Can We Trust the Polls?

Democratic consultant Mark Mellman argues that reporters should emphasize the uncertainty of public opinion polling and “Just say the polls aren’t clear.”

It is abundantly clear that polls on the presidential contest conducted by respected survey-research firms are all over the place. Within a few days recently, two of the oldest names in the business had dramatically different results that would lead to conclusions about the race that are diametrically opposed. Harris said John Kerry was ahead by one point, while Gallup found Kerry trailing by eight. At about the same time, Pew had President Bush ahead by one while CBS pegged his lead at eight points. Truth is, that has happened during many weeks of this campaign. Back in March, Gallup had Kerry ahead by six points. Less than a week later, a CBS/New York Times poll reported Bush ahead by eight. On almost exactly the same days in June, the AP reported definitively that Bush was ahead by one point while the Los Angeles Times was just as sure that it was Kerry who led by seven points.

The reasons for these divergences are fascinating grist for future columns. But, clearly, not all of these results can be definitively true at the same time. The conclusion is inescapable: at minimum, the polls produce an uncertain portrait of the state of the race.


In each case, though, the story on the state of the race is based almost exclusively on that outlet’s own poll. Journalists with whom I have discussed the problem lean toward one response: “We trust the poll we paid for.†This is understandable as an economic criteria, but not as a journalistic one.

Mellman is correct in arguing that media outlets are guilty of touting their own polls at the expense of the big picture. It does not follow, however, that the real story is that we have no idea of where the race stands at any given moment. The average of all the reputable polls puts President Bush up somewhere in the vicinity of 5 percent. That seems to be a pretty good indication of where the race stands, as evidenced by the reactions of the two campaigns. Noting that Bush is up at the moment is not at all the same as saying that Bush will definitely win in November; too much can happen between now and then.

Republican pollster David Hill argues that, while not perfect, we have developed pretty good techniques for screening “likely voters” from a survey. He contends that pretending otherwise is not only foolish but nothing more than political correctness.

[T]here are good ways of determining likelihood of voting that are totally based on demographic profiling. Simply knowing the region, age, gender, marital status, work status, length of residence, race and ethnicity of voters allows powerful prediction of likelihood of voting. Any sophisticated political observer who doubts the value of “profiling†to predict likelihood of voting should review copies of the Census Bureau’s biennial reports on voting and registration. This week I reviewed “Voting and Registration in the Election of November 2000.†Even though this is a federal publication, the political-correctness police haven’t discovered it, so it tells the truth. Many racial or ethnic groups vote less than do non-Hispanic whites. For example, it reports that among all voting-age Hispanics, just 28 percent voted in 2000. Even among Hispanic “citizens†of voting age, just 45 percent voted. And among Hispanic citizens who are registered to vote, only 79 percent voted. Similar patterns were noted among Asians and Pacific Islanders.

These data are not reported to denigrate these ethnic or racial minorities but simply to make the objective and data-based point that these groups vote less than do non-Hispanic whites or blacks. Racial profiling may be wrong for policing, but it is right for pre-election polling.

The census study also reveals that women, older people and married people are more likely to vote. The gender factor is particularly interesting. Until 1984, men were more likely to vote than women. But as more women won degrees and jobs, the polls were reversed. More education is a powerful factor in voting, even among the better-educated. For example, 75 percent of bachelor’s degree holders voted, but 82 percent of those with an advanced degree went to the polls in 2000. Age is a demographic responsible for huge differences in registration and voting. Only 51 percent of citizens age 18-24 were registered to vote in 2000, and just 36 percent of them voted. By comparison, among people age 65-74, 79 percent were registered and 72 percent voted. So the ultra-liberal “Rock the Vote†crowd is pretty much wasting its time, thank goodness. They’d be better off getting Chubby Checker to do a twist contest at the condos in Fort Lauderdale.

One of my favorite data points involves couples. Among married people whose spouse is present, 68 percent voted. But among married people whose spouse is absent, on account of illness or death, just 52 percent voted. Just 53 percent of the divorced voted. Among the separated, only 46 percent voted. And merely 44 percent of those who had never married cast a ballot in 2000. Voting is surely a social act. When we have someone to discuss it with over the breakfast table, we̢۪re more likely to go to the polls.

While I can understand how some sensibilities may be offended by these data, it̢۪s just a fact of life that some citizens vote and some don̢۪t. We̢۪re not all equally likely to vote. And the sooner pollsters factor that into their equations, the sooner we̢۪ll get to the truth.

Most of this is so well established that it’s taught in introductory level political science classes.

Predicting human behavior is a dicey enterprise, simply because people are so incredibly complex. With respect to voting, there are a number of variables that affect turnout even in addition to the demographic ones noted above: everything from the weather on election day to the news reports of the public opinion polls themselves can have an impact. In a particularly close election, where the outcome of the race may fall within the margin of error of the polls, these things can be decisive. But the fact that we can’t issue predictions with one hundred percent certainty is hardly a reason for pretending we have no clue at all.

FILED UNDER: Campaign 2004, Political Theory, Public Opinion Polls, , , , , , , , , ,
James Joyner
About James Joyner
James Joyner is Professor and Department Head of Security Studies at Marine Corps University's Command and Staff College and a nonresident senior fellow at the Scowcroft Center for Strategy and Security at the Atlantic Council. He's a former Army officer and Desert Storm vet. Views expressed here are his own. Follow James on Twitter @DrJJoyner.


  1. Bithead says:

    funny though, ow we didn’t hear this from the Democrat pollsters when their guy wasn’t losing, both now, and in the last three elections….

    Could this caution from them now, be self-serving?