On the 9th February 2021, Savanta ComRes for the Scotsman returned the first No majority in nearly a year. The Scotsman had misled readers with its reporting of the data but even after weighting it still did suggest a No lead.
Since then, we’ve had almost exactly as many No as Yes leads and a wee cluster of apparent No leads in the last 5.
But 2 of the last three have been for the Scotsman and done by Savanta Com/Res. Look back over the body of results since Savanta ComRes’s first in February and there’s a pattern.
6 out of the 26, all by Savanta ComRes for the Scotsman, have all found a No lead or in 1 case only, a tie. Savanta ComRes has only once found a Yes lead, for ITV News back in February. Survation, mostly for DC Thomson, also tend to find No leads with 3 of those and only one narrow Yes lead.
In contrast, Panelbase found 2 Yes leads and 1 narrow No lead and both Ipsos-MORI polls found a Yes lead.
A pattern suggests something different being done, perhaps in the construction of their sample?
ComRes has been accused of an ‘inbuilt conservative bias‘ in the past: https://www.ministryoftruth.me.uk/2013/05/02/comres-polls-an-inbuilt-conservative-bias/
Survation was investigated after doing political work’ for Nigel Farage: https://www.bloomberg.com/news/articles/2018-09-06/brexit-pollster-faces-scrutiny-over-secret-political-operations