Thomas D. Elias: Gripers over polls usually wrong
Every election year at this time, radio and television airwaves, newspaper columns and political websites are suffused with poll results. Some track voter preferences daily, like www.realclearpolitics.com, which carries a daily compendium of polls on presidential and other significant contests at the national and state levels.
But are these polls accurate? As a rule, whichever party the polls report as trailing complains about things like skewed samples, political push polling where questions asked by pollsters lead respondents into preferred answers, and the fact that polls usually can't reflect the very latest events.
And yet … most published polls are devastatingly accurate, especially when they account for likely voter prejudices and the latest in early-voting availability, whether absentee or electronic.
For most of the last two months, national polls and surveys in so-called battleground states like Florida, Ohio and Colorado have showed President Barack Obama with small to moderate poll margins over Republican challenger Mitt Romney. They also predict there will be no real contest in California, where both Obama and Democratic Sen. Dianne Feinstein are projected as easy winners.
This does not sit well with GOP activists, who know that everything from free television coverage to campaign donations and voter turnout can hinge on what the polls say.
So the tea party-linked Campaign to Defeat Barack Obama began bleating about poll bias six weeks before Election Day, just about the time the first early voters were casting ballots.
The public polls include too many Democrats in their samples, the group argues in email after email to its adherents. There is also the argument that many public polls survey among registered voters rather than winnowing their samples to include only likely voters.
Using registered voters, both the daily Rasmussen Reports and Gallup tracking polls have shown Obama with a slight national lead for the last six weeks and larger leads in some swing states, most notably Ohio and Wisconsin, where his edge in the surveys has ranged as high as 12 percent.
Neither the current Republican gripes nor similar ones by Democrats during the George W. Bush election years explain why public polls, whose very survival as businesses depends on accuracy, would deliberately skew their samples to favor one party or the other. If any public poll develops a reputation for inaccuracy, push-polling or favoritism, that poll will quickly disappear.
Says Jay Leve, director of the New Jersey-based Survey USA firm whose polls are paid for and used by many California television outlets, "(The complaints) are an assault on the integrity of an entire profession and I find it repugnant, as should any thinking person."
As a rule, private polls run by campaigns provide somewhat more detailed information than public ones, but plenty of detail can still be gleaned from analyzing the results of public surveys like Reuters-Ipsos (www.ipsos-na.com/news-polls/pressrelease.aspx?id=5764), which polls for several newspaper and television outlets.
One Reuters-Ipsos poll taken Sept. 7-10, for example, found Obama ahead nationally by 48-45 percent among registered voters. It used a sample including 38 percent who leaned Republican and 42 percent leaning Democratic, which Republican complainers instantly said was responsible for the poll's result. But the sample also showed 51 percent of those responding identified with at least some of the tea party's conservative message. And other national surveys indicate slightly more voters identify themselves as Democratic, hence slightly more Democratic leaners in the sample.
Like its competitors, Reuters-Ipsos risked its reputation on its sampling choices. And most similar polls have been very accurate in the modern era.
The surveys learned a lot from the 1982 election of Republican George Deukmejian over Democrat Tom Bradley as California governor. Polls published the day before that vote showed Bradley the likely winner and exit polls on Election Day also had him ahead.
But Deukmejian won. One reason was that large-scale voting by mail was then a new thing and the polls had not taken it into account. They never made that mistake again, and came within 1 percent of predicting the actual presidential outcome in 2008.
Polls were also inaccurate in two California primary election races in 2006, failing to account for both the number of individuals who have cellphones but no land lines and not considering the impact of last-minute email blasts by the campaign of current-Secretary of State Debra Bowen.
There have been problems with polls using only robotic telephone calls, too, but those also have mostly been repaired.
The bottom line this year: Despite complaints from those who don't like their findings, chances are the public polls have been accurate so far and will be pretty reliable right up to Election Day, unless one or more campaigns have introduced some new wrinkle that no pollster knows about.