We’ve read a lot about how Americans’ confidence in the national media has curdled during the current campaign.
Watching and reading the predominantly poll-driven coverage this year has been like making your dinner out of a jumbo bag of cheese curls. News consumers, who currently have a huge appetite for political reporting, have been left feeling bloated but unsatisfied, knowing that they’ve gorged themselves on something unwholesome.
There’s little doubt that the blathery and biased coverage of the 2008 election will help hasten the undoing of traditional newspapers and broadcast television. After seeing the new circulation numbers and ratings, it seems that there may not be much thread left in those worn old cloaks, anyway.
But what about the pollsters on whom the panting pack of political reporters have based their coverage?
The kind of solidity once seen in a few careful polls, like Mr. Gallup’s, was akin to the confidence investors had in blue chip stocks like the
Pennsylvania Railroad and GM. And just as the blue chips have become junk, the handful of authoritative polls have been replaced with a crowded market of shoddy commodities.
Was Barack Obama winning by 15 points as the Pew Center told us on Tuesday? No. Was the race tied as the Associated Press told us last week? Hardly.
We know that there are always outliers, but some of these polls have been so bad that they feel more like out-and-out liers.
To be fair, dishonesty is not the issue. The real problem, especially with academic and media polls, is that the epochal nature of this year’s election has allowed pollsters to engage in magical thinking.
We don’t really know what turnout will be like in this strangest of all modern elections, so pollsters are free to imagine that it will be whatever they hope for. And at most media outlets and colleges that means that all past truths will be washed away by the rising tide of Obama.
Think of a poll as a machine. Before turning it on, the operator has myriad knobs and dials to set. If he thinks that women will vote in much larger numbers than four years earlier, he turns the dial until a sample of 1,000 voters contains 600 women.
Women vote for Democrats more than men, so by turning the dial he has changed the results of the poll.
Coca-Cola can take weeks and hundreds of thousands of dollars to see how Coke Zero is being received because opinions about soft drinks don’t whipsaw on the results of a debate or a market meltdown. Pollsters have to do their work in three days and on a budget.
That’s why the dials are necessary. A good pollster decides based on history and his gut what the electorate will look like, weights the sample accordingly, and turns on the machine. A bad pollster engages in magical thinking and sets the knobs wherever they feel right.
Consider the silly poll that Newsweek put out this week that showed Obama with a 12-point advantage. The magazine doesn’t give us much detail into the sample used for the poll, but in what was released we can see some obvious problems.
When pressed on party affiliation, their respondents were 52 percent Democratic and 40 percent Republican. As far as historical voting behavior in the Newsweek sample, there were 471 John Kerry voters and 447 George W. Bush voters.
So Newsweek was right in the sense that if Democrats flood the polls and Republicans stay home, Obama would crush McCain.
But serious pollsters know that the electorate may have changed slightly since 2004, but that massive turnout will still reflect past performance. It will be the same, just much more of it. Consider what’s going on with early voters in swing state Nevada.
According to the Las Vegas Review Journal, a quarter of the state’s electorate had already voted by Sunday. But of those voters, just 20 percent were Hispanic, 14 percent were under 30, and 15 percent didn’t vote in the last three elections.
Early voters this year look pretty much like what we’ve seen in Nevada before. If that’s the case, polls based on a huge shift in the electorate that show Obama with a doughty lead would come up a cropper.
Nationally, many pollsters have discounted the possibility that everyone is fired up about this election, not just black Democrats and college students. That may explain much of the erratic nature of the results.
McCain’s top pollster, Bill McIntruff, rightly predicts that voter turnout nationally this year will be at least 130 million voters compared with the 122 million in 2004.
The only problem is that we don’t know which 130 million.
Chris Stirewalt is the political editor of The Washington Examiner.
