Politics & Policy

The Weak-End of Polling

Poor methodology is making for some poor results.

According to the conventional buzz, George W. Bush has been steadily running an 8-to-10-point lead among the major polling surveys, and therefore looks to be a big winner in November. Some estimates, such as the Battleground Survey, show the Texas governor with a 12-point lead, hinting at a possible landslide that might also translate into a Senate and House GOP sweep.

This will give the Bush Republicans the biggest mandate to govern since Calvin Coolidge in the mid 1920s.

Or will it?

Are these polls really trustworthy? Actually, are they even forecasting a Bush win? I’m an economist and not a political scientist. And while it’s true that economists forecast early and often, often we can’t keep up with what these pollsters are putting out.

For example, as I surfed the Web this morning, I found a new CNN/USA Today Gallup poll reporting that 48% of likely voters would choose Bush, and 46% would pick Gore. A week ago, Gallup told us Bush would be an easy 50-41 winner. Then I came upon a new CBS News poll that shows 43% of registered voters say they will vote for Bush, and 41% will go for Gore. So my spirits sank early in the morning.

Then I punched up the Rasmussen “Portrait of America” presidential tracking poll. They report that interviews conducted over the past three days show Bush leading Gore 43-31. What’s more, they give Bush 310 Electoral College votes for a clear victory. And, in the race for Congress, Rasmussen shows Republicans beating Democrats by a 38-34 margin. Ah, I thought. More tax cuts are coming. My Dow 50,000 forecast is still alive and well.

A couple of weeks ago the same polling confusion prevailed. In the week of June 28, the LA Times poll had Bush ahead by 10. The Battleground poll showed Bush a 12-point winner. And the CNN/Gallup survey put the Republican nominee up by 13 percentage points. In the very next week, however, Newsweek showed a much narrower 4-point lead for Bush, and an AP poll had the race dead even.

So I called one of my old polling friends, Republican Ed Goeas, who worked with me years ago in Christine Todd Whitman’s tax-cutting 1993 gubernatorial victory in New Jersey. Along with Democrat Celinda Lake, Goeas publishes the highly-regarded Battleground Survey. He told me to be careful about reading the polls. For one thing, it really matters if polls are conducted during the week or over the weekend. He told me that “political pollsters don’t poll on the weekends. They prefer Sunday night through Thursday night. Weekend results are just not reflective of where a given race really is.”

Goeas explained that more Democrats are found at home on the weekends, especially blue-collar Democrats. He added that “anyone who spends 20 to 30 minutes during the weekend talking to some pollster is not normal.” Goeas also pointed out that surveys which canvas likely voters are much more significant than those who cover registered voters. This is especially true in the current presidential race, because Republican voters are much more intense in their anti-Clinton feelings, whereas intensity among Democrats is very low for Al Gore.

Matthew Dowd, who is George Bush’s director of polling and media, agreed with Goeas. Dowd also warned that voter samples and weighting, including geographic balance, are very important in determining the accuracy of a poll. “It’s terribly important to capture the right percentage from each state and region,” said Dowd. “This includes the weightings for women, and Democratic and Republican historic voting patterns as revealed in exit polls from prior elections.”

Dowd told me that through the month of June, the average of all public polls favored Bush by 6 percentage points, compared to a lead of only one-and-a-half percent last March.

So I went back to look more carefully at the polls just printed today. Sure enough, the CNN/USA Today Gallup poll of likely voters (that’s good) was conducted July 14-16. Turning to my calendar, I discovered that — yup — it was the dreaded weekend survey. Not to be trusted.

Then I looked again at the CBS poll and — sure enough — it was another weekend survey. Even worse, the fine print stated: “Registered voters.” Another big error.

Then I turned back to the Rasmussen poll, which indicated that the interviews were conducted July 15th, 16th and 17th. So while it started on Saturday, which is bad, at least it got through the valuable Sunday night and Monday windows that Goeas says are OK. Good for Rasmussen.

Finally, I pulled out of my inbox the latest John Zogby poll, which has Bush ahead of Gore by 4 percentage points — half the 8-point lead Bush had in June. Zogby polled likely voters, but just to confuse the issue more, his survey was conducted July 14-17, which is Friday, Saturday, Sunday, and Monday. Methodologically, you could call that one the bad, good and the ugly.

Of course, as an economic forecaster, I know that at the end of the day the numbers I publish include a significant dose of my visceral computer — found deep inside my middle-aged belly instincts. So that’s how I’m viewing all these polls. Something tells me this will be a solid Republican year. Maybe not a Calvin-Coolidge-style year, but good enough to defeat the Al Gore Democrats in all three Washington houses.

That is, provided that George W. Bush works his tail off and sticks to his solid issues agenda.


The Latest