Professor John Robertson OBA

In the last year, from the YouGov poll of 3-7 June 2024 until the Norstat poll of 27-30 May 2025, the average Yes or No lead:

Norstat (6 polls) – Yes lead average 1.83;

YouGov (5 polls) No lead average 10;

Survation (5 polls) No lead average 4.2;

Find out Now (4 polls) Yes lead average 3.5.

So, all of these pollsters registered as following the same guidelines consistently, should not be discovering such wildly diverging populations. That YouGov should find, over the year, an average of a 10 point lead for No and Find Out Now should, theoretically looking at a scientifically similar sample of the same population, find an average Yes lead of 3.5 points – a 13.5 point average divergence – means they are not following the same guidelines and one or other, or both, is skewing the results by biasing the sample.

The most obvious explanation, to my mind, lies in the decision by YouGov and Survation to adjust their initial sample to make sure it repeats the 55/45 ratio from the 2014 referendum result and to predispose another No outcome.

Norstat do a similar but smaller adjustment, to between 52/48 and 53/47, and get a consistent but only marginal Yes lead.

Find Out Now which find the largest consistent Yes lead, do not adjust their sample according to the 2014 result, at all.

What does this mean? Find Out Now does not adjust their sample at all and thus reflects more closely, reality – there has been a steady shift, in the last 11 years, as the electorate membership changes, from a 55/45 No lead to a 46 or 47/53 or 54 Yes lead.