Biased Polls? Election Rigging? Nah!

As certain Republicans did in 2012, Trump and his acolytes are claiming that public opinion polls showing Clinton leading by an average of 5.7 points are “biased” and part of a plot for “rigging” the election. That argument is just as silly now as it was then, when a lot of GOP time and effort was spent looking under the hood of various polls and trying to “unskew” them.

Leave aside the science that lies behind polling and think about it from a purely economic angle. Organizations that conduct polls make their money in a very competitive industry; they do that NOT by choosing a sample of voters to skew the results toward a particular result, but by being right.

Polling organizations will be held up to public ridicule if they badly miss the election result. Just google “Gallup presidential 2012.” More critically, they will lose commercial clients, as well as political ones, if they’re badly wrong in their marquee prediction: a presidential race.

But predicting the future is hard. Taking a poll today to predict what will happen in the presidential election 10 days from now requires a pollster to make educated guesses about a number of variables under conditions of uncertainty. A critical one is whether the person being surveyed is actually going to vote, which is what is meant when you hear that a given poll is reporting results only for “likely voters.”

So pollsters look at the characteristics that social science research has shown to predict whether someone is likely to vote. For example, within racial/income/gender groupings, the more education you have, the more likely you are to vote. That is, if you’re looking at white women, the more educated a woman is, the more likely she is to vote. Although voter turnout is affected by other socioeconomic variables, the same finding holds: within a racial/ethnic group, a gender group, a socioeconomic group, or a combination, the more educated the voter, the more likely he/she is to vote.

Education is just one socioeconomic variable that matters; there are many others. How a pollster weights them to determine who is a likely voter determines in part how accurately he/she predicts the election outcome.

Attitudes also matter. Enthusiasm is one of them, but so are hostility, disgust, and fear. This year, Trump has given many voters plenty of reason to experience dislike, disgust, and fear. Republicans make the same argument about Clinton, of course, and thus these attitudes may matter more this year in predicting turnout than in years when the candidates are more middle-of-the-road. The evidence for attitude’s effect on a voter’s propensity to vote, however, is not as well-understood as are socioeconomic variables, which affects the accuracy of predictions.

But pollsters are trying to get it right; they aren’t trying to bring about any particular result. So the next time you hear someone “unskewing” the polls and asserting the “real” results are more favorable to their preferred candidate or that some “fix” is in, back away slowly. They are leading you down the path that leads to crazy.

For example, google a fellow named Dean Chambers, who had a website in 2012 called “Unskewed Polls.” Chambers examined how different polls weighted predictive variables, decided that the polls were overweighting this or that, and adjusted the results to show what the “truth” was. Lots of Republicans were sure he was right and every other pollster was wrong.

Oops. Chambers’ predictions missed the election result by a mile. He later explained that he would have been totally accurate, totally, except that he did not account for what he claims was suppression of the GOP vote and massive voter fraud, for which neither he nor anyone else ever has produced a shred of evidence. Curiously, Chambers had no explanation for how the pollsters whose results he “unskewed” got it right, despite the fact that they didn’t adjust for the alleged suppression and fraud either.

But if you’re convinced that the mainline pollsters’ accurate predictions of most elections since 1948 have all been part of a grand conspiracy to award the results to whoever won a given election, you aren’t going to be moved by logic, reason, or history. If you want to cling to what’s called an unfalsifiable hypothesis, you will, all evidence to the contrary notwithstanding, and there’s no point to talking to you about it.

So. How ‘bout those Cubbies?

Leave a Reply



This site uses Akismet to reduce spam. Learn how your comment data is processed.