The Race Tightens

Interpreting Polls Legacy blog posts The 2004 Race

As long as I’m back on the topic of debates, it is worth remembering that media coverage of the debate typically has more impact than the immediate reaction of debate viewers. In the ten days since the first debate on September 30 almost every major public poll has released a new survey, and we are now in a position to see if the voters’ preferences have changed.

Although some see conflicting trends, it is now clear that the race has narrowed since first debate. The RealClearPolitics weekly average of the three-way vote (including Nader) had George Bush running ahead of John Kerry by 6-7 points (roughly 49% to 42%) now has Bush ahead by less than two points (47.5% to 45.8%). Of course, RCP simply averages whatever polls were reported in a given week, and their approach is often less than apples-to-apples.

To create more comparable averages, I used results of polls from eight organizations that were in the field just before and after the debates (once between September 17 and 28 and again between October 1 and 7). The organizations were ABC/Washington Post, AP/IPSOS, Gallup/CNN/USA Today, Fox/Opinion Dynamics, ICR, The Marist Institute, Time and Zogby. All eight reported results for “likely voters;” all but Fox and Zogby also reported results for all registered voters.

  • Among “likely voters,” the Bush lead narrowed from an average of six percentage points (50% Bush, 44% Kerry, 2% Nader) to just two (48% Bush, 46% Kerry, 2% Nader).
  • Among all registered voters, the change was nearly the same: the Bush lead narrowed from seven points (50% Bush, 43% Kerry, 3% Nader) to just two (48% Bush, 46% Kerry, 2% Nader).

Two other thoughts: First, remember the incumbent rule. If the election were held today, Kerry would likely get most of the average 4% that is still undecided, resulting a photo finish.

Second, the shift over the last two weeks has been a modest 2% net change for each candidate. It may be that only 4-5% of the voters have shifted allegiance, yet the race is close enough that such small changes have great potential consequence. Keep in mind, if they could shift Kerry’s way, they could also move back.

———-
Note: The last time I averaged poll results, several readers questioned the wisdom of averaging. From the perspective of pure sampling theory, they have a point. The different polls have different sample sizes (likely voters ranged from n=630 to 1,036), ask slightly different horse race questions, use different likely voter models and different weighting schemes. If we were designing a survey from the ground up and wanted to apply precise statistical tests to the resulting data, averaging is a bad idea.

Having said that I do not agree that averaging in this context attempts to “commensurate what is effectively incommensurable,” as one reader put it. I selected polls that were in the field just before and after the first debate, so that each pair of surveys is methodologically comparable. I also averaged results for likely voters and registered voters separately. While different weighting, sampling and likely voter models can produce different results, there is some benefit to pooling the combined “art” of eight different pollsters.

Mark Blumenthal

Mark Blumenthal is political pollster with deep and varied experience across survey research, campaigns, and media. The original "Mystery Pollster" and co-creator of Pollster.com, he explains complex concepts to a multitude of audiences and how data informs politics and decision-making. A researcher and consultant who crafts effective questions and identifies innovative solutions to deliver results. An award winning political journalist who brings insights and crafts compelling narratives from chaotic data.