Election Polls Push Right

A recent Gallup poll showing George W. Bush with a 13 percent lead over John Kerry has spooked many people who fear another four years of Bush. A closer inspection of that poll reveals something more disturbing: deliberately manipulative polling techniques intended to create the false impression of a commanding lead for Bush. Similar methods were used by United States polling firms in Venezuela during the lead-up to the recall vote on Hugo Chavez. Ultimately Chavez won that vote decisively in spite of polls that showed him losing. Gallup is using some of the same tricks.

When Gallup calls households they always ask to speak to the youngest male of the household first. This produces a disproportionately male, and therefore disproportionately Republican sample.

Like most poll firms, Gallup deliberately tilts the people they poll based on their party identification in order to produce results that reflect the views of “likely voters.” Unlike most other firms, though, Gallup’s figures are based on the assumption that 40 percent of voters will be Republicans and 33 percent will be Democrats, even though all the data (including the past several elections) indicates that the reverse is much more likely to be the case.

Gallup has a reputation as a venerable polling firm, but their methods have become less and less reliable since they were sold by the Gallup family. The Gallup family jealously guarded the appearance of neutrality. By contrast, the present CEO of Gallup, James Clifton, contributes thousands of dollars to Republican candidates.

In the 2000 elections, Gallup showed Bush with a double-digit lead right before the election, when all the other major polls showed a neck and neck race. As we all now know, the Gallup polls in 2000 were dead wrong.

Political polling has been made doubly unreliable this year by two new factors. The first is that pollsters only call landlines while increasing numbers of people only have or only answer their cell phones. Anger at Bush has energized communities with traditionally low voter turnouts, producing a nationwide surge in new voter registrations. It is difficult for polling organizations to know how to accurately take these factors into account, so mainly they don’t.