By-election results suggest poll once again over-estimating Labour

Anything other than an exit poll in a Parliamentary by-election runs the risk of being undone by a large, late swing of the sort often found in that very sort of contest. So even if a poll is accurate at the time, it can look inaccurate by the time the votes are counted a few days later.

Armed with that caveat, it is still useful to look at the trend across the Parliamentary by-elections in this Parliament. Polling is a little sparse, but overall there are enough polls to spot a consistent trend if one exists.

And a trend does indeed exist, as can be seen by comparing Labour’s actual result with what the by-elections polls had previously shown:

  • Oldham East & Saddleworth: Labour polled 2% less than ICM and 4% less than Lord Ashcroft*
  • Barnsley: Labour polled 2% less than Survation
  • Leicester South: Labour polled 3% less than Survation
  • Feltham & Heston: Labour polled 1% more than Survation and 2% more than Lord Ashcroft
  • Corby: Labour polled 6% less than the final Lord Ashcroft poll
  • Eastleigh: Labour polled 1% less than Populus, 2% less than Lord Ashcroft and 3% less than Survation (in all cases taking their final poll)
  • Wythenshawe & Sale East: Labour polled 6% less than Lord Ashcroft

On 11 occasions, then, Labour underperformed the polls nine times and exceeded them only twice. Or in 6 out of 7 by-elections, Labour polled less than the polls recorded. Overall, on average, Labour polls 2.4% less than the polls.

Not a huge margin, but enough to remind me of the old days when polling companies regularly had problems with over-estimating Labour support. Nothing yet to match up to the scale of the pollsters’ nadir of 1992, but something to bear in mind if the national polls get even closer between Labour and Conservatives. The Parliamentary by-elections suggest that, whether it’s due to pollster error or late swing, Labour currently falls short of its polling ratings when it comes to actual votes cast.

* I’ve excluded the Survation poll as it had a large number of undecided voters (23%) who were not reallocated according to its usual processes. 

One response to “By-election results suggest poll once again over-estimating Labour”

  1. Exclude “Lord” Ashcroft, who might — just might — be partial. Then take what’s left, and see what is not within “the margin of error”.

    Now reprise …

Leave a Reply

Your email address will not be published. Required fields are marked *

All comments and data you submit with them will be handled in line with the privacy and moderation policies.