The Misuse Of Polls

A very interesting analysis of polls, polling and the misuse of the same by Charles Arlinghaus appears in the New Hampshire Union-Leader today. It is short and well worth the time to read it.

Polling about politics and public policy is incredibly misleading and counterproductive. Don't get me wrong, it's not that I think pollsters are incompetent or that their information is wrong. Today's polls are surprisingly accurate measurements of public opinion. The problem is the way they are used.

A poll is merely a quick snapshot of people's immediate reaction to a very specific question at one particular time. That piece of data is very useful if used correctly to assess baseline attitudes or initial impressions of personalities. Unfortunately, the numbers are most often used superficially to cause inaction or make judgments that aren't true.

As an example, in 1999, with the first contest more than a year away, the Republican primary looked like a tight contest between Elizabeth Dole, then leading in the polls, and George W. Bush, the only other candidate with significant support. Bush and Dole were the only two candidates anyone had heard of, so their numbers were naturally high. A nobody named John McCain was at 2 percent. But in the end Dole dropped out and McCain won the primary going away like Secretariat at the Belmont.

The poll was largely a measure of name identification before anyone started campaigning and before most candidates had a chance to become known.

The fixation with a flat numeric answer is the worst abuse, of course. But there are others including falling into a false interpretation of a poll's meaning by failing to put the snapshot into a larger context. The trend is more revealing than the snapshot of a single poll number. The analysis is written from a New Hampshire point of view, but applies just as much to national polling.