Readers can click through to Smith’s post and the embedded links for more on the debate about whether poll results should be weighted for party ID. But there’s another important point here: even among surveys that take a similar approach to this question, like The Washington Post/ABC News poll, Quinnipiac’s findings are an outlier. As Mark Blumenthal of HuffPost Pollster wrote Wednesday:
In Florida, a state widely considered a must-win for Romney, results ranged from a whopping 9 point Obama lead (53 to 43 percent) on the new CBS/Times/Quinnipiac poll, to a 4 point Obama advantage on the Washington Post poll (51 to 47 percent), to a much narrower 1 point edge for Obama (48 to 47 percent) on a Tampa Bay Times/Miami Herald/Mason Dixon poll.
Most of the other polls have been closer to the Post and Mason-Dixon results
Blumenthal’s tracking model, which incorporates data from all public polls, puts Obama up by nearly five points in Florida, while the Real Clear Politics polling average puts the president’s lead at about three points. Both are meaningful advantages, but well below the wide margin of the Quinnipiac poll.
As Blumenthal notes, decisions about how to weight samples and select for likely voters are as much art as science. But even setting aside methodological disputes, polls will occasionally produce outliers, so news accounts should always take a broader look at the data. On this score, the Sentinel post actually did better than most—after that lede that emphasized the poor numbers for Romney, it noted that other recent polls had found a closer race. The News Service of Florida article, by contrast, delves into the Q poll’s results for different subgroups but never pauses to look at other recent surveys.
Meawhile, The New York Times itself used the poll as to add drama to its portrait of the Romney campaign on the brink of defeat in the final weeks of the campaign—without pointing readers to other data that would have offered a less stark picture. (The Times provides its customary note on methodology but does not appear to have addressed the recent critiques of the Quinnipiac poll.)
I have long been a critic of polls and news reporting of those polls. In January, I took The Miami Herald and Tampa Bay Times to task for questionable reporting of a poll those papers sponsored. As I said at the time: “News organizations are getting increasingly sloppy with reporting on—and addressing the shortcomings of—their own polls, not to mention asking tough questions of all the other polls that seem to pop up every day.”
Unfortunately, this complaint still holds true. The unpleasant fact is that news organizations have always seen polls as a marketing tool. The scores of polls being done by news organizations of all sizes are more about name recognition for the organization than about informing voters. Meanwhile, the steady stream of polls has become a reliable source of fodder for content-hungry politics blogs around the country—even if, as in the case of the Tampa Bay Times here, a second post has to go up hours later debunking the first post.
It is long past due for news organizations to reconsider the worth of these polls (especially at the presidential level, where we have more data than we need); do a better, more transparent job of explaining them and responding to criticism; and think twice before publishing a poll that they have not been directly involved in producing. What is happening now is a disservice to voters.