If you're not familiar with using betting data in a situation like this, here's the one thing that's important to know: since about the mid-'90s, the betting data has been far more reliable in predicting the outcome of elections than polls have been – and that difference is increasing over time. Either bettors are getting better information, pollsters are getting worse (or are more overtly biased), or both. I'm guessing both...
My wife loves your blog. She showed me this post last week. Would you mind answering my question as to what evidence you have to support your statement, "since about the mid-'90s, the betting data has been far more reliable in predicting the outcome of elections than polls have been"? Thanks.
ReplyDeleteHi, Unknown! :) There have been dozens and dozens of studies on this phenomenon over the years, often as a specific example of the "wisdom of crowds" testing academics have done using both "real" wagering lines (as at Betfair), or the more academically-oriented sites like the old Intrade (now shut down by government regulators). I haven't kept track of any of them in particular, but a quick web search (which you can easily repeat, with terms like "betting lines more accurate than polls") led me to lots of hits, of which I list just two examples:
ReplyDeletehttp://money.cnn.com/2011/11/15/news/economy/intrade_election_polling/
http://www.telegraph.co.uk/news/general-election-2015/11457970/If-you-want-to-know-who-will-win-the-election-ask-a-gambler.html
There are, to be sure, other studies that call these findings into question. I recall reading a meta-study just before the last presidential election that concluded, in what I found to be a convincing fashion, that the evidence was overwhelming that the betting lines were substantially more accurate than polls - but that there was no evidence about why that should be so...