What is the purpose of a political poll, and how does one figure out which polls are worth watching? Or to put it another way: What’s a good poll and what’s a bad poll?
In a presidential election year — especially one in which the race appears quite close and the outcome promises to say a great deal about where our nation will be headed for the foreseeable future — these are not just casual questions.
To help, I turned to a longtime friend and political strategist. Tom Edmonds and I co-authored a couple of “by the numbers” policy books a number of years ago. Edmonds is a past president of both the American Association of Political Consultants and the International Association of Political Consultants.
This strategist was not shy in laying out the problems with many polls today. He noted that it’s “very difficult” for the average person to figure out what’s a reliable poll.
The key problem is: “Polls used to be for measuring public opinion.
Now, polls are for molding public opinion,” Edmonds declared.
He used the example of poll aggregates. That is, when various polls in a particular race are averaged together to gauge where the race might be. Edmonds noted that campaigns have figured out how to “skew” aggregates. So, before an independent, legitimate poll is released, campaigns will release their own bought-and-paid-for poll to shift the aggregate in their favor.
Edmonds advises: “Look at individual polls, not at aggregates.” He added that voters should “not trust any campaign polls.” Whether Democrats or Republicans, Edmonds said, “campaigns are not going to release anything negative.”
The issue of polling likely voters or registered voters was brought up by Edmonds as well. Other than being “entertaining on late night TV,” polls of registered voters are “useless,” Edmonds said. He added, “Everybody thinks they’re registered to vote.” He summed up, “If polls are supposed to be predictive, you need someone likely to vote.”
Edmonds emphasized that registered voter polls amount to “garbage in, garbage out,” amounting to a “legitimate measure of garbage.”
How polls are weighted matters as well, but again, it’s not so easy for the average voter to figure this out. Edmonds took to task a recent Washington Post poll on the Virginia Senate race between Democrat Tim Kaine and Republican George Allen that had Kaine up by 8 percentage points. Edmonds has a long history with and understanding of Virginia politics. He pointed out that the state does not have party registration, so it takes some work to gauge who comes out to vote. He noted that it generally breaks down as a third Republican, a third Democrat and a third independent in Virginia. But the poll in question had a very different breakdown: 24 percent Republican, 32 percent Democrat, and 35 percent independent. Edmonds sarcastically added, “Guess what? The Republican was down.”
It’s worth noting that out of the 11 polls published on that Senate race during August and September, nine showed a statistical dead heat, with each candidate polling in the mid-to-high 40s. But Kaine was leading comfortably in two — the Washington Post poll and a CBS/New York Times survey. That CBS/New York Times survey also had a breakdown of party ID skewed (24 percent Republican, 32 percent Democrat and 35 percent independent) from the norm or trend as explained by Edmonds.
Similar questions crop up in terms of national polling on the race for the White House. For example, a Quinnipiac University Polling Institute survey released on Oct. 2 put President Obama at 49 percent and Mitt Romney at 45 percent among likely voters. As expected Obama leads by a huge margin among Democrats (94 percent to 5 percent) and Romney among Republicans (91 percent to 7 percent).
Elections almost always come down to the independent vote. Quinnipiac had Romney leading among independents, 47 percent to 45 percent.
Hey, wait a minute. Then how was Obama leading? Well, it turns out that the breakdown of likely voters polled by Quinnipiac was 24 percent Republican, 30 percent Democrat, and 38 percent independent.
That’s a huge assumption given that Republicans, as Gallup has pointed out, typically have turnout advantages. Even in 2008, when Democrat enthusiasm was extremely high while Republicans were discouraged, Republicans made up 32 percent of the vote.
It’s hard to find an explanation for such a skewed voter breakdown if the purpose is to get the most accurate picture of where the electorate is.
What polls can be trusted? Edmonds pointed to pollsters that care about their reputation, citing Rasmussen Reports, for example. In 2008, Rasmussen predicted a 52 percent to 46 percent win by Obama over John McCain, with the final tally being 53 percent to 46 percent.
Edmonds bemoaned the loss of the “valuable tool” of independent, reliable polling, calling it “another lost institution.”
What should voters do? Edmonds wisely advised, “The public needs to be skeptical, and needs to vote on what’s the right thing to do whether or not they think their candidate is behind.”
Indeed, if the people get molded by polls, that speaks very ill of the people. After all, it’s the issues and the ideas that matter.
— Raymond J. Keating, chief economist, Small Business & Entrepreneurship Council.