More statistical proof that Yes can win (courtesy of Fraser Nelson)

One of the hoary old myths that has occasionally been trotted out in this campaign is that there is some kind of 'iron law' in referendums that the Yes vote will always fall and the No vote will always rise as polling day approaches, meaning that the Yes campaign need to build up a substantial early lead to have a good chance of winning. But even on the rare occasions when that claim is unwisely given the time of day by a serious academic, most of us on the pro-independence side are entirely untroubled by it, mainly because we're aware of the obvious historical example that drives a coach-and-horses through the whole notion - ie. the astonishing surge achieved by the Yes campaign in the months leading up to the Quebec independence referendum in 1995. And Dr Matt Qvortrup helpfully reminded us a few months ago that the Yes campaign also increased its support (albeit to a more modest degree) in the Montenegro independence referendum of 2006, enabling it to pass the artificial 55% threshold that had outrageously been insisted upon by the EU.

However, some people go on and on about this supposed 'law' to such an extent that I've sometimes found myself wondering if there's at least a degree of truth in it, and if Quebec and Montenegro were merely unusual exceptions to a general rule. Luckily, I've just had my mind set to rest on that point - and, weirdly, I have Fraser Nelson to thank for it. Earlier tonight, he triumphantly tweeted a link to one of those perverse academic studies that produce a great deal of high-quality evidence, and then cheerfully ignore most of it in order to reach the conclusion that the author had in mind from the start. However, if we set aside the author's commentary and simply look at the actual evidence, it's fair to say that any Yes supporter reading the study is bound to come away feeling more optimistic. Here's what it shows -

1) Big swings of opinion are more likely to happen in referendum campaigns than in regular election campaigns. That always seemed intuitively likely, given that no-one goes into a referendum with an established and hard-to-break habit of voting in one particular way. However, it's still good to hear that this common sense conclusion is borne out by statistical evidence. As we all know, the swing required for Yes to win is significantly less than the SNP achieved in the 2011 Holyrood campaign. It's sometimes said that there were special circumstances in 2011 and that the Yes campaign will be hard-pressed to repeat such a feat - but the reality is that it's actually much easier to do it in a referendum.

2) In six out of twenty-three referendums looked at by the study that took place between 1980 and 1999, the Yes vote increased over the course of the last month before polling day.

3) In five out of eleven referendums looked at by the study that took place after 2000, the Yes vote was higher on polling day than it was in an average of the polls taken in the period between six months and one month earlier. (The Montenegro referendum is not included in this list, because the author only looked at referendums held in "stably democratic countries".)

Those last two points represent a clear-cut open and shut case - the supposed "iron law" has been comprehensively disproven. Not only is it perfectly possible for the Yes side to increase its support during a referendum campaign, the statistics show that it actually happens fairly often. Nevertheless, as you've probably anticipated, the study author thinks he has a get-out clause up his sleeve, which is that in most (but emphatically not all) cases any increase in the support for Yes is relatively modest. But here's the thing - we're still just over eight months away from polling day, and the most recent poll was several weeks ago. The study makes no attempt to assess polling trends in the run-up to referendums over that kind of long timescale - indeed, in the case of the 1980-99 referendums, it only looks at the trend over the very last month of campaigning, thus missing the vast bulk of the Yes surge in Quebec. I think most of us would concede that if, with one month to go, we still find ourselves in the current position (excluding undecided voters) of Yes 40.3%, No 59.7%, then we'd be facing an uphill struggle - although even then the situation wouldn't necessarily be irretrievable, because the study obligingly identifies an extraordinary case in New Zealand where the Yes vote increased by almost 20% over the last month of the campaign. But the reality is that we don't actually expect the big Yes surge (if it arrives) to occur in the final month. The crucial period is more likely to start in late May, when the campaign proper gets underway and the broadcasters face a legal requirement to start providing balanced coverage.

The study author's other tactic for undermining his own evidence is to indulge in a dubious assumption that public opinion in Scotland is fairly settled at present. In support of this, he points out that - according to the polls - the proportion of voters who are undecided is similar to the position prior to the 1997 devolution referendum. But with all due respect to him, those of us who actually lived in Scotland back then know on an anecdotal basis just how much more uncertainty there is in the air this time around. I was talking to my sister the other day, who was a strong Yes-Yes voter in 1997. It quickly became obvious that she was only at the very start of the process of grappling with the issues - she was asking me about extremely basic stuff like the currency, the status of the Royal Family, the nuclear issue, whether it would be possible for Labour to be in power in an independent Scotland, and so on. It was abundantly clear that a) she will definitely vote, and b) she doesn't have a clue how she will vote. And yet if a pollster asked her the referendum question now, I have a sneaking suspicion that she would say 'No', particularly if the silly formulation of "if it was taking place tomorrow" was used. It's no coincidence that TNS-BMRB, a pollster that doesn't use that formulation, produces a much, much higher proportion of undecided respondents than other companies. It seems highly likely that there are many people out there who don't know how they will vote, and yet are still saying 'No' to certain pollsters.

This goes to the heart of a basic piece of human psychology. Suppose I offered you a free cup of coffee. Unless you dislike coffee, the chances are you'll say 'Yes', because it'll seem like a small pleasure with no downsides attached. So I bring you the coffee, but then mention that I thought I might have seen someone spit in it out of the corner of my eye. At that point, you're likely to decide not to drink it to be on the safe side - after all, it's no big deal, it's only a cup of coffee, you're not really losing anything by turning it down. That's roughly analogous to what happened in the AV referendum in 2011, which hardly anyone really gave a monkey's about (although they should have done). In the early polls, Yes had the lead because people were only thinking of the upsides in a situation where nothing much seemed to be at stake, and there was nothing to lose. But then as the campaign progressed, the No side came along and sowed doubts in people's minds that there might actually be something to lose after all - the equivalent of the possible spit in the coffee. At that point the logic turned on its head, and the fact that there wasn't much at stake suddenly meant that there wasn't enough to gain by voting Yes - the "better to be on the safe side" instinct kicked in, and people swung to No in huge numbers.

But an independence referendum isn't like that at all, and it's no coincidence that many of the examples given of a Yes campaign increasing its support happened in major constitutional referendums. A better analogy in those cases might be a situation where your spouse is threatened with redundancy, but is offered the chance to move to a better job overseas. To begin with, your gut reaction is likely to be a provisional 'No', because this is a decision with huge stakes, and in the first instance you're primarily thinking of what you might lose - your current home, your familiar surroundings and routine. But as you get closer to decision day, you weigh up the pros and cons in a hard-headed way, you gradually become accustomed to the thought of a major change, and the potential benefits start to sink in more and more. You might even begin to get rather excited at the prospect of making a fresh start, rather than finding it scary as you initially did. The nagging thought will also start occurring to you that a transformative opportunity like this is not going to come around again any time soon. That, in broad terms, is likely to be why the polls moved in favour of Yes in Quebec and Montenegro, and it's also why it's monumentally silly to lump referendums of all types in with each other in an academic study. The correct comparison for us is with other independence referendums (of which there are very few examples in "stably democratic countries"), and to a lesser extent with other major constitutional referendums.

It should also be pointed out that while the Yes vote increased dramatically in Quebec in 1995, it fell over the course of the earlier independence referendum campaign in 1980. So there's certainly nothing inevitable about a Yes surge, but unlike Ian 'Complacency' Smart we're not actually claiming any inevitability - all we need is the knowledge that if Yes run the right campaign, they have every chance of winning. And fortunately we've got that knowledge.