Editor's letter: The shifting polls

last updated 8 January 2015

We’ve heard an awful lot about shifting polls in the last week, as President Obama’s apparently rock-solid lead dissolved after a single debate with Mitt Romney. Whether you despair or rejoice at that outcome, it reveals a basic political reality that the fancy modern science of political campaigning might have obscured. All the careful formulations on health care, all the calibrated appeals to women and Hispanic voters, all those millions of dollars’ worth of television ads designed around granular research into the American voters’ fears and desires—none of it moved the needle as much as an old-fashioned mano a mano battle.

That may unsettle political quants who thought they had this election all figured out, but it won’t have come as a surprise to Nate Silver. The founder of the blog FiveThirtyEight.com predicted the 2008 election correctly in 49 states, and was subsequently hired by The New York Times. Yet in his fascinating new book, The Signal and the Noise (see Reviews: Books), he argues that human action has proven devilishly resistant to analysis, and thus to accurate prediction. As Big Data gets bigger, we get better at predicting the course of hurricanes and flu epidemics. Yet, as Silver says, “There is no reason to conclude that the affairs of men are becoming more predictable. The opposite may be true.” His account of the monumental failures to predict the financial crisis, for instance, shows how desperately we seek data that confirms our predilections—and how easy it is to find. That’s something to remember as we all assess the analyses and predictions of “experts” in this nail-biting campaign drama. The truth won’t come until Election Day.