Pollsters weigh in on polling accuracy, "conspiracy" theories


by MARK WIGGINS / KVUE News and photojournalist KENNETH NULL

Bio | Email | Follow: @MarkW_KVUE


Posted on September 28, 2012 at 7:04 PM

Updated Friday, Sep 28 at 7:21 PM

AUSTIN -- New poll numbers suggesting Democratic President Barack Obama may be pulling away from Republican opponent Mitt Romney have some talking conspiracy.

"Most of them that are doing these polls, they're trying to make news, not reflect it," conservative talk radio host Rush Limbaugh told his audience earlier this week. "They're advancing an agenda. They're all Democrats, they're all liberals."

Whether they buy into the conspiracy theory or not, many conservatives including Texas Governor Rick Perry have questioned the accuracy of some polls showing a jump ahead for the president.

"Go to 'unskewedpolls.com' and take a look," Perry told KVUE Thursday. "I think they make a compelling argument that the polls are skewed."

The website run by conservative "Q Star News" claims national polls are weighted too heavily toward Democrats. The site also purports to "unskew" them by shifting the weight towards Republicans, with the result showing Mitt Romney winning in nearly every poll.

"The underlying theory that somehow all of the polling companies are colluding to support President Obama is fundamentally implausible," Texas Politics Project Director and University of Texas professor James Henson told KVUE Friday.

Co-director of the UT/Texas Tribune poll, Henson said the "unskewed" approach doesn't work in part because party affiliation is dependent on how respondents choose to identify and can change.

"We know that party identification in fact does move. It's not a fixed characteristic like your gender or your race or your age, especially," said Henson. "So when we weight polls, and everybody weights their results, you do it to try to match the population. You do that according to demographic characteristics. Sure you go and look and you check your partisan results to see if it seems plausible, but your party identification as a Democrat or as a Republican or as an Independent is not fixed."

Henson said different polling organizations weigh certain factors differently and outcomes may vary as a result but the aggregate results of multiple polls taken with different methodologies may offer the best indication of where the race is at.

Beyond methodology, the mechanics of polling are undergoing a sea change as well and Henson suggests inaccuracies in national polls may be more likely to do with polling technology reaching a critical crossroads.

"I think what we're seeing is that polling has become much more politicized in part because the actual practice of polling is in flux because of changes in phone lines, use of Internet technology, etcetera," said Henson. "That flux creates the space for people to question polling for political reasons."

"I think that there is a good case to be made based upon what we're seeing right now that some of the polls are too Democrat," said Austin-based political pollster Mike Baselice.

Baselice also expressed concerns with the "unskewed" results.

"It's an interesting exercise. The problem is it's a crude way of weighting it and if you think about it, the 'unskewed' data is being skewed."

Baselice has spent more than 20 years as a political pollster for clients including Rick Perry, David Dewhurst and Greg Abbott and said he doesn't buy the conspiracy theory either. He said that generally the polls get it right, although he does question some recent polling results.

Baselice cited the fact that Republicans voted in much higher numbers than Democrats in the 2010 mid-term elections as an example, a trend he believes will hold true again.

"We're seeing that in almost every poll again this cycle. And so to think or say that the country is more Democrat than Republican by ten points, I think is a stretch in the wrong direction."

Like Henson, Baselice said obtaining a representative sample has grown more difficult due to changes in culture and technology. He believes the true numbers lie somewhere between the "unskewed" results and the aggregated results of dozens of national polls tracked by the polling reference website RealClearPolitics.com.

When it comes to what effect poll numbers can have on a campaign or electorate in a big election, Baselice said quantifying reaction is difficult.

"It can have the effect of suppressing votes for the candidate who is not doing so well," Baselice said. "But it works the other way. If Obama's winning, then his donors and supporters say, 'I guess he doesn't need my help.'"

"There's one that matters and that's on the sixth of November," Perry said Thursday. "So I wouldn't waste too much time."

With time ticking down to the general election, it's a controversy that will soon be resolved.

News You Can Use Video
More Video