Which Polls Count
It is very clear that the polls this election year are wildly out of sync. Consider the four polls which were based on several days of surveys ending on September 16. Pew Research had Obama up by 8%, NBC/Wall Street Journal had Obama up by 5%, Monmouth/Survey USA had Obama up by 3%, and Rasmussen had Romney up by 2%. These polls had margins of error which were significantly smaller than the differences between the polls. Some of the polls are bad science or worse.
Which polls should be trusted, and which should be treated with great suspicion? Some news organizations report self-generated poll results, like the CNN/ORC poll here, which now reports likely voters but before that reported registered voters. The polls are not conducted every day, but rather over periods which seem to be determined by the news organization. Moreover, no one is paying CNN for its polling information.
Compare CNN to Rasmussen, which sells access to its polls and reports results every day. Rasmussen asks the same question each time, it asks the question every day, it provides a rolling average (rather than a snapshot), and people fork out money to see what Scott Rasmussen has found out about America. Rasmussen's business is polling America and reporting accurate results. Although Scott Rasmussen makes little attempt to conceal his conservatism, he also reports polls which must make us -- and him -- from time to time go "Ouch!"
Gallup has similarities to Rasmussen. The Gallup Poll is a rolling average based upon polling surveys taken every day with the same questions and wording. Gallup, like Rasmussen, is in the business of polling, and it sells its services to businesses. In fact, Gallup and the GWU Battleground Poll seem at times to minimize their own findings, as with the remarkable fact (in Gallup's case) that self-described conservatives outnumber self-described liberals in every state, or almost every state, every time Gallup asks the question.
Gallup has asked for ideological self-identification four times in the last 30 months. In February 2010, Gallup reported "Ideology: Three Deep South States Most Conservative," but what the poll found is that conservatives outnumbered liberals in every single state. Six months later Gallup reported "Wyoming, Mississippi, Utah Rank as Most Conservative States," while the poll found that in every state but Rhode Island, conservatives outnumbered liberals. In February 2011, Gallup reported "Mississippi Rates as the Most Conservative U.S. State" on a poll Gallup conducted which found that conservatives outnumbered liberals in every state. One year later, when Gallup discovered that conservatives outnumbered liberals in every state but Massachusetts, the title of the story by Gallup was "Mississippi Most Conservative State, D.C. Most Liberal" (D.C., of course, is not a state).
The GWU Battleground Poll is another poll which is rigorous and accurate, but which obviously prefers that its findings cheer the left. Last November, I recounted the wording of questions in its poll, which wording leaves no doubt about what the polling organization hopes will be the answer. Yet it is this very Battleground Poll which has shown in consistently for more than ten years that an overwhelming majority of Americans are conservative.
These three polls, taken together, give Obama today a lead of 3 points, which is very close. This is very different from the other polls, which are issued by news organizations and which seem intended less to reflect opinion than to create it. Fox on September 27 had Obama up by 5 points; Bloomberg on September 26 had Obama up by 6 points the day before; the National Journal had Obama up 7 points on September 21.
The difference between Gallup and Rasmussen on one hand and the news organizations whose poll finding are their news stories on the other is that the credibility of the first two is their whole business. When people generally believe that either of those agencies is trying to create news, the whole reason for their enterprise is gone. (The same is true of the GWU Battleground Poll, which is bipartisan and prides itself on accuracy.)
In the same vein, bond rating services these days are giving state governments, our federal government, and several nations in Europe very unwelcome news which can -- or will -- push the interest rates up on sovereign debt issued by these governments. This is not because the bond rating services are conservative. Standard & Poor's, Moody's Investor Service, and Fitch Rating are business rivals, not ideological creatures.
Count seriously those polls produced by polling organizations whose bread and butter is accurate polling. In the next month, watch closely Gallup, which includes registered voters (which understates conservatives), and watch closely Rasmussen, which includes likely voters. Both those polls show a very tight race which either candidate can win.