Two polls on the U.S. Senate race in Texas were released in the final weeks of September. The first, conducted by the University of Houston Center for Public Policy and Rice University School of Social Sciences, put Republican candidate John Cornyn ahead by six points against Democrat Ron Kirk. Commissioned by the Houston Chronicle and KHOU-TV Channel 11, the poll had a margin of error of plus or minus 3.3 percentage points. A week later, a California-based polling firm called Fairbank, Maslin, Maullin & Associates released a survey that had Kirk ahead in the race by four points. That poll had a margin of error of 3.6 percentage points.
Did Ron Kirk do something special in the week between the two polls to surge ahead by 10 points? No. The contradictory polling simply underscores a dirty little secret in today’s politics: Horse-race polling, especially in tight races, is all but meaningless. And now, even former adherents of this black art confirm it.
“Politicians and pollsters must admit the new shortcomings of their once highly reliable polls,” writes former political consultant Dick Morris in The Hill. Morris points out that in 2000 almost all of the national polls were wrong in their predictions that W. would garner more votes than Al Gore. Many of those same pollsters predicted a tight Senate race between Rep. Rick Lazio and Hillary Clinton for the open Senate seat in New York. (Clinton won going away, 55 to 43.) He blames inaccurate polling on a general unwillingness of the citizenry to respond to telemarketers. A clear sign of this is the growing popularity of “no-call” laws now embraced by 28 states, including Texas. In-person polling doesn’t work as a replacement because it’s too expensive and few polling firms are willing to send their workers into the kinds of neighborhoods necessary to get a representative sample. Nor will Internet polling do, because blacks and Latinos are not online in the same numbers as whites. It is also difficult to find reliable lists of voter e-mail addresses.
In his column, Morris doesn’t delve into the difficulties in accurately forecasting black and Latino turnout or first-time voter performance. This might be an issue in the Texas gubernatorial campaign, where polls have indicated a sizable lead for Republican Rick Perry. Democrat Tony Sanchez is counting on a big Latino turnout, including many who have only just been registered.
Although internal polling by campaigns can be helpful for judging trends and taking the temperature on issues, it is the newspaper horse race polls that are most problematic. Newspapers love them because they are a cheap, easy way to give the illusion that they have good political coverage. But the danger of relying on polling is that surveys, while often inaccurate, can influence public opinion. The small percentage of the American public who still believe what they read in the media might actually consider the numbers underneath the banner headlines to be true. Most newspaper polling stories don’t include lengthy descriptions of the samples or the wording of the questions asked, which might, if properly explained, give the reader a better sense of how reliable they are. Poll results are like any statistical exercise: Strip away the veneer of objectivity, and you’ll find a “science” that can be molded to any agenda.
Nowhere is this more alarming than in the debate over whether the U.S. should invade Iraq. Polling has consistently shown that Americans favor an invasion, while anecdotal reporting indicates that there is widespread unease over the prospect. When poll questions mention possible U.S. casualties, support plummets. What results would we see if the pollsters mentioned widespread unease among our allies? Or raised the prospect of Iraqi children killed by U.S. bombs? Reporters and editors should get back into the business of bringing us the information we need to make informed decisions, instead of telling us what we already think. –JB