Man-in-Pub-View of YouGov poll showing Labour ahead

The people who actually answered YouGov's question about who they were voting for showed the SNP 13% ahead of Labour. Not behind. Their analysis is almost certainly wrong.

My view of YouGov's probable reasoning, and wait a bit for why it's not totally barking in principle for YouGov to do what they do but merely wrong in practice, goes something like this.

First bit of reasoning by YouGov: We found a huge number more SNP voters than actually voted less than a year ago, so there must be a flaw in how we selected the people to interview. YouGov conclusion: We accidentally talked to too many SNP voters.

Second bit of reasoning by YouGov: We asked these people who are now saying they will vote SNP what party they identify with and overwhelmingly they said they identified with the SNP. YouGov conclusion: These people are consistent SNP voters thus proving that we are correct to discount their number in line with the 2010 Westminster election.

Third bit of reasoning by YouGov: After we apply the discount we get a pattern in Scotland that is consistent with what has happened to the “Other” category of voters in UK surveys since the 2010 election. YouGov conclusion: we are correctly reporting Scottish voting intentions.

So what may be wrong with this? Two key flaws spring out.

When people, immediately after being asked which party they will vote for, are asked which party they identify with, they are very substantially predisposed to name the same party. Even asking which party they voted for last time is fraught with difficulty.

This is because of a “false memory” effect. Intrinsically people don't like to exhibit inconstancy of opinion. So there is a strong propensity to state that the view now held has been held for longer than it has been in reality. Only a few do this deliberately. Most do so unconsciously.

Thus when YouGov discount SNP respondents because these people appear to be long term SNP supporters, they are likely to be in error for a substantial proportion.

The second flaw is more simple. What happens in Scotland need not, and evidence over the long run shows does not, mirror the patterns observed at a UK level.

One only has to consider the Labour result in Scotland in 2010 in a UK context to see that.

Bottom line is that when the interviews show strong support for the SNP, there is probably something in it.

If you want the didactic discussion about the YouGov poll that puts Labour well ahead of the SNP for the May Election look at the good article by Alex Porter at http://bit.ly/i9Mumw.