exactly right and ten “put their reputation on the line” by publishing polls conducted very close to polling day.
YouGov’s Anthony Wells, writing on UK Polling Report, has provided a handy summary of those final surveys, which we have listed below in the order of their projected Conservative share (rounded to the nearest whole number).
- Deltapoll: CON 45%, LAB 35%, LDEM 10%, BREX 3%
- Survation: CON 45%, LAB 34%, LDEM 9%, BREX 3%
- Opinium: CON 45%, LAB 33%, LDEM 12%, GRN 2%, BREX 2%
- Ipsos MORI: CON 44%, LAB 33%, LDEM 12%, GRN 3%, BREX 2%
- Kantar: CON 44%, LAB 32%, LDEM 13%, BREX 3%
- Panelbase: CON 43%, LAB 34%, LDEM 11%, BREX 4%, GRN 3%
- Qriously: CON 43%, LAB 30%, LDEM 12%, GRN 4%, BREX 3%
- NCPolitics: CON 43%, LAB 33%, LDEM 12%, GRN 3%, BREX 3%
- YouGov: CON 43%, LAB 34%, LDEM 12%, GRN 3%, BREX 3%
- ICM: CON 42%, LAB 36%, LDEM 12%, BREX 3%
- SavantaComRes: CON 41%, LAB 36%, LDEM 12%
- BMG: CON 41%, LAB 32%, LDEM 14%
It is interesting to note that, whilst several companies did get the Conservative share correct, the average still under-estimated it by two points. The spread is entirely in that direction, with a long tail of pollsters beneath the mainland Conservative share of 45 per cent but none erring in the other direction. Is this because there are still ‘shy Tories’, or is there another reason?
With a solid majority, which the polls largely forecast even if commentators and politicians alike refused to believe them, the polling industry can perhaps look forward to a few years of respite. Questions raised during the campaign about the impact of polls on the race, and whether they should be restricted during elections, will take on less urgency.