I've been using an argument a lot lately, that one way to interpret (i.e., be encouraged by) this year's polling is to look at what the polling thought of Al Gore in late-October/early-November 2000. I've been saying that right up to the election, most polls had him down 2 points or so -- and I believe a similar under-reporting of Democratic votes is happening, and will happen, in 2004.
This morning I actually looked up the numbers, presented by the essential pollingreport.com. If you look at the last days of the cycle then, Bush mostly led by more than 2 points -- 3, 4, 5, 6, 9 point leads in the last days. (Interestingly, many poll tightened up on 11/6, the eve of the election.) There's some good analysis out there on the "incumbent rule," notably at mydd.com, which predicts that undecideds always skew toward the challenger. Combine the two trends (which I'm sure are inextricably linked and more like 1+1=1.4) and a polled tie looks pretty great.Posted by dbrown at October 05, 2004 08:38 AM