All commentators and observers during elections put enormous faith in the polls. They give us our only real empirical measurement of how an election may play out. To help our clients and property industry colleagues understand where the pollsters are, the REC team has produced a simple graphic for you:
This lists all the published voting intention polls that have occurred since the General Election was called. A few observations:
First, it demonstrates in stark detail the enormous range in the polls; a Labour lead of both 12 points and 27 points within 48 hours of each other clearly means at the very least that both pollsters cannot be correct. This is why we always caution clients to ignore any individual poll heralded in the media, usually by the media company that has just paid for it! It is the average or trend, the so-called ‘poll of polls’, that probably gives a more accurate indication.
Second, we have split the pollsters into two groups: those which have consistently polled Labour with a lead of +21 points and above (ie above the overall average) and those which have consistently polled that lead at +20 points or below (ie below the overall average). This is the interesting split in the pollsters which a few journalists have been writing about in recent days, the so called ‘excluders’ and ‘guestimators’; those pollsters which exclude respondents who answer ‘don't know’ (tending to give Labour a larger lead) versus the pollsters which try to guess, using various methods, how the ‘don't knows’ might vote (which tends to give Labour a lower lead). As we have said before, polling is a black art not an exact science. This issue is more important in this election because there’s an unusually high percentage of ‘don’t knows’, which could impact the accuracy of the polls and in fact affect voter opinion.
Third, the ‘guestimators’ seem to be moving slightly. It’s small so far but they are slowly creeping closer to the ‘excluders’. Now why is that? It could be that as the election goes on, the ‘don't knows’ are being more specific about who they'll vote for. Or it could be we are beginning to see pollster ‘herding’. This is where, not wanting to be an outlier on election day, pollsters start altering their algorithms to get their results closer to the average. If we dive into the numbers for both groups we can see that the ‘excluders’ rolling average has stayed quite static with Labour around 24 points ahead. Meanwhile, the ‘guestimators’, having had Labour as low as 12 points ahead, have slowly crept up to now having Labour 18 points ahead on average.
The question is where do the polls go from here? Traditionally in the latter half of the short campaign the polls narrow, partly because more people make up their minds as the campaign unfolds – manifestos are published, gaffes occur etc – and therefore report more accurate responses to the pollsters, and partly because the pollsters fiddle increasingly with their algorithms to ‘herd’ to that safe average.
REC will keep watching for you.
Comments