I once heard a speech on statistics entitled Probability is Hard but if you believe some political pundits like David Brooks a more appropriate name would have been Probability is for Wizards. While believing that polls and the economy can be useful indicators of winning an election Brooks apparently doesn’t believe one can ever place a numerical value on the chance of any outcome. As he told PBS earlier this month:
If you tell me you think you can quantify an event that is about to happen that you don’t expect, like the 47 percent comment or a debate performance, I think you think you are a wizard. That’s not possible. The pollsters tell us what’s happening now. When they start projecting, they’re getting into silly land.
The italics are mine, the basic misunderstanding of probability is his. Some, most notably Ezra Klein, have pointed out this failure of math is widespread among the punditocracy as, among others, Politico’s Dylan Byers and Jonathan Martin, The National Review’s Josh Jordan, The Daily Caller’s Matt Lewis and MSNBC host and former congressman Joe Scarborough have all recently demonstrated an obvious inability to comprehend statistics. Ironically, in attempt to invalidate statistical models many of these people have demonstrated the overconfidence bias as their subjective confidence that their holistic judgements of where the presidential race stands far exceeds the objective accuracy of subjective holistic judgments.
They also seem to be failing to grasp two fundamental concepts in probability. Firstly, they don’t seem to comprehend that a prediction that an event has a 75% chance to occur isn’t refuted if that event fails to occur. After all there was a 25% chance that it wouldn’t occur. None of us would approach a surgery with a one in four chance of death as certain to succeed.
Secondly many, notably Brooks, don’t seem to understand probability is specifically designed to handle situations in which we aren’t sure about the outcomes. He demonstrates this when he says:
First, I should treat polls as a fuzzy snapshot of a moment in time. I should not read them, and think I understand the future.
If there’s one thing we know, it’s that even experts with fancy computer models are terrible at predicting human behavior. Financial firms with zillions of dollars have spent decades trying to create models that will help them pick stocks, and they have gloriously failed.
In old news, the past is often a good guide to what happens in the future. Collecting data on past events and people’s declared intentions and using that to help predict what will happen in the future isn’t something unique to political polls rather we all use this method, instinctively at least, to navigate our own lives. There simply isn’t a reason why an organized and formulaic modeling of data should not be used but our intuitive subjective feelings should. If anything, where possible quite the opposite should be true. Moreover, simply because one has trouble predicting one field of human activity, like stock prices, says almost nothing about the ability to predict other human activities, like voting. Just because some fields are unpredictable, doesn’t mean all fields are.
Ultimately, while it is quite reasonable, and productive, to attack specific models on their potential flaws and biases, many pundits have taken the approach of attacking probability itself. Of course, some of these people could simply be hacks, saying whatever it takes to help their cause, but after reading screed after screed against the utility of systemically modeling polling data I wouldn’t bet on it.