The Staggers 23 June 2016 My final prediction is in: and it's Remain Remain is on course to win, says Peter Kellner. Photo: Getty Sign UpGet the New Statesman\'s Morning Call email. Sign-up Take your pick. If the telephone polls are right, then Remain is heading for victory, possibly by a comfortable margin. If the online polls are right, then the result could be extremely close, and we may not know the outcome until breakfast time tomorrow. Here are the final polls, excluding don’t knows: TELEPHONE POLLS: ORB/Telegraph: Remain 54 per cent, Leave 46 per cent Survation/IG index: 51-49 per cent ComRes: 54-46 per cent Ipsos-Mori: 52-48 per cent TELEPHONE AVERAGE: REMAIN 53 per cent, LEAVE 47 per cent ONLINE POLLS: Opinium: Remain 49 per cent, Leave 51 per cent TNS: 49 per cent-51 per cent YouGov: 51-49 per cent Populus: 55-45 per cent ONLINE AVERAGE: REMAIN 51 per cent, LEAVE 49 per cent Those figures incorporate the chosen turnout filter for each company. This sometimes makes a big difference. Here are two sets of figures from TNS and ORB; I have underlined the ones they chose to make their headline projection, on which they wish their accuracy to be judged ORB: All voters: Remain 51 per cent, Leave 49 per cent, certain to vote: 54-46 per cent TNS All voters: Remain 49 per cent, Leave 51 per cent, certain to vote, 46-54 per cent In other words, there is little difference between TNS and ORB’s initial figures, when they count every respondent who takes sides But if we count only those who are certain to vote, then a chasm opens up, with ORB reporting an 8 per cent Remain lead, and TNS an 8 per cent Leave lead. Confused? You should be. Online and telephone polls have mostly been telling different stories; moreover, the polls can’t agree on whether, or how, to filter their figures to allow for differential turnout. Indeed, TNS has muddied the waters even more by changing the basis of its headline figure. Last week, it stressed an 8 per cent leave lead, based on those it regarded as likely voters. This week, it looks at first sight as if the Leave lead has collapsed to 2 per cent. Not so. The Remain-Leave gap is precisely the same as last week, according to both ways of calculating the numbers; but this time, TNS has preferred the measure that points to a close outcome rather than a big Leave victory. To predict the outcome, then, we must do two things: assess the state of play before the start of voting, and judge whether there will be an on-the-day shift. Neither can be done with absolute certainty. Here, with no money-back guarantee, is how I see it. The simplest way to estimate the eve-of-referendum vote shares is to average the figures from the eight polls listed above. This poll-of-polls gives us: Remain 52 per cent Leave 48 per cent Such a calculation, however, offers spurious precision. As I think the online surveys may be overstating support for Brexit, I reckon that the likely Remain vote ahead of today’s vote was 51-55 per cent, with 45-49 per cent plumping for Leave. However, if the outcome is very close, then two groups of voters not covered by the polls might tip the balance: the 23,000 Gibraltarians with the right to vote, and possibly 200-300,000 expatriate voters living abroad. (Sadly we shall not know afterwards what the true number is, for these votes will simply be incorporated into the counts at the local authorities where the expatriates previously lived.) Both groups are likely to vote mainly Remain. This could add 0.2-0.3 per cent to Remain’s percentage and so widen Remain’s lead, or narrow Leave’s lead, by around 0.5 percentage points Now to the possibility of an on-the-day shift. YouGov’s on-the-day Scottish referendum poll, it became clear that more voters were making a last-minute switch from pro- to anti-independence, and that the anti-independence supporters were slightly more likely to vote at all. Together these factors moved the dial two percentage points, from a 52-48 per cent lead for Better Together in YouGov’s previous survey, to 54-46 per cent on the day – close to the 55-45 per cent result. I believe it’s likely, though not certain, that there will be a similar on-the-day shift today to the status quo. Answering a pollster can be done cost-free. Casting a vote is a decision with consequences. We know that some voters are torn between heart and head: the emotional pull of Brexit versus the worries of what might happen to jobs and prices. Do some people respond to pollsters with their heart, and then vote with their heads? My judgement is that, if there is an on-the-day effect, it will help Remain rather than leave; so the overall Remain share will be 0-2 points higher than it was yesterday. Let’s assume the polls haven’t screwed up completely, and the true eve-of-referendum position, including Gibraltar and expatriate voters, was Remain 51.2-55.3 per cent, Leave 44.7-48.8 per cent. Adding in on-the day effects that hover between neutral and a 2 point lift for Remain, the final UK result should be somewhere in the range of Remain 51.2-57.3 per cent, Leave 42.7-48.8 per cent This gives us a mid-point prediction of an 8.5 per cent lead for remain, or a majority of around 2.5 million of votes cast. But don’t be surprised if the gap is less than one million – or as much as four million. And if the phone polls have been systematically overstating support for Remain throughout the campaign, then a victory for Brexit is perfectly possible. My apologies if that is not precise enough for you. If you need a more exact forecast, I suggest you toss a coin or ask an astrologer. This post originally appeared on the Politics Counter. › What the EU referendum looks like when you have nothing else to lose Peter Kellner was President of YouGov from 2007 to 2015. Prior to that, he worked as a journalist for Newsnight, the New Statesman, and others. Subscribe For more great writing from our award-winning journalists subscribe for just £1 per month!