The 650 | 31st May 2017

Projecting turnout is the pollsters’ new battleground

By

Nerves are now certainly jangling in Conservative Central Office, with a YouGov poll last weekend showing a drop to only a 5-point lead, before easing to a 7-point lead yesterday. Survation, with a phone poll this morning split the difference with a 6-pointer for GMTV.

This, from an ICM 22-point Conservative lead just three weeks’ ago.

We too see a continuing Tory tumble with our latest Guardian poll showing a more comfortable, but still rapidly dwindled 12-point Conservative lead. The Tories have dropped two points since our last Guardian poll a week ago, and one point compared to our Sun on Sunday poll published yesterday. Labour remain stable or are up one, depending on your comparison preference.

This poll was completed before last night’s leaders’ grilling on Sky and Channel 4.

The dramatic shifts in polling numbers have been argued in many places to be a function of a sudden surge in young voters and/or 2015 non-voters, motivated by Jeremy Corbyn’s populist platform including the abolition of student tuition fees and return to state funded grants. The Survation poll this morning revealed that 82 per cent of 18-24s would/already have voted, which compares favourably with the next two older age cohorts and is only a tick below that of the uniformly voting 55+ cohorts.

Either this a full re-writing of the psephological textbook or needs to be viewed with extreme caution. Our own poll, suggests that about half that number (44 per cent saying 10/10 certainty) of 18-24s will actually vote (even when full unweighted, it was only 50 per cent).

Clearly, this difference does bring to a head the new methodological battleground. Some pollsters, especially ICM, believe that the turnout scale no longer has value in disentangling voters from non-voters, because the fieldwork process (phone and telephone) predominately fails to reach the latter who are less interested in politics and by corollary, less interested in answering survey research.

ICM stopped phone polling after the EU referendum, partly because we found it incredibly difficult to reach certain demographic groups – especially 18-24s.

We, along with other pollsters typically reached half or less of the 18-24 target (by phone). We note with interest that Survation did a brilliant job in reaching them though – a full 80 per cent of the target number (up-weighting them takes care of the missing residual).

Whether Survation achieved this through full random digit dialling or whether they utilised some targeted sample would be interesting to know, but either way, the great irony about being good at their job is that this success could easily introduce the very skew that kills the poll’s accuracy.

If the 18-24s reached are in some way different to the 18-24s not reached, i.e in saying they will disproportionately vote and vote Labour at that – when their wider counterparts will not and do not – it’s likely that the same failings of 2015 will be very much embedded in this sample.

So how pollsters address the turnout issue is now central to what the poll says. We at ICM turnout weight using a matrix that assumes younger people will be less likely to vote than older, and less affluent people will be less likely to vote than the wealthy. This has been the general pattern of general election’s for an age, and whether you believe our poll findings or those of others will depend on whether or not you think Jeremy Corbyn can actually buck that trend.