About two minutes after ICM had suggested the online/telephone divide was a thing of the past, it seems to be opening up again - online polls from other firms now seem to be mostly pointing to a reasonably decent Leave lead, while phone polls from other firms are showing a dramatic swing to Leave, but one that leaves the two sides more or less level-pegging. That's very much the story of tonight's new phone poll from ComRes.
Should the United Kingdom remain a member of the European Union or leave the European Union? (ComRes, telephone)
Remain 46% (-6)
Leave 45% (+4)
It has to be said that at a moment when polls from credible firms are showing a spread that runs from a 7-point Leave lead to a 2-point Remain lead (arguably it runs all the way to a 5-point Remain lead, depending on which version of the ORB phone numbers you prefer), it really is incredibly silly for otherwise intelligent people like George Eaton and Dan "Hokey Cokey" Hodges to look at the ComRes poll and say "Leave needed to be 4 points ahead at this stage if they were to withstand the late swing towards the status quo, they're actually 1 point behind, therefore they've lost". Somebody should update the list of classic logical fallacies to include "only the last poll I looked at matters, and it negates all of the others".
These commentators obviously don't like having to cope with uncertainty, but I'm afraid to say we've got it by the barrel-load at the moment. For example...
* Nobody knows whether online polling or telephone polling is more accurate. If online polling is pretty close to the mark, it seems more likely than not that Leave are currently ahead by four or more points. If phone polling is on the money, it's not even clear which side is in the lead.
* Leaving aside the specific effect of data collection method, nobody knows which pollster has the best-founded methodology more generally. The ComRes datasets don't seem to be out yet, but in past months their turnout model seems to have been particularly favourable to Remain, which may explain much of the difference from the ICM phone poll.
* Even if a pollster's methodology is bang-on accurate, there is a standard margin of error for each individual poll in any case. The ComRes numbers are - just about - consistent with a Leave lead of 4 points (or indeed with a much bigger Remain lead).
* The 4-point lead that Leave supposedly need at this stage is a number plucked out of thin air. Yes, there is a tendency for there to be a late swing towards the status quo, but there's no "iron law" - nobody can be sure it will actually happen, and if it does, nobody can be sure what the scale of it will be. Much will depend on the turnout - if it's very low, I wouldn't be totally shocked if Leave actually do slightly better than the late polls suggest.
* * *
SCOT GOES POP POLL OF POLLS
A couple of weeks ago, I switched to using ORB's turnout-filtered numbers in the Poll of Polls sample, because the firm finally seemed to have come into line with their client (the Telegraph) by presenting those as the headline results. Now they appear to have changed their minds again, and have reverted to headlining the unfiltered numbers while the Telegraph carry on using the filtered ones. I can't keep going back and forth like a yo-yo, so I've made an executive decision that the Poll of Polls will continue to assume that the filtered numbers are definitive. That helps Leave, but only slightly. Just over 60% of ORB's phone sample say they will definitely vote, so the filtered numbers would probably be most accurate if the turnout is in the low 60s or thereabouts - although it must be remembered that one of the problems at the general election was that certain types of people (who were disproportionately likely to vote in a particular way) overestimated their own chances of getting to the polling station.
50/50 ONLINE/TELEPHONE AVERAGE :
Remain 44.6% (-0.2)
Leave 46.6% (-0.3)
ONLINE AVERAGE :
Remain 42.3% (-0.7)
Leave 45.4% (+0.1)
TELEPHONE AVERAGE :
Remain 46.8% (+0.3)
Leave 47.8% (-0.7)
(The Poll of Polls takes account of all polls that were conducted at least partly within the last two weeks. The online average is based on eleven polls - five from YouGov, two from ICM, two from Opinium, one from ORB and one from TNS. The telephone average is based on four polls - two from ORB, one from ICM and one from ComRes.)