On the lack of Republican engagement in modern “big data” analysis of politics, Robert Schlesinger writes:
Is anyone surprised that the GOP—the party of climate science denial, polling trutherism, and proud rejection of “smart people”—is uninterested in working with eggheads?
Before I get into that more, here’s some context.
So there’s this thing called probability. It’s like chance. A coin is heads, tails, or lands on edge. It’s not a “fifty-fifty”, it’s something like 49.9999999999/49.9999999999/0.0000000001. But we say fifty-fifty, and in most situations, the odds of that coin landing on end are so infinitesimal so as to be irrelevant in any coin-flip probability calculation.
Probability is not “prediction” though. And probability is not statistics, either. Statistics (as in plural statistic) are data or measurements of a thing. Statistics (as in the singular science) is the “study of the collection, organization, analysis, interpretation, and presentation of data.” Probability is “a measure of the expectation that an event will occur or a statement is true.”.
When a meteorologist—or weather forecaster—says it’ll be 65 and sunny with a 5% chance of rain, he or she didn’t consult a crystal ball, but examined both historical data (statistics) and current conditions, then ran models, or simulations—essentially, a complex break-down of a problem utilizing:
- measurements of real data,
- a good historical understanding (which is allowed to change and evolve over time), and
- a question to answer.
That simulation (or model, or experiment, if you will) gets run thousands of times, over and over, and in our example here, the meteorologist reported the most common result, which indicated no precipitation in 95% of simulations, a most-common temperature of 63-67 degrees in 90% of simulations, and perhaps a 75% chance of full sun to partly cloudy. But that’s wordy, it’s unclear, and “normal” people like a single, simple, piece of information. “65 degrees, sunny, 5% chance of precipitation.”
Taking this practice—taking metrics, stats, and unbiased queries and turning them into models, which may then be run thousands of times to gain an understanding of the many ways a scenario may result—is old. It’s not well understood, and historically practitioners of this method (“quants”, or people who believe that “immeasurables” can, in fact, be quantified) were seen as heretics within their communities. And still are. This goes for the “moneyball” stuff, also called sabermetrics, where statistical analysis and simulations are used to gain a better understanding of baseball strategy and performance. It goes for my own line of work, trying to build models to manage information risk. It goes for weathermen. And recently, it goes for Nate Silver and FiveThirtyEight, too.
Choice pull-quote from the article:
[A]s the Washington Post’s Ezra Klein argued yesterday, you can also see that quant-gut tension in some of the media commentary toward Silver. “If you had to distill the work of a political pundit down to a single question, you’d have to pick the perennial ‘who will win the election?’” Klein wrote. “During election years, that’s the question at the base of most careers in punditry, almost all cable news appearances, and most A1 news articles. Traditionally, we’ve answered that question by drawing on some combination of experience, intuition, reporting and polls. Now Silver—and Silver’s imitators and political scientists—are taking that question away from us. It would be shocking if the profession didn’t try and defend itself.”
The reason is: Silver pundits better than the pundits, he does it without bias, and perhaps most importantly, he does it intelligently without whinging on and on. (Though he does spend a lot of time discussing biases, where they can be introduced, and how he has identified biases in the data he uses—from particular polling houses, for example.) His analysis is rational, it is transparent, and it is darn accurate. But it is not a prediction in the singular. It is a statement that, say, “in 75% of simulations run, and given current polling data, President Obama will win.” That doesn’t mean he’ll win with 75% percent of the vote, and it means that 25% of the time, Romney wins. Anyway, go read this article, then go visit Silver at FiveThirtyEight and bask in the glory of his gory data analysis. And then pray that he completely makes punditry obsolete as mass media have morphed it.