Pollster Ratings

by Eyck Freymann


Nate Silver, author of fivethirtyeight.com, has written up an excellent description of the pluses and minuses of most of the big pollsters. Although I disagree with his methodology slightly, I generally found it a breath of fresh air; when one looks at the polls each day one is bound to notice partisan trends (due to sampling) from the various polling agencies.

If I find time (after the election), I'll take a comprehensive look at the polls. Rather than examining, as Silver does, the individual pollsters' track records in previous elections and states, I would examine how many points to the right or left each one was to the RealClearPolitics average at the time of its release. In other words, assess a pollster's true lean by measuring its tilt in pre-election polls. This way our judgment won't be based on a single poll, one which could easily skew unusually right or left.

You will find that Rasmussen, despite a decent track record from '04, is to the right of the national average over 80% of the time. This, I suspect, is due to its methods: it uses robocalls with a slightly smaller than average sampling of minorities (who trend Democratic).