We’re coming up on primary season and the Presidential campaign seems to be never-ending this year due to the fact that there is no anointed successor to the incumbent. This leads to endless speculation about the likely winners. Of course everyone likes a horse race—especially the media, ’cause it sells—so an entire industry has built up around predicting the winner of elections. The most popular by far is personal “I think” kinds of things, but it turns out experts are often, well, not-so-expert, and Joe Average is even worse. (That won’t stop us here at AMB: We’re going to blow hard on this topic as the days go by not because we’re likely to be right, but because it’s fun….) Polls are better than experts, but it turns out that getting polls right is really, really difficult:
- Getting an accurate cross-section of the population is amazingly, amazingly hard, and in the era of caller ID, cell phones, etc., harder than ever.
- People change their minds (duh, why else would there be a campaign?) and in many cases don’t decide which way to vote until the last minute.
- Figuring out who’s going to vote is also very tough. Just figuring out turnout rates is next to impossible AFTER the fact (mostly because figuring out who’s eligible to vote is difficult). Add to that the fact that modern campaign strategy focuses on mobilizing your guys and demobilizing the other guy’s guys.
In short, polling has known issues that limit its accuracy and, in my opinion, it’s only going to get worse, not better. Telephone polls are horribly expensive and soon to be a thing of the past, much like other technologies to which era they better belong, such as punch cards, desk calculators and vacuum tubes. For those of you who are morally offended by the fact that the population is as studied as it is, this might sound like good news, but not so fast. Demise of the telephone poll only means we don’t hear about the fact that we’re being studied.
One of the innovations in political prognostication over the last several years has been the futures market. Many of us have heard of the Iowa Political Markets, which allow trading of futures to predict elections. It turns out the Iowa Market does quite well as an aggregator of information, outperforming most polls. (Of course, the market benefits enormously from published polls. Buyers need to have information to aggregate, after all, even imperfect information.) Now National Journal has gotten into the act, opening things up to a broader set of players. Right now, things aren’t looking good for the Republicans in the presidential arena, but it’s a long way out so don’t rejoice or despair, as the case may be—a year in politics is a long time.
If you want a fairly accurate popular press account of this and other things, check out Ian Ayres’ book Super Crunchers, who chronicles the “econometric revolution” quite nicely. (Highlights include the markets I just mentioned, experimental web pages used to maximize click-through, and taking John Lott’s infamous gun studies to the woodshed for continuing to rely on data with serious coding errors even after said errors were demonstrated numerous times.)