Perverse Incentives in Campaign Polling

In many lines of endeavor, competition between competing firms is beneficial to the customers. Polling on US presidential campaigns tends to be something of an exception. A bunch of different firms are very capable of doing “good enough” opinion surveys, at which point the best way to further improve accuracy would be to aggregate across different polls which media outlets are generally reluctant to do. Instead from a business perspective what you want to do is stand out from the crowd by innovating. Except when you innovate in a space where the state of the art is already good enough, you tend to end up with a lower-quality product.

Thus, USA Today’s aggregate poll of 12 battleground states.

There are two kinds of polling on the presidential election that make conceptual sense. One is nationwide polling. Broadly speaking, the winner of the popular vote almost always wins the electoral college and doing a nationwide poll is much logistically simpler than state-by-state polling. Alternatively but much more laboriously, you can go around and poll particular states that you’re interested in. The close correlation between the popular vote and the electoral college isn’t a law of nature, so state-by-state polling can add value if you’re willing to expend the resources. But doing an aggregate poll of Colorado, Florida, Iowa, Michigan, Nevada, New Hampshire, New Mexico, North Carolina, Ohio, Pennsylvania, Virginia, and Wisconsin doesn’t do anything to improve on nationwide polling. If you try to imagine an election where the popular vote is close enough for the electoral college breakdown to matter, then you’re already imagining an election in which Obama loses North Carolina and wins Wisconsin. If you want to know whether Romney might be slightly up in Ohio or Virginia despite being slightly down nationally, you have to do state-by-state polling of the states you’re interested in.

But doing the same thing as everyone else is boring and doesn’t necessarily get you clicks. To get page views you want to do something new, and if doing the right thing is old hat that means you need to try to do the wrong thing.