by Alex Porter
An opinion poll out today, commissioned by the Green party, shows that Labour in Scotland has taken a lead of 9% over the SNP.
The survey, undertaken by YouGov and with a sample size of over 1200 people, puts Labour on 41% on the constituency vote for Holyrood and the SNP on 32% with the Conservatives and Lib Dems on 15% and 8% respectively.
The gap between the principle Holyrood contenders widens to a 14% lead on the regional list vote with Labour on 40% and the SNP on 26%. The conservatives poll equally on the list vote with 15% but the Lib Dems appear to be hemorrhaging votes as only 7% support raises the prospect of the Greens closeby on 6% overtaking them.
Weighting for Godot
The YouGov poll is in sharp contrast to another poll last week by IPSOS Mori which showed the SNP marginally ahead on both constituency and list votes.
Observers might point to the IPSOS Mori poll and suggest that it is a rogue given that the same organisation gave Labour a 10 point lead as recently as November last year. However closer analysis shows that it may be YouGov who are giving Labour continued hope of a victory which may never come.
YouGov’s problem is its weighting methodology. Of those who actually responded to the YouGov poll the SNP have a 13% lead but after YouGov’s weighting the poll transforms that SNP advantage into a 9% disadvantage. How so?
It is perfectly reasonable for opinion poll organisations to adjust their samples so that they reflect the population as a whole.
IPSOS Mori tend to weight their polls according to age, sex and working status using census data. That seems to work out ok with a margin of error which is small enough to mean their survey is reasonably reflective of national opinion.
For YouGov one critical weighting factor is ‘Party Identification’. This works fine in a tribal political system but when you have two parliaments with different voter party identification for each parliament like Scotland does the system starts to show cracks.
YouGov keeps figures on which party Scots identify with when it comes to Westminster elections. Then when they survey for the Holyrood elections they weight the sample until it reflects what they expect in Westminster elections and hey presto a 13% SNP lead becomes a 9% Labour lead.
In the Scottish political system there is even greater party ID differentials between the constituency and list votes. The consituency vote might just take it within a reasonable margin of error but the list votes, as one might expect, show a greater disparity given that the party ID methodology is of lesser value than for the constituency vote.
Not surprisingly this is pointed to by the SNP as a major flaw in YouGov’s methodology. The party themselves uses YouGov and so is perhaps familiar with its strengths and weaknesses.
In England this party ID system may work fine but the further you get away from the Westminster model the more this weighting methodology becomes redundant and the more appropriate becomes the demographic weighting method which IPSOS Mori applies.
As you would expect Labour are happier with the latter of the two polls. Their election co-ordinator John Park says:
“Only last week the SNP were declaring victory before a single vote had been cast. There is only one poll that counts and Labour will spend every day between now and May knocking on doors, speaking to people about jobs and the economy.”
The SNP’s campaign organiser was keen to highlight the anomaly in the YouGov methodology whilst also sending out the message to their activists and supporters that the latest poll news doesn’t mean the party is not very much in the hunt:
“The election is all to play for,” he said. “As voters look toward Scotland’s election in May it is the SNP’s strong record in office, our known and trusted team, and the vision we have for our nation’s future that can win the trust and support of the people of Scotland.”