The eve-of-voting polls are remarkably consistent pointing to 48-52, with 5-10% undecided, in favour of no, so this is going to be the baseline of my prediction prediction. But the pollsters are not at all confident of their weighting methodology.
- 'Don't knows' typically break for the status quo in such referendums.
- There are an unusually large number of people refusing to talk to pollsters. If these break one way or the other, this can make a mockery of polling.
- One side is much noisier and more enthused than the other and there has been intimidation. This can lead to an under-reporting of one side
- There are a lot of people who're voting for the first time and for whom no previous elections can be used to compare.
- Baseline 48-52 for 'No'.
- Don't knows at 5% breaking 2-1 for 'No' gives 47 1/4% to 52 3/4% for no.
- It's simply impossible to know how the Silent voters will vote, but in my experience as a teller, they tend to be older, male, and well educated. Older lean 'no', male leans 'yes' and education is a weak predictor of 'no'.
- I suspect 'No' voters are less likely to take part in online surveys, and be keener to avoid letting on they vote no, for fear of Nationalist flash mobs. I suspect there is a shy 'no' vote nudging it a couple of percent, or possibly more.
- First time voters, and newly registered voters are likely to be under weighted in pollsters methodology, especially if the turnout is very high. It may be this is sufficient to outweigh the 'shy nos'.