Monthly Archives: March 2016
The last time I wrote this blog back in January, I made a series of off-the-cuff, not so serious predictions for the coming year. Some were pretty spot on, such as my yearly Oscar best picture prediction or the fact that something called the Zika virus was about to become big news. Others, such as the potential that, despite a slow start, Marco Rubio might eclipse his rivals – well, not so much.
Since I wrote that list I have read Tetlock and Garner’s Superforecasting, which reinforced something I think we all know but rarely practice conversationally — namely that forecasting the future in any sort of absolute terms is a little silly. Back in 2008, one of my colleagues insisted, while gently hitting a table, that Barack Obama could simply “no longer be elected President.” More recently I’ve had colleagues assert that they thought that Jeb Bush was going unify the party, that the most likely outcome of the current primaries will be a brokered convention, and that Donald Trump had little chance of winning the Republican nomination. None of these outcomes is likely, even if, statistically, none of them is impossible.
That we live in a world defined by probability rather than determinism is a difficult thing to instill intuitively in both colleagues and the students they teach. At most universities in most courses, students learn that things work in concrete “provable” (“prove” is a word I ban from my courses) ways – whether it’s in math or science or in the typical humanities or social science essay that calls for a strong thesis supported by a bunch of cherry-picked evidence to support the thesis. Reality, however, is filled with uncertainty and randomness, and the ability to feel one’s way through what’s random and what’s not, the very cornerstone of critical thinking in my mind, is simply not a skill set that students encounter often at their university.
What’s even worse is that the question of what-will-happen-in-the-future seems to be the most common people ask or want to ask. Students often want to write papers about future events like whether or not Great Britain will leave the EU only to have me remind them that it is difficult to “research the future.” Similarly, reporters often want “experts” on TV, radio, and print to use that expertise to tell them what the future holds. As Tetlock and Garner point out, bigger and bolder predictions make for better stories, more popular pundits . . . and worse predictions. It might make for good TV, but what we see with expert political predictions is simply what it looks like when someone enthusiastically expresses his or her personal cognitive biases.
Does that mean that we have no ability to predict the political future? Of course not. It’s just that most predictions that are likely to be right are also not very interesting. Will Putin still be president of Russia in six months. Most likely. Will John Kasich steal the Republican nomination at a brokered convention? Unlikely. It’s the in-between, medium-likelihood events that are the most interesting to predict.
The best way to predict such events – not whether they will happen for sure or not – but, rather, assessing what the likelihood of particular events happening is, is through crowd sourcing. There is a famous story of the discovery of the “wisdom of crowds” by Francis Galton over a hundred years ago. He observed that in a contest involving almost 800 people guessing the weight of an ox, that no one got the weight right. Collectively, however, the guesses averaged out to within one percent of the ox’s weight.
The same principal seems to work with political prediction markets. There are several that operate with “play money” and a couple that, under the justification that they are for research purposes, allow people to log on an invest their own money. Several studies have shown that both systems do a similarly good job of according with the frequency of real world events – much better, in any case, than do most individuals.
Playing the prediction market (in my case for modest sums of money) has not only been fun, but also educational. As a recent Daily Show piece illustrated, “betting” on politics is a sure way to get people politically involved and can raise political awareness and enthusiasm.
Of course, as an educator, I want to bring some of what I have learned to my students. While I don’t plan on introducing a Prediction Market 101 course anytime soon, I’ve raised a few eyebrows at work sitting down with a student office assistant on occasion who has taken an interest in the process. She’s someone who took well to the statistical intuition and basic methods I teach in a course she had with me, and it seemed like a good way to encourage her to continue her education in both politics and the nature of our probabilistic world. Plus, as I’ve reminded her, the 10 dollars she invested is less than 10% the cost of some of her textbooks – a pretty good price for teaching her not to “grow up” to be the next confident windbag.
As for the real answer to questions like: Who will be the next president? The best answer, according to prediction markets, is still that Hilary Clinton has about a 60% chance to be in the White House next year (and about a 25% to be indicted on federal charges before then). Donald Trump only has about a 1 in 4 chance of winning. A brokered convention is about 40% likely to happen and Great Britain is estimated to have about a 1 in 3 chance of voting to leave the EU in June. Jeb Bush is still estimated to have a 2% chance of getting the Republican nomination because, well, nothing’s impossible in statistics.
You won’t be everyone’s favorite conversationalist if you start speaking probabilistically about what you think will happen in the future, but you’ll nevertheless be the most correct person in the room.