by Ruy Teixeira
cross-posted at the Washington Monthly's Showdown '06 blog

When I last checked in about the state of the race--about ten days ago--things looked pretty good for the Democrats. Now they look even better.

Take the generic congressional contest, for example. In the nine polls finished since 10/20 that are listed on PollingReport.com, the Democrats' average lead is 14 points. That's huge by historical standards. Democrats haven't seen these kind of leads this late in an off-year election campaign, since the elections of 1974 and 1982, when they gained 43 and 26 seats respectively.

Of course, the generic congressional contest does not tell you directly about how the myriad individual races will turn out (we'll get to the race by race data in a moment) so some caution is advised in assessing just what this gaudy lead is likely to mean for the Democrats on election day. But here's some food for thought. Three political scientists, Joseph Bafumi, Robert Erikson and Christopher Wlezien, have recently released a paper that forecasts the level of seat shifts from the generic congressional vote question, using model-based computer simulations of the 435 individual House contests.

The gruesome methodological details may be found in the paper, but the bottom line is that their simulations predict a 32 seat pickup for the Democrats. As we shall see when we get to the race by race data, this is not such a crazy prediction.

But before we get to the race by race data, let me flag a couple of more things from the national-level data. One is the probable role of independents in this election. As I remarked in my initial post on this blog:

The Democrats are also running even larger leads among independents in the generic Congressional ballot–typically 6-7 points higher than their overall lead.....

With that in mind, consider the following. As far back as I can get data (1982), the Democrats have never had a lead among independents larger than 4 points in an actual election, a level they managed to achieve in both 1986 and 1990. Indeed, since 1990, they’ve lost independents in every congressional election: by 14 points in 1994; by 4 points in 1998; and by 2 points in 2002. So, even leaving questions of relative partisan turnout aside (and I suspect the Democrats will do better, not worse, in this respect in 2006), the implications of a strong Democratic lead among independents in this year’s election, if it holds, are huge.

National polls continue to confirm a very wide lead for Democrats among independent voters. For example, the latest Washington Post/ABC News poll showed the Democrats running an amazing 28 point lead among independents, a finding that was discussed at length in the Post story on the poll. As I have continually stressed--and the mainstream press is now starting to pick up on--the Roveian fire-up-the-base-and-screw-the-middle strategy only works mathematically if losses in the political center can be minimized. Now they can't and the GOP is likely to pay the price--and very probably not just in this election.

Let me also draw your attention to a very interesting study released by the Pew Research Center that, among other things, compares a wide range of demographic groups' current voting intentions to their voting intentions at this point in the 2002 campaign. If you read one poll in detail this election cycle, let it be this one. The Pew data show huge swings toward the Democrats among many important voter groups including seniors, middle income voters, non-college educated voters, whites, rural residents, married moms, white Catholics--the list goes on and on. In effect, these shifts have turned yesterday's swing voters into Democratic groups and many of yesterday's Republican groups into swing voters.

Turning to the race by race data, polls show an ever-widening number of House seats in play. And, according to data from the Pew Research Center study and from a series of polls from Democracy Corps, Democrats are favored overall by voters in these districts. The latest Democracy Corps analsyis--"Iraq Weakens Republican Hold" is particularly noteworthy, since it shows Democrats making especially good progress in "third tier" competitive GOP seats, where GOP incumbents had the best chances of holding onto their seats. The reason, as the memo's title implies: Iraq. As more and more voters insist on seeing a change in this area, the fault lines are going deeper and deeper into GOP territory.

The scale of the possible seat shift can be assessed by looking at a number of different sites. A project of RT Strategies and Constituent Dynamics called "Majority Watch" has been polling 60 competitive House districts--55 of which are currently held by Republicans--and currently characterizes 24 of these districts as strong Democratic, 18 as leaning Democratic and 2 ties. Splitting the 2 ties, that indicates a possible Democratic gain of 38 seats (43 wins minus the five seats they already hold in the competitive 60 seats).

Over at Pollster.com, Mark Blumenthal and Charles Franklin look over all the available public polling on House races and assign 222 seats to the Democrats with 25 tossups. Let's say the Democrats and Republicans split the tossups (though in a wave election like this one, it's more plausible that these races would break toward the Democrats). That would bring the Democratic total to 234 seats--a gain of 31 seats over where they now stand.

The political scientists' forecasting model prediction of 32 seats doesn't see so far-fetched in light of these data. It's also worth noting that Charlie Cook now predicts Democratic gains of 20-35 seats (with a hedge toward a higher number than 35). Using the midpoint of his range, that would put the Democrat gain at around 28 seats--again, not far off the 32 mark.

Turning to the Senate, the latest Pollster.com 5-poll averages show the Democrats up by 5 in their only truly competive seat (NJ) and leading by 5,6,10 and 11 points, respectively in MT, RI, OH and PA. Then Webb has popped into a 3 point in VA--reversing a deficit to Allen that he had running for a number of weeks. After that, McCaskill-Talent is now dead-even in MO and Ford is down a point to Corker in TN. Therefore, it is possible that the real nail-biter in this election could be MO, if VA does come through for the Dems and Corker manages to hold on against Ford in TN.

So, things look good--very good--for the Democrats. What could turn this situation around for the GOP? At this point, it appears that the GOP is putting its faith overwhelmingly in one factor: turnout. Rev up their 72 hour turnout program they argue and--presto!--many of those Democratic margins will disappear on election day and the Democratic wave won't amount to much more than a splash.

I am highly skeptical. As I and others have been arguing for awhile, the GOP turnout machine is overrated and is simply not capable of turning defeat into victory in the manner alleged by GOP operatives and apparently believed by many in the press and even some Democrats. In this regard, I strongly recommend reading Mark Mellman's very nice deconstruction of the GOP turnout myth in today's edition of The Hill. Read the whole thing, but here's some of what he has to say:

How much difference can turnout really make? Consider the punishing arithmetic. Take a House race that this year would otherwise be 52-48 Democratic. What would turnout efforts have to achieve to overturn the putative victory?

Use white evangelical Protestants as an example. They comprised 23 percent of the national electorate in both 2000 and 2004, so let’s say they are the same proportion of our imaginary Congressional District. Say the 72-hour program was spectacularly — increasing their turnout by 20 percent while every other segment of the electorate held constant. In that case, evangelicals would constitute 26.4 percent of the electorate.

Assume for the sake of argument they continued to give the GOP the same 78 percent of their votes they gave to George Bush in 2004. Such heroic efforts would still result in a Democratic victory. And if white evangelical Protestants only offered 68 percent of their votes to Republicans, all that work would result in less than a 1-point shift in the vote. And that calculation makes the very unlikely assumption that one side enjoys great success while the other does nothing.

How likely is a 20 percent increase in turnout based on a GOTV effort? The best serious academic estimate is that all the GOTV work in the presidential campaign of 2004 increased turnout not by 20 percent, but by about 3 percent....

Can’t micro-targeting help them achieve spectacular successes? Anyone who has ever modeled data knows there is much more salesmanship than science in Republican claims about these efforts. Our firm and others on the Democratic side have been using these models for half a dozen years or more and we know they can make our efforts much more efficient; expand our GOTV and persuasion universes; and provide message guidance. So when races are otherwise marginal, the lift models provide can make all the difference between winning and losing. But no model is going to turn what would otherwise be a 5-point loss into a victory.

But didn’t the GOP prove its efforts were much more effective than the Democrats’ in 2004? No. Check the data. In Ohio’s base Democratic precincts turnout was 8.2 points higher than it had been in 2000. In base Republican precincts, turnout increased by a slightly lesser 6.1 points. Winning a state is not the same as doing a better job on turnout.

As important as turnout and GOTV efforts can be, the GOP needs to find something more to hold back this wave.

Well said. Readers would be well-advised to keep Mellman's analysis in mind as they read the press coverage leading up the election and get ready for what should be a very exciting and interesting election night.