I’m not a poll wonk, but I do think we tend to judge pollsters by their last showing. The truth is, though, unless an election is a blowout, there’s enough wiggle room in the margins of error for most pollsters to come within a few points of being right in a close race, even if they pick the loser. One of them, by sheer luck, will be more right than the others.
If pollster X says Smith will beat Jones by 2 points in his final poll, and pollster Y says Jones will beat Smith by one point in his final poll, both would likely be within the margin of error if Smith did indeed beat Jones by 2. But we’d remember that X was exactly right, and that Y picked the loser four years later at the next election. But it may just have been luck that X was right. When dealing with a close election, where most of the possible outcomes are off the table (i.e., someone winning 63-37 or 58-42), there are only a few possible point spreads and someone’s bound to hit it just about right just by coincidence. The other pollsters, who missed it by a point or two, look worse by comparison, especially if they predicted the loser to win 51-49 and instead the other guy won 51-49.
I think Rasmussen got close to being right in 2004. But that probably doesn’t mean he’s better, just that he lucked out. Someone else may luck out this time.
Probably so.
But I still like Rasmussen and Battleground better than the rest mostly because they don't jerk us up and down as fast as most of the polls. Part of that is methodology I'm sure but some of it is, I'm hoping, a reluctance to play politics with their data which is common in many of the polls that we read daily.