Ignore the Iowa caucus polls

With the Iowa caucuses a little more than 10 weeks away, polls will proliferate, so I am officially sounding my quadrennial warning — beware of caucus polling. It’s always fun, but often misleading. 

Caucus polls are bedeviled by one ingredient common to all primary surveys and another that approaches the unique.

{mosads}Because most Democrats support the Democrat on the ballot and most Republicans vote for their party’s candidate, general-election polling exhibits a fairly stable structure. Absent the anchor of party, primary polls fluctuate much more dramatically, a tendency on particularly vivid display in Iowa. 

In 2008, 26 consecutive Iowa caucus polls from the end of July through just before Thanksgiving put Mitt Romney in the lead by an average of 10 points. Two of the next four polls gave Romney a 1-point lead, and the others put Mike Huckabee 3 to 5 points ahead. On caucus night, Huckabee won by 9 points.

Even that brief recitation masks considerable volatility. A Strategic Vision poll, taken Sept. 21-23, 2007, gave Romney a 13-point lead, while an American Research Group survey five days later put his margin at just a single point. Only about two weeks after that, the University of Iowa claimed a 23-point lead for the eventual loser. Dizzy yet?

Would the average of the very last polls have yielded the right winner? Yes. Would any single poll have provided real confidence? No. Would polls in September, October or November have offered useful guidance? None.

Such volatility is confined neither to the GOP, nor to 2008. Leading up to 2004, nine polls through October predicted Dick Gephardt would win the Democratic caucuses. His widest margin was in a Survey USA poll that put him 13 points ahead of Howard Dean, less than a week before a Des Moines Register poll put Dean 2 points ahead of Gephardt. The Register poll had eventual winner John Kerry with just half as many votes as Dean, while ultimate second-place finisher John Edwards tied for fourth with Joe Lieberman at 5 percent.

Zogby International’s nightly tracking would have us believe that John Kerry zoomed from third place with 15 percent to first place with 22 percent in two short days, while at the same time Howard Dean dropped 6 points. Talk about volatility!

The second great obstacle to accuracy in Iowa is polling the right people. Surveying Kansans won’t tell you much about the Iowa caucus results, and neither will polling the 2 million Iowa adults who won’t attend a caucus. Yet that is exactly whom most Iowa caucus polls survey.

Most polls that purport to provide Iowa caucus results survey the 2.3 million adult residents of Iowa, asking whether each respondent is registered to vote and then either whether he has participated in, or is likely to attend, a caucus. The tendency to say yes results in responses so highly inflated as to be worthless. If everyone who said he had attended, or would attend, a caucus actually showed up, turnout would be four or five times higher than it was at its highest.

Roughly half the participants will come from a Lilliputian group of voters who have participated in caucuses before. Campaigns lavish attention on that segment, rendering them a different species from the ordinary Iowa registrant. Public pollsters have no real idea who these people are, relying again on their inflated self-reports. 

Even so, only polling those who have attended previously is insufficient. Usually, 45 percent to 55 percent of those who attend the caucuses have never participated before. 

It’s an interesting mix of two hard-to-locate segments — grizzled veterans who are vastly underrepresented in any statewide “caucus” poll and highly motivated newbies in love with one of the candidates.

So while you breathlessly await the next set of Iowa poll results, feel free to breathe deeply and ignore them.

Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients Senate Majority Leader Harry Reid (Nev.) and House Minority Whip Steny Hoyer (Md.).

This post has been updated.