How Many Polls Does It Take to Screw Up an Election? | Letters | Chicago Reader

News & Politics » Letters

How Many Polls Does It Take to Screw Up an Election?

by

comment

To the editors:

In your most recent City Hall column [February 27], you assert that the increase in voter support for Mayor Washington seen in the second Chicago Sun-Times poll of the past campaign was the result of my arbitrarily adjusting the proportion of black voters in the survey sample upwards from 36 percent to 42 percent. This simply isn't true. Your conclusion is based on a misreading of the survey data and a misunderstanding of our election polling methods.

Both surveys were based on Chicago registered voters. Blacks made up 36 percent of the sample in the first survey and 39 percent in the second survey. This three percentage point difference is within the margin of error due to sampling, and is not the six percentage point difference you cite.

The 42 percent blacks figure applies to the second survey's likely voter base. As I stressed to your reporter, a key difference between the two surveys was that the second one did a better job of discriminating those who would actually go to the polls from those who would stay home. The closer the timing of a survey to an election, the more reliable such information becomes. The 42 percent blacks figure was derived from the internal survey data itself -- blacks expressed a greater interest in going to the polls than whites -- and was not imposed from the outside.

It has always been the policy of The Gallup Organization to open up our methodology to public scrutiny. I would tie happy to provide more information if something still isn't clear.

Larry Hugick
Senior Study Director
The Gallup Organization

Gary Rivlin replies:

Five weeks before the Byrne-Washington primary election, a Gallup poll (commissioned by the Sun-Times and WLS) showed Byrne ahead of Washington by two percentage points among likely voters. Every previous poll showed Washington ahead by 10 or 12 percentage points. What explained that turnaround? Reading the Sun-Times article, you'd have to guess that it was some sort of mighty pro-Byrne surge. Yet the answer had nothing to do with the changing views of the electorate but rather with Gallup's assumption that blacks would make up a smaller percentage of the electorate than other polling organizations assumed.

Gallup released another poll a few weeks later showing Washington ahead of Byrne by 12 percentage points among likely voters. Was it a pro-Washington surge? No, Gallup decided to hike up the proportion of blacks included in the survey (from 36 to 39 percent) and add an extra step to its polling techniques (they not only set up a category for those saying they'd likely vote, but also adjusted the racial proportions of this category), hiking up the racial makeup of the likely voter category from 36 percent to 42 percent black. This brought theirs more in line with other surveys. As any pollster can tell you, and as the primary election showed, asking people whether they plan on voting is not a very reliable measurement (white turnout was actually greater than black turnout).

The point is these surveys are based on certain assumptions that are not consistent from poll to poll. In the case of the Gallup poll, the Sun-Times and WLS never mentioned these changes, giving the false impression of a fickle electorate rather than an undecided pollster.

Add a comment