Welcome

Welcome to my new blog. In the coming days I'll be publishing some insights from a new poll I've commissioned about support for Scottish Independence. If you follow my Twitter account you'll know that I have a lot to say about polling, sampling and weighting. So, I decided to commission my own poll, so that I can look at how different approaches to weighting might influence the findings.

Roughly speaking there are two approaches to weighting Independence polls:

• Based on past vote (particularly the 2014 Indyref)

• Based on demographic factors

The problem with weighting based on the 2014 Indyref voting is that, as we move further away from 2014, the current voting population becomes more and more different from that electorate. Among those who voted in 2014, No voters were much more likely to be older as reported from the Scottish Referendum Study here. Some 67% of those 70 and over voted No; while 62% of those aged 25 to 29 voted Yes. This means that, nearly ten years on, voters who have since passed away are almost twice as likely to be No voters as Yes voters. Meanwhile around 20% of the current electorate did not vote in 2014. So weighting based on the 2014 result will artificially bolster support for No. Oddly, most pollsters still use this approach including Yougov, Panelbase, Survation and Savanta Comres.

But, the longstanding pollster Ipsos Mori doesn't use 2014 voting (or indeed any political events) to weight its samples. It relies solely on demographic and socio-economic characteristics like age, gender, socio-economic grouping and education level. Although this risks having a fluctuating proportion of 2014 Yes/ No voters, it removes the baked-in advantage for No that other pollsters use. You can have a look at polling figures and filter for different pollsters on John Curtice's excellent What Scotland Thinks website. You'll see that Ipsos tends to show a higher level of support for Yes.

My new poll is from Find Out Now which, although a relatively new pollster, is a member of the British Polling Council and Market Research Society. They provide an accessible and transparent approach to polling and an ability to look at results across a wide range of factors. Watch this space for more details over the next few days.