Are all pollsters equal? Maybe it’s because he’s a Lord, but Lord Ashcroft’s polls seem to be considered more important than most.
Two polls were released today. Ashcroft put the Tories ahead six. Twitter was ablaze and Labour MPs reportedly dismayed. But only a few hours earlier, Populus had put Labour up five.
Ashcroft’s poll is released later in the day, and his considerable Twitter following creates a mini-news cycle, but his Lordship’s poll would still probably bulldoze Populus’ if it was released at 5am. Maybe it’s a remnant of feudal Britain. Or because Ashcroft has enjoyed such a high profile path from Tory grandee and tax non-dom to non-partisan pollster.
Either way, his weekly Monday poll has impact. So what’s suddenly changed since his last poll in December?
When a new poll shows something very unusual, we can see what has really changed by studying the polling tables, or tabs. These break down the voters in a sample into different demographics: from party affiliation and gender to age and class.
There is a fairly clear culprit for today’s shock numbers: Labour’s support among 2010 Lib Dems is only 15 per cent.
We can use these tabs to see how, say, 2010 voters have shifted from one party to the next. Doing this with Lord Ashcroft’s polls unveils a fairly clear culprit for today’s shock numbers: Labour’s support among 2010 Lib Dems is only 15 per cent. In Ashcroft’s last poll it was 28 per cent. And, as May2015’s The Drilldown shows, it is usually around 30 per cent.
The Lib Dems won 23 per cent of the vote in 2010. 30 per cent of that is 7 per cent of the overall vote. Halving that number takes 3-and-a-half per cent off Labour’s vote share. Labour’s share has fallen by 4-points overall (a slightly lower level of support among 2010 Labour voters could be said to account for the remaining difference).
But has this sudden collapse in Labour’s 2010 Lib Dem support actually happened? We need far more polls than just this one, which is based on a sub-sample of just 60 Lib Dems (not that this is untoward: that is the sample size a 1,000-person polls creates). We need dozens of polls with dozens of sub-samples to have enough 2010 Lib Dems to have a basic understanding their intentions.
But has this sudden collapse in Labour’s 2010 Lib Dem support actually happened?
The Populus poll offers a strong contrast. 31 per cent of 2010 Lib Dems in Populus’ poll offer their support to Labour – twice as many as in Ashcroft’s latest. Populus’ poll then puts Labour ahead by five because it shows 85 per cent of 2010 Labour voters will remain loyal in 2015. The Drilldown – i.e. the amalgamation of all pollsters – suggests it will be slightly lower.
The point is, we should only be jolted by shock polls when the changes in the underlying data make sense. Today they make little sense. Labour’s 2010 Lib Dem support has been drifting down recently, but it hasn’t halved. Why would it have? And Labour’s vote hasn’t suddenly become more loyal.
Both sides will seize on the poll most favourable to them this evening, and many pundits will suggest such aberrant polls prove how little pollsters know. Instead, you could just focusing on tracking the trend with May2015’s Poll of Polls, and stay updated with how the polls are likely to translate into seats on election day.