In a story about how George Bush, not John Kerry, got a “bounce” from coverage of the Democratic National convention, Susan Page disclosed something significant about a poll done on the matter. She revealed, “Because the results were a surprise, USA TODAY extended the survey an additional night, to Sunday, to create a larger and more reliable sample.” In other words, because the media feared the results would undermine Kerry, they did their best to find more people for their “survey” who would validate the media’s assumption that Kerry would get the bounce. But it was not to be. They couldn’t find those Kerry supporters.
Despite the blatant manipulation of the polling data, Bush came out the winner. But the admission by Page demonstrates how these polls could possibly be rigged to produce the desired result. It shows that when there is a “surprise” in the polling data, the pollsters always have the option of extending the survey by polling more people over a longer period of time to produce different results.
The CNN/USA Today/Gallup poll showed the Kerry/Edwards ticket trailing Bush/Cheney 50 to 46 among likely voters, with independent candidate Ralph Nader at 2 percent. “Before the convention,” USA Today reported, “the two were essentially tied, with Kerry at 47, Bush at 46.” In her original story about the poll, Susan Page did not disclose that the survey was extended because the results were a “surprise” to those expecting a bounce for Kerry. Instead, she said that, “USA Today extended its survey to Sunday to get a fuller picture of what’s happening.” By using the phrase, “a fuller picture of what’s happening,” Page made the exercise seem scientific.
On July 12, CNN announced the result of a CNN/USA Today/Gallup poll finding Kerry beating Bush 50-45 percent with Nader getting 2 percent. But the fine print showed that it was based on contacting only 706 people considered likely voters. An observer of media coverage of polls, Joachim F. Van Ells, wrote to us to ask, “How many of the 140 million potential voters in this country do you think know that these percentages, reported as the trend in the American public, are based on such small samples? In fact, if you check with the major polling organizations you will find that most of the polls conducted have sample sizes of between 1,000 and 2,000 participants.”
In a letter to the Federal Communications Commission citing the poll, he said that, “If this was a commercial product, I am sure your agency might take a close look for false advertising and you would most likely require a disclaimer. It seems clear to me that the news media is well aware of the impact of using percentages as opposed to actual numbers when reporting results. I strongly believe that your agency should investigate and take corrective action.”
He believes the media should be required to prominently report actual numbers or at least the total sample size, in addition to the percentages. That will lessen the impact of the results, he says, but he told the FCC “it is your job to protect [people] from any conscious attempts to mislead them.”