In fact, it seems that this referendum is such a gamble that the media, both here and in Britain, have been jettisoning polls and have begun replacing them with bookmakers’ odds. The Spectator has been offering predictions on the referendum’s outcome based on “implied bookies’ odds”. Meanwhile, a business reporter on CBC radio the other day was heard to claim that bookmakers’ odds are perhaps more accurate than polling data because they capture “people putting their money where their mouth [sic.] is.” If this seems a bit outlandish, consider this: The Bank of England is apparently using bookmakers’ “implied betting odds” on Brexit to assess how much the referendum uncertainty has hurt the pound.
Now I recognize that opinion polling has produced some spectacularly erroneous predictions of late, including in both the last British election and the Scottish independence referendum. But despite the challenges that polling faces in the age of digital media, is it really wise to rely on bookmakers odds to set policy and predict voting intentions?
The short answer is no, for the following reasons.
First, betting is not in any way analogous to voting. A wagerer’s “vote” (i.e. wager) does not necessarily express how he personally feels about Brexit. Rather, it is an expression of his belief about how other voters – other real voters – will vote on Brexit. I may personally believe that, all things considered, leaving the EU is the best thing for Britain, while at the same time believing that I am in the minority. Hence, I may vote for Brexit while I bet my money on Remain.
Second, just as a wager is not the same thing as a vote, a wagerer is not the same thing as a Briton. Foreigners and others not eligible to vote can easily place a wager. Considering wagers as votes essentially means counting the votes of many ineligible voters, further polluting the data one gets from bookmaker’s odds. And a wagerer can place multiple bets, while a voter only gets one vote.
Third, even if each wagerer were also an eligible voter, and even if voting intentions reliably corresponded to betting intentions, the odds offered by bookmakers almost by definition will not correspond to those intentions exactly. If a bookmaker accurately and dutifully offered odds that reflected wagerers’/voters’ actual intentions, he would make no money.
For an oversimplified example, let’s say that precisely 50% intend to vote “Leave” and 50% intend to vote “Remain”. If the bookmaker constructed his book such that, if “Remain” wins their wagerers get all the stakes of the losing “Leave” wagerers, then there is no profit for the bookmaker. He must construct his book as if there was, say, a 49% chance of either outcome, so that whatever happens he can pocket the stakes of the 2% in between. In other words, his odds will never reflect the 100% chance of the “either Leave or Remain” outcome. The house must always make its end, whatever the outcome. That’s why it’s a racket (look up “Dutch book” to see what I mean).
(The bookmakers’ odds in the Spectator article referred to above do add up to 100%, which, to be fair, is why they’re called “implied” odds. However, it is not clear how the implied odds are arrived at, nor how, say, undecided voters fit into all of this.)
Fourth, although the bookmaker must construct a Dutch book, some Dutch books are better than others. For example, if one outcome looked lopsidedly more probable than the other, too few people would be attracted to bet on the less favoured outcome — there wouldn’t be enough money in the losers’ pot to pay out the winners. The bookie would take a bath. So the bookmaker must offer some inducement, in the form of a greater payout, to induce enough people to place bets on the long shot. Insofar as this attracts some people to “vote” for the unlikely outcome who otherwise wouldn’t, it further distorts the “polling” data.
No one ever went broke betting on the ignorance of journalists. It’s rather more disturbing, however, to find that the smart money is on the ignorance of the Old Lady of Threadneedle Street.