Though many of us have a healthy skepticism about opinion polls and recognise the limitations in interpreting Scottish sub-poll samples of between 90 and 180 respondents, the results of a series by the same pollster can still be suggestive of real distribution in intended voting behaviour.
So, when a pollster repeatedly diverges significantly from the others, we can reasonably suspect that there is something in their sampling or other methods causing that. From 1st June 2019 until last week and ignoring those with only one poll in that period, we get:
Pollster Number Scores SNP average
ComRes 7 Six 3 and one 4 3.1%
Deltapoll 3 One 5 and two 4 4.33%
YouGov 19 Three 5, fifteen 4, one 3 4.1%
Survation 5 Four 4 one 3 3.8%
Opinium 3 One 5 two 4 4.33%
Ipsos MORI 2 Two 4 4.0%
BMG 3 Two 3 one 2 2.66%
As five pollsters cluster around 4% of the full sample and thus greater than 40% SNP support in their Scottish sub-poll, ComRes and BMG are clear outliers. Not only are their averages notably lower but they repeatedly record support at 3% when this is a rare result for the other five pollsters. The non-inclusion of 16 and 17-year-olds may be keeping the scores down but, as far as I am aware, all of them do that. I can’t tell easily whether ComRes and BMG have a sample skewed toward the older respondents known to be less supportive of the SNP but that might do it.
ComRes has attracted criticism recently. See this from the Independent in June::
Predictions of a Margaret Thatcher-style landslide for the Tories under Boris Johnson has prompted a row among data experts. The new poll, conducted by ComRes for the Daily Telegraph, said Mr Johnson was the only leadership contender who could offer victory for the Tories in a general election, translating to a majority of 140 seats……Analysis by Electoral Calculus, a separate firm, translated this support into a 140-seat majority if Mr Johnson was prime minister – only four less than the party won under Mrs Thatcher in 1983…..However several experts cast doubt on the interpretation of the figures, with one saying it was “bonkers” to project national opinion polls onto constituency results during such political uncertainty.
ComRes also hit headlines earlier suggesting that Labour’s anti-semitism problem was a big electoral threat.
I welcome comments from readers of a psephological bent, or anyone, really.
A clue from the Scotsman in 2014: