Polling understatement of the Tories is MUCH less likely to happen when they’re clearly ahead
Why we shouldn’t rely on the “add a few points to the CON poll shares” this time
One of the ongoing themes of this election is that the polls always understate the Tories. Certainly there have been a lot of instances where that has happened as the chart from 538 above illustrates.
But it is a lot more subtle than that. Just go through it general election by general election. The polls were almost running level last time and CON understated. GE2010 saw CON with big leads and in the end LAB was understated. Going back to the general elections from 1992 to 2005 LAB was leading in the polls and the Tories were understated.
In 1987 we have the exception that proves the rule – CON poll leads and CON understatement. GE1983 CON had big leads but LAB was understated.
Mrs. Thatcher’s GE1979 saw a small understatement of the Tories but look at the two elections in 1974. In the first, in the February, CON led in the polls but LAB was understated. In October it was the reverse.
GE1970, the general election I worked on, LAB had a substantial but Ted Heath swept to power.
This is Nate Silver’s comment:
“.. It’s perhaps also noteworthy that while Conservatives have outperformed their polls in eight of nine elections when Labour was ahead in the pre-election polling average — but only four of 10 when Conservatives held the lead to begin with, as they do this year. In 1983, for instance, when the Tories entered the election with a 20-point lead, they wound up winning by “only” 15 points instead. The empirically derived rule of thumb “add a couple of points to Conservatives” could just as easily be “add several points to Conservatives if Labour is ahead in the polls, but don’t make too many assumptions otherwise.” Or it could even be “discount large leads in the polls, because parties tend to underperform them,” which would work against Conservatives this year…“
I’ve no idea what’ll happen on Thursday but I’d be surprised if the “add 5 to the CON share” notion applies – certainly not to ICM and ComRes.