The polls for the primary and caucuses coming up at the very beginning of 2008 are all over the place. In the national sampling for the Republican nomination, for example, Scott Rasmussen shows Fred Thompson leading the pack with 25 percent of the vote, with Rudy Giuliani in second place at 23 percent. He has Romney in third at 13 percent and McCain in fourth at 10 percent. His sampling dates are Sept. 27-30.
But Gallup disagrees. They have Rudy in the lead at 30 percent with Thompson a distant second at 22 percent. They show McCain still in contention in third place with 18 percent of the vote. Gallup has Mitt Romney trailing badly with only 7 percent. Gallup's field dates were Sept. 14-16. Did the world change dramatically in the weeks between them? No way. The polls just disagree.
The difference is the screening process. It is very, very difficult to predict who will vote in Republican primaries such as those in New Hampshire and Michigan. It is even harder to tell who will participate in caucuses in Iowa. Yet the ability to predict who will vote and who won't is pivotal to an accurate reading of the likely outcome.
We won't know who is right and who is wrong until the votes are actually cast. And then it will be too late.
Rasmussen has a tighter screen than Gallup for Republican primary voters. When I asked Scott Rasmussen about the differences between his polls and those of Gallup, he answered, "I screen for likely primary voters." Gallup's screen is looser and lets in more voters.
So who is right? Neither or both, depending on how you look at it. Gallup is betting that voters have not yet focused on the primaries and have not decided whether they will vote or stay home. Rasmussen is hoping that they have decided already whether or not to vote and can accurately predict their behavior in January even though it is only October.
To make things more complicated, both Iowa and New Hampshire permit independents to vote in either party primary, making the process three-dimensional. Voters in the Republican primary not only have to decide whether to back Rudy or Thompson or McCain or Romney (or, in Iowa, Huckabee) but must also figure out if they would rather vote for one of the Republicans or go into the Democratic primary to vote against Hillary. Tough choice! Tougher prediction!
The conventional wisdom is with Gallup. Five months before a caucus, one's voter screen should be loose, since who is paying attention this early? By prematurely screening out voters who don't yet know who is running, the argument goes, one is blinding oneself to the real results, which will only be apparent once the poorly educated voters decide to participate.
But my money's on Rasmussen. Why?
Iowa is a caucus state, so a really tight screen is appropriate. If you are going to sit for three hours in a meeting in January to vote, you probably know in September that you are going to do it.
The race started nine months ago. This unusually early starting gate suggests that voters probably know a lot about the race very early on. With one-third of Americans getting their news from cable TV, all the likely voters are already players in the process and should know if they are going to vote or not.
With Hillary running, voters probably have a pretty good idea of what they are going to do. With Hillary polarizing voters as sharply as she does, voters probably already know in which primary they are planning to participate.
But anyone is entitled to his or her own opinion. And the truth is probably somewhere in between.