Beware of Pollsters Bearing Gifts

The Review-Journal and Channel 8 have a new poll in the field, and it tells us … not much, and too much.

According to the R-J report, based on a SurveyUSA telephone survey of 869 registered voters statewide, Barack Obama leads Mitt Romney, 47-45. That may be true: The national numbers are close and Nevada tends to mirror those.

But the poll was automated, which tells less than those conducted by real live humans. To poll so few really doesn’t give enough of a statistical sample, especially for understanding county returns. So, with the margin of error of 3.5 percent in so small a survey, Obama or Romney may lead a questionable poll. That tells us very little.

The pollster concludes that Romney’s selection of Rep. Paul Ryan as his running mate is a “net wash,” which may or may not be true. For now it can be told: I was among the polled. And the poll asked whether the Ryan selection affected my vote. I honestly answered—well, I honestly punched the number, since the poll was automated—that it didn’t. That doesn’t tell what I really thought of the Ryan selection, does it?

Making matters worse, according to the R-J, “Those polled included 39 percent registered Democrats, 36 percent registered Republicans and 23 percent registered as independents. That roughly reflects the current Nevada electorate, although Democrats account for more than 41 percent and are expanding their advantage.”

Similarly, the poll shows Obama and Romney nearly tied among Hispanic voters. But national polls show Obama well ahead of Romney among Hispanics. If Nevada is the swing state everybody makes it out to be, this poll would be unique, to say the least.

So if they surveyed fewer Democrats by percentage than the actual registration, doesn’t that skew the result? What’s more, the R-J got rid of its previous pollster, Mason-Dixon for its 2010 performance, when it joined many pollsters in missing the boat on Sen. Harry Reid leading Sharron Angle, and more than one cynic suggested that Mason-Dixon merely produced the results R-J wanted. This isn’t the way to build up confidence.

The poll extended to other races, and the other statewide battle: Sen. Dean Heller leading Rep. Shelley Berkley by five points, 44-39 percent. Again, consider the problems already mentioned. One key result also leads to the question of which Democrats the poll missed, or where those Democrats lived.

According to the results, Heller leads Berkley in Washoe County, 53-32 percent and in the rural counties and Carson City, 64-20. That makes sense, since she’s a Las Vegan to her toenails. But the results show her leading in Las Vegas, 45-44, and the rest of Clark County, 44-37—in other words, the area she has represented in the House for seven terms and where she has won easily, even overwhelmingly, for the four of those races. This suggests the following possibilities:

• The numbers are flawed: After she won so easily for so long, it can’t be that close.

• She’s in trouble on her home turf, meaning those huge victories didn’t mean anything, which seems dubious, especially in light of other numbers.

So, those numbers may or may not be meaningful. With all due respect to pollsters, though, I’d like to mention Ralph Denton, whom I eulogized in Vegas Seven not long ago. In 1958, working on Grant Sawyer’s gubernatorial campaign, Denton took a call from a reporter working on a story about a poll showing his opponent ahead by several points. Denton paused and said, “Not according to the Oklahoma poll.” The what? He explained that a group from Oklahoma had just done a survey showing Sawyer in the lead. The reporter bought it. After he told that story, I took a survey of myself and found 100 percent support for not being 100 percent certain of what polls say.