Why Ranking Polls and Pollsters Is Useless

Polls are tools, and are as valuable as predicting the outcome of a given race as the odds at the race track. The favorite does not always win.

By + More

By Peter Roff, Thomas Jefferson Street blog

Far too many people fail to understand the role that polling plays in elections.

They are a measure of public opinion among a certain segment of the population--adults, registered voters, likely voters, men, women--at a given moment in time. They are a snapshot, not a movie, and they have limited predictive value.

Blogger Nate Silver’s analysis of the performance of a cluster of polling firms in the last election, which my bloleague John Aloysius Farrell wrote about earlier today, while interesting, is of little value overall.

Ranking pollsters based on the accuracy--meaning how well their numbers in late polls matched up against the actual election returns--may satisfy our urge to see precision where none exists, but really it does not help us understand elections or the thinking that goes into how voters make their choices any better than we already do.

[See a slide show of 5 key issues in the 2010 elections.]

Some voters make up their minds late in the cycle. Some, as was the case in the 2000 presidential election, clearly change their minds as the election comes closer. Others choose not to vote for reasons of their own. And, finally, some just like to lie to pollsters, thinking that their surveys constitute an invasion of privacy and that, by being dishonest, they are somehow scoring a kind of personal victory.

The function of polls is to show what voters may be thinking at a given moment in an election and to determine if there are discernible trends that can help a campaign hone its message or perfect its strategy. They are useful in showing if a candidate’s support is rising or falling, if a particular media strategy is working, or if something not thought to be trouble is actually a major crisis. What they do not do is predict with any degree of certainty how the election is going to turn out to the percentage point.

Silver’s analysis, while excellent as far as it goes, only serves to bolster the case that polls will tell us how an election will turn out and--again as we saw in the 2000 presidential election--that is not necessarily the case. It is worth remembering that the Democrats who supported former Vice President Al Gore in that race pointed in the contest period to polls showing that Gore might have been on his way to winning Florida, as though that was somehow definitive proof that he should be awarded the state’s critical electoral votes. In fact, as every independent post-election recount showed, Bush did win Florida, the left-liberal mythology not withstanding.

It is important to remember that polls are tools, and are as valuable at predicting the outcome of a given race as the odds at the race track. Needless to say, the favorite does not always win.

  • Check out our editorial cartoons on the 2010 campaigns.
  • See who is donating to your member of Congress.
  • See a slide show of 5 key issues in the 2010 elections.