LondonJohn
Penultimate Amazing
- Joined
- May 12, 2010
- Messages
- 21,464
It is not the same as an opinion poll because of the trading of money. That's why if someone predicts something and another says "Wanna bet on it?" it introduces a different consideration, and putting one's money where one's mouth is does not have the same consequences as merely opening one's mouth. Ladbrokes and other bookmakers also have online trade it isn't necessary to go into a betting shop, which may be a social occasion for a certain strata of punter (no idea really)
That said there is bound to be some betting with emotion (rather than analysis) even if money is involved. Bookies can get stuff wrong just as polls can (the traded odds also said "remain" in 2016 for instance, but by a smaller margin than polls). So they are not unbiased estimators but there are grounds to expect them to outperform opinion polls.
Well firstly I'd actually separate this out by class of betting instrument: the dynamics of betting exchanges are markedly different from those of "traditional" bookmakers. In the former, the odds reflect matched trades (e.g. in a two horse race, the amount of money being bet on outcome one x the odds offered on outcome one = the money on outcome two x the odds on outcome two), after allowing for the exchange commission. So in that respect, the odds offerered by exchanges should, all other things being equal, represent a far closer reflection of bettors' intentions on either side of the bet than the odds offered by old-style bookies (who adjust odds on all sides of any bet downwards as a way of protecting their book and making their money).
And secondly, as you and others have alluded to, the intentions of people who actively place bets is far from a proper statistically-appropriate representation of the wider population (in this instance, the population being all those people who voted in the first referendum or who would intend to vote in any second referendum).
In fact, it's extremely difficult to forecast outcomes in this sort of contest. Professional polling & market research companies can certainly select sample populations which at least come pretty close to representing the actual voting population. But, as you point out, what people SAY they would do to pollsters, and what people actually DO do, are often different.
In short, I don't think anything - whether publicly-traded odds, or polls/market research/focus groups of any sort - can predict the outcome of this sort of contest without a margin of error which tends to render the prediction effectively of very low worth. (But of course that's definitely not something that the odds-setters, or even moreso the polling & mkt research operators, would like to admit)