The Red Queen

I’m not a professional pollster, so perhaps I don’t know the methodology, but here’s my take.

National polling is paradoxical. Because it’s not possible to survey everyone in a national poll, a pollster can only survey a sampling of the electorate. So he must make a preliminary decision about what sample group is representative. He must make an educated guess about voter turnout. Yet that, itself, is a prejudgment about the future election.

In that respect, it’s not the poll that predicts the election, but the pollster. In other words, the polling results aren’t simply predictive of how voters will vote. Rather, they reflect back on the pollster’s anticipation of voter turnout. So there’s a circular element to the procedure. In a sense, the pollster had to get ahead of himself. Prior to the polling results is his chosen sample group. In that respect, the polling results aren’t direct evidence of how voters will vote; rather, they are evidence of how the pollster predicts they will vote, considering which sample group made the cut. In a way, he’s having to predict the answer in advance of the answer. He must have a hunch of how many voters will answer the question which way before he even poses the question. Select the respondents in light of the anticipated response.

At that level, pollsters are in the same boat as other political junkies. Their accuracy depends on having their ear to the ground.

This also raises the question of whether posters are right because they made a good guess, or because they made a lucky guess. After all, psychics and astrologers sometimes get it right, but when they do, we chalk that up to luck. (Mind you, I allow for the possibility of precognition, but right now I’m thinking of charlatans.)

I expect that when they get it right, that is sometimes due to the fact that, as political junkies, they have a good feel for the political mood of the country apart from polling. But I also suspect that in some cases, they just got lucky.

In the nature of the case, we’re more impressed by accurate predictions than inaccurate predictions. Since we know the future is unpredictable to some degree, it doesn’t surprise us when the prediction is off. So mispredictions are quickly forgotten.

Don’t these polls basically tell you that Republican respondents will probably vote Republican, Democrat respondents will probably vote Democrat, and which way independents lean? The key is getting the proportions right.