Political polls aren't perfect, but are far from meaningless

Tis the season of polls.

Tis the season of polls.

For the next 14 months, we will be barraged with political polls that will often seem more confusing than enlightening. For example, Rudy Giuliani leads the race for the Republican nomination in three August polls, but his lead over Fred Thompson is either a whopping 13 points (Gallup and Quinnipiac) or a meager three points (Rasmussen). On the Democratic side, Hillary Clinton leads the race in four August polls, but her margin over Barack Obama is either 22 (Gallup), 20 (CBS), 15 (Quinnipiac), or 14 points (Rasmussen).

Discrepancies like these will lead inevitably to the quadrennial questions about political polls' validity. We'll hear that polls this early in the race are meaningless (no, just not predictive of the outcome); that they are affected by the bias of the pollster (most pollsters care more about their reputation than the success of their favored candidate); that so few people agree to participate that the results are skewed ("non-response" is a growing problem but, so far, does not significantly skew results); that mobile phones have ruined reliable telephone interviewing (maybe someday, but not yet); and that online surveys are not scientific (complicated issues, but the best are pretty good).

When you hear attempts to dismiss political polls' accuracy, remember this fact: In the end, the best are nearly always predictive of the electoral outcome within the margin of error (usually around three percentage points). This is so more often for general-election than primary-election polling because it's easier to predict, and thus sample, turnout in a general election. There are few professions in which there is greater bottom-line accountability than political polling. If you consistently call elections wrong (or are off by more than the margin of error), people won't hire you.

How should we evaluate this barrage of polls? Based on 20 years of conducting research for candidates, corporations, and nonprofits, I follow these rules:

1. The farther out from the election, the less predictive of the outcome. Candidates wage campaigns, circumstances change, voters focus and change their minds. However, these polls are not "meaningless." They affect media coverage, fundraising, endorsements, and many other very "real" factors.

2. Dismiss polls conducted by the campaigns - for obvious reasons.

3. Focus only on reputable pollsters - you know who they are or can easily find out.

4. Average polls. The average of the poll results cited above are more likely accurate than any one of the individual polls.

5. Dismiss "outliers." If three current polls are close and one is very different, throw out the different one.

6. Pay attention to methodology. Most polls are conducted on the telephone, automated telephone, or online. Telephone is generally considered the most accurate, with automated telephone and online gaining in popularity and credibility.

7. Bookmark these links, which will do much of the work for you: www.realclearpolitics.com/; www.pollingreport.com/; www.pollster.com/polls/; www.politicalarithmetik.blogspot.com/.

Greg Schneiders is a founding partner of Prime Group, a consultancy that specializes in helping clients understand, plan, and execute change. Greg@primegroupllc.com.

Have you registered with us yet?

Register now to enjoy more articles and free email bulletins

Already registered?
Sign in