renton

Unless they are going from their last ICM/Scotsman poll. Because it's Y down 6, N up 6 based on the 700 size poll from Saturday and 1 down based on the Guardian's telephone poll.

It's definitely 48 Y 52 N

3.8% down to 3%, it's not really a huge difference, it's just that saturday's poll threw out one at the extreme of it's distribution.

No......

There you have it, massive outlier from the 700 size sample on saturday. You can take it as massively down on that poll or you can take it as 1 down from the ICM phone poll. take your pick. Still statistical tie, looks likely we'l be going in at this kind of level into referndum day.

They've been better with embargos since that 51/49 poll, apparently they keep getting city traders phoning up looking for inside information.

Behave. Last one was a massive outlier, I'd take 50% right now.

I see bet fair have decided it's all over: now paying out on a no vote.

It's true we can't ignore previous referndums, but at the same time, different actors with different demographics,different socio-economic factors, different policy issues and 20 years of global change means theres a limitation as to how much read across from Quebec you can have.

really, I was under the impression that this was a group that never needs much in the way of upweighting, and indeed has been relatively stable across all the pollsters.

Depends on which you do a poll of polls, there are a couple of different methodologies - either the last set of polls in a month, or the last six polls by all the involved pollsters, at the moment doing a poll of polls base don the latter methodology is likely to depress the Yes position as it includes an oldish Ipsos-Mori (though it looks like that will be shortly remedied). The other issue with poll of polls is that for a long time they've merely generated a false middle through wildly different methodologies - that's lessened now but again, with ipsos in there you are still likely to see a big No lead even though everyone else reckons it's even stevens. based on the run of the last few polls from PB, Survation, ICM, YG and TNS I'd reckon it's 50/50 - we've seen Yes in the range 46 to 54 over those polls but removing the outliers sees yes in the range 49-51. I'd also go back to something I noted previously, that big yes swings were in Younger groups and C2D2E percentiles - both groups likely to make up a lot of the newer voters who've been undersampled in BPC pollsters. Against that, you've got the 'silent majority', 'shy no' narrative of which there is absolutley no proof of it being a major phenomenon, but there is the more real possiblity of Yes shedding a couple of points based on previous referendum performances. How do those two factors cancel out? Does the yes group have a soft edge? Given how the Yes vote held up after last weeks blitz you'd think not much, but then they might waiver when in the actual booth. Will those big swings to Yes in lower income and younger age groups seen by the RIC and the BPC pollsters actually turn up on the day? I don't think there is much in it, I don't on the face of it see a reason to suppose that No has an advantage at this point, that might change tonight with TNS due out at 5pm and you'd assume the YouGov one out late tonight for publication in the papers tomorrow.

A No vote is as likely as a Yes vote based on the current polls.

No, merely that an inability to find enough 'real' respondents in a given poll leads to a large upweighting in those groups, and with it and upwieighting in the statistical margin of error for that group which can lead to inaccurate or volatile results (see Survation and it's inability to find enough 16-24s). I figure that Ipsos consistently No friendly results are something to do with their demographic breakdown of the populace - upweighting in certain age groups would introduce volatility from one sample to the next, but not a consistently depressed Yes vote with repsect to the other pollsters.

Yeah, they've has that problem through the entire campaign.

Wrote a long rambling thing about it a month or two ago, I said then I thought it would be 45/55 No, based entirely on what the BPC pollsters were showing. In that time Yes have managed to close the gap with YouGov, Panelbase, TNS and Panelbase now all showing a statistical dead heat, so Yes looks good for anything in the range 48-52. Even when I've been hopeful of a Yes win, I've never seen us winning by a big margin - assuming the Yes vote holds up between now and then and you get them out on the day, I reckon 52-48 Yes. That's flipping a coin, it could easily be the other way, but mind that the BPC guys all carry a +/- 3% margin so with the current polls there is a huge degree of uncertainty - hence why YG is doing a large sample poll to try and get that polling MOE down to +/- 1%

Mebbes Aye, mebbes No. Pollsters struggle with false recall, so that you get folk who say they voted in (say) 2011 but actually didn't, or the other interpretation is that they genuinely do find more folk who voted than didn't, thus oversampling on one part of the population. With online polls I'm inclined to the latter, as they tend to self select volunteers from people who are politically engaged. Having said that, they do see large swings to Yes in the C2D2E percentiles and a better Yes turnout in younger age groups, both of whom will be prevelent in those new voter classes. Also, RIC canvassing returns and registration drives did show strong yes returns amongst that C2D2E sub section of the populace (on the other hand they usually threw up enough DKs so that if they broke No, those impressive yes percentages would be eroded quickly). More anecdotally, I met a guy on Saturday who'd never voted once in 35 years, pretty much his whole voting eligible life, but had registered with the intention of voting No. He didn't particularly care who ran the UK or how they did it, just so long as there was a UK. Takes all sorts. So yeah, the official pollsters do see positive news for Yes in terms of those groups who likely make up large chunks of the newely registered voters, and this is somewhat backed up by the RIC canvassing, but I'll bet it's bloody fragile and while there is a probability that new voters will swing to Yes, there is only limited data to support that and enough uncertainty for me to make it more a matter of faith than fact.

Every poll going breaks down the data by gender, as well as post voting behavoiour and social group. Yeah, women do tend to lag behind men but the gap has closed to a degree.

rounding errors.

And No have all three major 'national' parties and their support networks, a media machine that largely favours the status quo and 300 years of historical inertia. These are all 'massive' advantages as well. ... and no one is saying the media is the reason yes will lose, the stories of this week were all on the face of it 'newsworthy' even if I personally found the reporting style sensationalist and simplistic (but that goes for most reporting). Fact is though that having thrown everything from the IMF and banks moving to increased food prices, and the Yes vote didn't melt away as clearly was hoped for. It's still too close to call, and all the polls have basically converged to a statistical tie, with yes ahving gained about 20 points over the last couple of years.

Some guy tries to analyse the referendum outcome by studying the flow of 'big data'.... make of it what you will. http://blog.majesticseo.com/research/scottish-election-poll/

I don't think anyone is saying that, I think there is a general sense of relief/surprise that the Yes vote has held up after a week where Westminster has clearly pushed what it believes are the consequences of a No vote very hard, not a case of 'blaming the media' more that there has been no discernable drift away frm yes after a concentrated drive by the No campaign.

Well, there has been a year long hunt for 'shy no' by the pollsters, of which the Kellner correction is the obvious example. There is not even any real anecdotal evidence to suggest it either - it's your opinion that it favours Yes and you could well be right, at the same time the sheer weight of negative press Yes gets and particularly in the last couple of weeks could easily make folk shy about admitting a yes vote in the face of being told they don't love their families or they are voting for a financial basket case. Not saying your wrong, just saying that there is perfectly legitimate arguments for why it might favour the other way as well.

Like TNS and their face to face polling, or the RIC and their face to face canvassing, actually talking to another human being about your opinion tends to cause a larger degree of reticence than the online polls where anonymity is guarenteed and you are self selecting from particulalry politically engaged people

Screw that, that's a great poll for Yes, on a traditionally bad methodology for yes, No ahead by a literal bawhair, a weke to go and with the onslaught of the last week, the Yes vote held up.

Combination mobile/landline poll. Will be interesting to see as it's the only ICM phone poll to date - and traditionally Yes have struggled in the Ipsos phone polls. They've brought it forward massively as well which suggests a bad result for Yes as well.