• by Frank S. Foster , November 3, 2008

I received my Bachelor's degree from a state university in the Midwest many years ago. My tenure there was hardly notable, but the administration bestowed upon me a diploma and I moved on to warmer if not better things. So it was with genuine surprise when I learned that someone claiming to be affiliated with my former university wanted to speak with me.

"If this is about that honorary doctorate, I thought I made it quite clear-I will graciously accept," I began when the call was transferred.

"I'm sorry? No, no, Mr. Foster. This is about a panel discussion on political polls in battleground states."

"Really?" I asked.

"Yes. We would like you to participate in a discussion about presidential polling."

"No kidding? This is a surprise."

"I know, I spoke with your assistant and she said you hate polls," she continued. "But we would like you to consider our invitation. We need someone who is ... well, notorious, when it comes to research."

"When you say it that way, how could I possibly decline."

"I am sorry. Again." There was silence on the line.

"My dear young lady, you've got me all wrong. I don't hate polls-, " I corrected her.

"I'm glad to hear that," she interjected.

"I hate pollsters."

There was a moment of silence before she said, "So does that mean you are not interested?"

"Well, I tell you what. Throw in that doctorate thing and we'll see what we can do."

So early last Monday morning, I participated in a conference call with several noted pollsters and a number of academics with qualifications much more robust than my own. For nearly two hours they talked about various "groups" within the Florida electorate. There were the I-4 Corridor Independents, Protestant Latinos, Jews Over 65 living in South Florida, Successful Black Small Business Owners, and my favorite, College Educated Crackers. All of which were very interesting and entertaining -- which is why the television news programs are full of such analysis.

At the end I was asked if I had anything to add. I said I had only one question for each of the pollsters, "What percentage of the people approached to participate in the poll completed the questionnaire?"

In my view, the crux of the problem associated with presidential polling is panel participation, or more appropriately, nonparticipation. Unlike television research, small sample size in this environment is not nearly as problematic. But for some of the pollsters involved in our conference call last week, nearly nine out of 10 people initially chosen to participate in the poll refused to participate. Which means quite literally, if a researcher went up to the door of ten houses on a block (or in a grocery store or more likely on a telephone) nine shut her down before answering all of her questions.

How big of a problem is this? Well, that depends. Suppose those who did not participate shared a common characteristic. If that particular characteristic were to affect their choice of candidate on Nov. 4, the pollsters have a potentially significant problem. Take wealth, for example. Assume for a moment that wealthy people are much less likely to participate in polls. Suppose those same voters are much more likely to vote Republican for President because they feel the Democrat will significantly raise their taxes. Because of the non-response bias, the poll's data would likely underestimate the support for McCain in wealthy households.

"But the number of wealthy Americans is significantly smaller than the number of poor Americans," dismissed an academic. "Your bias will be overcome by Democratic turnout within the poor segment of the electorate alone." Perhaps. But the bias in our example is real and the error associated with the poll's results does not reflect that.

Bias is not limited to small samples of the electorate. Suppose a voter's enthusiasm for a candidate makes them more likely to participate in polls. Suppose Obama supporters are much more likely to be enthused about their candidate than those who support McCain. This "candidate enthusiasm" bias may cause the poll to inaccurately reflect support for both candidates. This comment about bias was greeted with silence from all parties on the conference call. I thought I had been cut off.

What did this meeting have to do with television? Put simply, almost nothing. But its effect on television means almost everything. The Wall Street Journal stated the other day that the last 139 major polls had Obama winning the election. You need only tune into NBC, CBS, ABC, CNN and MSNBC to see that the medium as a whole has run with the pollsters. Without knowing the full extent of the error and bias of the estimates, television has made it clear that Obama is the shoe-in winner. And everyone in America likes a winner.

But what if the polls were wrong? Suppose for a moment that McCain were to win. What will happen? Smarter men than I have suggested the country may be at risk for racial strife the likes of which it has never been seen. Will minorities feel wronged? Undoubtedly. Will it be the fault of television -- unequivocally. Let us hope for the sake of television as a medium that Barack Obama wins tomorrow. Whether or not television has affected who will become President, let us pray the medium avoids its "Dewey Defeats Truman" moment for another election cycle.