DR, like everyone else, has been following the latest polls out of Iowa with great interest. One can reasonably infer from these polls that the race is tightening, as, for example, the Zogby tracking poll indicates.
But, beyond this, the results of these polls should be treated with extreme caution.
The reason for this is simple: Any poll is only as good as its sample. In this case, the sample Zogby wants to draw, in an ideal world, is a sample of those Iowans who subsequently do show up at the Monday caucuses. Of course, they can’t do this, so they proxy those voters by drawing a sample of “likely caucus-goers” using various screening questions to do so.
How good is this sample likely to be? As a general rule of thumb, the lower the expected turnout (and in Iowa it will be perhaps one-sixth of registered Democrats and 4-5 percent of the adult population) and the more complex the voting process (like, say, the Iowa caucuses), the less reliable a sample of this kind is likely to be. You just don’t know you’re getting the “right” voters, since the voters you’re screening in may or may not be the ones who show up on caucus night.
In the end, the voters most likely to show up may be the ones who are organized into going by those with the best ground organizations (generally acknowledged to be Dean and Gephardt). That distinction won’t (and can’t) show up in the kind of polls being conducted by Zogby and others.
These problems are accentuated if turnout of caucus voters is unusually high on Monday, as some believe it might. This means the samples drawn by pollsters, which are based on historical turnout patterns, will reflect even more poorly the pool of actual voters who show up at the caucuses.
So, enjoy all the Iowa polling. But keep a big cellarful of salt handy to sprinkle over the results.