John Zogby's polling is tracking 15 swing House districts, and he finds Democratic leads in 13. Since Dems need only 15 to take control - and will doubtless pick up several not on Zogby's list - it seems we're in for several years of Speaker Nancy Pelosi.

In the Senate, only Tennessee seems to be holding for the Republicans. (There's no justice: Rep. Harold Ford Jr., the Democrat now losing to Republican Bob Corker, is the best of the crop of Democratic challengers).

Other Senate races? Ohio, Pennsylvania and Rhode Island are gone for the GOP. Republican fantasies of a rebound in Montana are falling short. It's down to Missouri and Virginia: Democrats need to win both to take control. In these states, pollster Scott Rasmussen has GOP incumbents Jim Talent and George Allen below the 50 percent mark - usually a sign of doom. Rasmussen has Allen tied with Democrat Jim Webb at 49 percent and Democrat Claire McCaskill ahead of Talent 49-48.

Even with those nail-biters too close to call, 2006 will go down in history as one of the worst years for the Republicans.

Let's examine the Senate first, because it is easier to understand. It boils down to this: Democrats are trying to seize seats from states that traditionally vote Republican. Even in a bad year for Republicans, it is difficult for Democrats to overcome the triple advantages of incumbency, superior fundraising, and a population that usually leans rightward. Not only that, but the national economy — which, when it is strong, usually boosts incumbents tremendously — is arguably the strongest in the entire history of the world.

On the presidential level, Virginia and Montana almost always vote Republican, Missouri and Ohio usually do, and Tennessee does so more often than not. Pennsylvania, a state that tends to lean only a little leftward, has not elected a Democrat in a regularly scheduled Senate election since the 1970s. And in Rhode Island, a very liberal state, Democrats must overcome a habit of supporting three different generations of the (liberal Republican) Chafee family for statewide office.

Meanwhile, Democrats this year are defending several seats of their own that, except for the national anti-Republican trend, feature unique circumstances that should make them nervous.

Maryland has been trending a bit more conservative anyway, and this is the second straight election cycle where Maryland black leaders are expressing a serious dissatisfaction with the Democratic Party. With Republican nominee Michael Steele being a highly charismatic black man, Democrats have reason to fear his inroads into a normally Democratic constituency.

Washington state and Michigan, meanwhile, feature two of the least accomplished, least powerful members of the Senate — respectively, Maria Cantwell and Debbie Stabenow. Many Washington state voters are still sore that the Democrats snatched the governorship two years ago in a still-disputed election, and many Michiganders are unhappy with what is arguably the worst state economy in the whole nation, which has occurred under Democratic state elected leaders.

I’ve noted that the polls in the last few elections have almost all tended to start out favoring Democrats, and then veer Republican. I think one potential explanation is the element of self-selection in today’s polls.

We all know Internet polls, in which the respondents are self-selected, don’t really tell us much about how the population at large feels. But it’s clear to me that even professional polls have some degree of self-selection by the respondents, because of the capability to opt out. People can avoid pollsters via caller ID, or refuse to take a poll, or whatever. Some people use mobile phones as their primary phone at least partially because unsolicited calls from telemarketers and pollsters are very rare on them.

Pollsters have noted the increasing turn-down rates they’ve been getting on calls for polls. So the element of self-selection has becoming more prevalent over in recent years.

Now here’s the hypothesis. Citizens who are dissatisfied with the those currently in power are, I think, more likely to respond to pollsters (not opt out) earlier in the election cycle. Thus they might have a tendency to be over-represented in early polls.

As the election nears, those who are more complacent about the party in power become more interested in the election and less likely to opt out, and to express a more definite opinion when they do take polls. It might be that the pollsters just can’t get these folks on the phone as much as they’d like early in the election season, but they can later, and that accounts for the drift in favor of Republicans.

Now I realize that the pollsters do the best they can to balance out their samples. But the more "self-selection" (because others opt out) is going on, the harder it is to create such samples. Maybe they just have not found a way in their methodologies to correct for such a effect that varies over time

Just an idea.