It’s technically Tuesday morning early as I write this, but I’m going to use the polls released on Monday, so this will be filed as this Monday’s projection update, as always built with generic ballot polls from Real Clear Politics.

Last week the Republicans fell off from historic gains to a result with a small majority. Let’s see if the trend continues on down or not.

As usual, to run my projection I take the recent generic ballot polls from Real Clear Politics, subtract out the undecideds and third party votes to get a pure two party vote, then compare that with the two party vote of 2008.

From there, I take a weighted average of the swing from 2008 to now in each poll, with Likely Voter polls counting double the weight of Registered Voter polls. That average is then run through the Swingometer to get a projected House composition.

We’re back to most polls being in a narrow band of 18-22, with registered voter polls from McClatchy and Marist College, Gallup, and Ipsos for Reuters showing up as the odd polls out.

Taking the weighted average I get a mean swing of 17.2. Plugging that swing into the Swingometer gives a projected Republican gain of 49 seats from 2008, giving the GOP a 227-208 House majority. The projection is still below 1994, thanks to three of the four RV polls being much kinder to the Democrats. But again, I said from the start I would work off the polls and the Swingometer, warts and all, and that is what I will continue to do.

Possibly Related Posts

• October 4, 2010 -- House Projection for October 4
• September 20, 2010 -- House Projection for September 20
• November 1, 2010 -- House Projection for November 1

There’s something to be said for consistency. And no matter who hires Greenberg Quinlan Rosner to gauge the House outlook for Democrats, bad news seems to come back.

That consistency continues with GQR’s poll of “voters,” both likely and not, for Citizen Opinion, H/T to Liberty Central.

GQR again sorts voters into two categories: likely voters (LV) and drop-off voters (DV). Among LVs, 98% voted in 2008 and 94% voted in [2006]. Among DVs, 92% voted in 2008 but only 25% voted in 2006. 99% of LVs are “probably” or “almost certain” to vote in 2010, while 56% of DVs are “50-50″ or worse to vote in 2010.

So again, we have the difference between the 2008 electorate and the 2010 electorate identified. Can we then describe the two groups? GQR does.

In partisan ID, DVs fall 53% toward the Democrats, 34% toward the Republicans, and 12% in the middle. LVs split 44 D – 48 R – 7 I. The total voter poll splits 46 D – 46 R – 7 I. Boiling that down to two-party splits, DVs swing 22 points toward the Democrats from the total electorate, while LVs swing 4 points to the Republicans. As we’ve seen in the past, Democrats are losing a chunk of their base in the transition from the 2008 Presidential election to the 2010 midterm.

GQR asked these voters if they voted for, by name, their district’s Democrat or Republican in 2008. Republicans and Democrats tied 41-41, but LVs favored Republicans 43-41, and DVs favored Democrats 47-26. LVs also favored John McCain 46-45 while DVs favored Barack Obama 61-30.

Now consider that a moment. GQR’s registered voters split evenly for the House, when the actual popular vote was 53-43. They also favored Obama by less than the actual national popular vote.

Are we at the point that dissatisfaction with Democrats runs so high, that people are lying about their 2008 votes? That’s what this poll suggests to me. And that’s an amount of doom that can’t be quantified.

Possibly Related Posts

• September 13, 2010 -- House Projection for September 13
• July 27, 2010 -- Dueling polls in New Mexico
• July 14, 2010 -- Incumbency Matters: The Joseph Cao story

Plug -2.8 into the Swingometer and you’ll see 5 districts swing. Louisiana’s second district was the fifth closest seat won by a Republican in 2008, and that Republican was Joseph Cao. He beat William Jefferson, the now-convicted felon who received bribes and kept the cash hidden in his freezer.

By election day Jefferson had already been caught, and the money had already been found. Why was the election so close? Incumbency matters.

Zoom ahead to the present. Most analysts have assumed Cao would at best have a difficult race this year, and at worst be crushed by the next Democrat on the ballot. Well, we now have a name for the most likely Democrat to face Cao in November: Cedric Richmond. So pollster Verne Kennedy of Market Research Insight took a look at the race.

That poll shows Cao almost doubling Richmond’s support, putting him ahead 51-26 (MoE 5). Even this early in the cycle, I’m amazed that a Republican in such a left-leaning district could poll over 50%, even if he’s working hard to keep his constituents happy and earning 54-9 (fifty-four to nine) favorable-unfavorable split. Incumbency matters.

All is not good news for Cao, though. Kennedy is making a specific assumption in this poll: He believes that, says Hotline, “turnout among black voters will top out at 57%” and weighted the results accordingly. So whether we believe this poll depends heavily on whether we believe that assumption.

For reference, Congressional Quarterly says that 64% of the district identifies as black. 57% would be under-representative, but not unreasonably so given the historically low turnout of that group in that region.

Possibly Related Posts

• May 3, 2010 -- Simulating the House election via Cook [Updated]
• August 24, 2010 -- Simulating the major House ratings
• August 4, 2010 -- It’s time for the newest dose of GQR Doom

Before we look at some of today’s primary races, here’s Rasmussen’s from a few days back on the Missouri Senate race.

John McCain barely won the state from Barack Obama, and apparently the Senate race is just as close.

Rasmussen shows Roy Blunt at 45 to Robin Carnahan’s 44 (MoE 4.5). It doesn’t take my fancy little modeling tool to know it’s basically a coin flip according to that, though it comes up with 54/45 chances of being ahead.

Missouri, like Pennsylvania and Ohio, looks to be one of those Senate races that could demand our attention all the way to November.

Possibly Related Posts

• October 6, 2010 -- A metric ton of new polling today
• September 9, 2010 -- Robin Carnahan is in real trouble
• August 25, 2010 -- Roy Blunt running away in Missouri

Rasmussen Reports came out with its generic ballot today, too. Having already explained in depth how I did Gallup’s, I’ll analyze the consequences of Rasmussen’s numbers in brief.

Rasmussen comes up with a 46-54 D-R two-party vote spread (47-53 for Gallup), resulting in an R+8 edge (R+6 Gallup). That gives us an 18 point swing from 2008 (16 Gallup), and a Swingometer-generated Republican gain of 51 seats (45 Gallup).

There’s little difference between the numbers, in other words.

Possibly Related Posts

• November 1, 2010 -- House Projection for November 1
• October 25, 2010 -- House Projection for October 25
• October 18, 2010 -- House Projection for October 18

The new Gallup generic ballot is out. Republicans have jumped to a 49-43 advantage, which National Review Online says is the largest Republican lead in 60 years.

Given the historical accuracy of the Gallup generic ballot in midterm elections, let’s plug this result in to the Swingometer.

In 2008, per Wikipedia, the national popular vote for the House went 53% for the Democrats and 43% for the Republicans. That translates to a 55 D-45 R two party vote when we subtract other votes. That gives the Democrats a 10 point advantage in 2008.

The current Gallup Generic Ballot reads 43 for the Democrats and 49 for the Republicans. That translates to a two party vote of 47 D-53 R, or a Republican advantage of 6.

From R-10 to R+6 adds up to a 16 point swing from 2008. If every district moved 16 points from 2008, the Swingometer projects a 45 seat Republican gain, for 223 seats in the House and the majority.

This will typically be a low end projection in a wave year, additionally, because the “national climate” influences fundraising, voter enthusiasm, retirements, and recruitment.

Possibly Related Posts

• June 1, 2010 -- Followup: The Swingometer on the Rasmussen Generic Ballot
• November 1, 2010 -- House Projection for November 1
• October 25, 2010 -- House Projection for October 25

In the North Carolina Senate race we already saw that Rasmussen Reports and Public Policy Polling are showing markedly different figures.

It looks like we’re going to see the same contrast in the Colorado Senate race, as Rasmussen showed Republicans doing well while PPP shows Democrats ahead.

Yes, it’s been a month since Rasmussen showed Republicans Jane Norton, Ken Buck, and Tom Wiens all running ahead of Democrats Andrew Romanoff and Michael Bennet. But PPP shows the mirror image of that result: both Democrats run ahead of all three Republicans.

The differences go down the line, too. Rasmussen has Colorado likely voters saying the PPACA will be bad for the country 55-37, but PPP has them opposing the plan 50-42, a 10 point swing in favor of the President.

There’s really no way I can reconcile these vastly different polls without chalking it up to assumptions which cannot be tested prior to November*. So that’s what I will do. We call can decide for ourselves who is right: PPP who appears to think 2008 represents a realignment in the American electorate, or Rasmussen who appears to dismiss the 2008 results entirely as they relate to 2010.

But for now, I’m expecting to see PPP and Rasmussen continue to disagree in any state George Bush won twice, but Barack Obama managed to swing to the Democrats.

* Yes, PPP was wrong in PA-15 and Rasmussen was right in the PA Senate Primary, but that’s not a head-to-head comparison firstly, and secondly a primary among Democrats would not be relevant to Rasmussen’s and PPP’s assumptions of the partisan makeup of the electorate. So I don’t see that relative performance as useful in favoring one of those two pollsters over the other this year.

Possibly Related Posts

• April 9, 2010 -- Colorado Senate Update
• August 11, 2010 -- Grading the Polls: It’s Buck, Bennet, and Maes
• August 10, 2010 -- PPP on the Colorado Primaries

Some researchers led by Noah Smith at Carnegie Mellon tried an experiment: Could they predict the results of traditional polling, which has as its core feature a genuine random sample of people, with careful monitoring of Twitter?

They think they got close, but I don’t think their results sound close enough to be useful. Apparently they got Twitter to have a 79% correlation with the Gallup tracking poll of Barack Obama’s Presidential approval. 79% sounds ok, but really it’s not.

Polling is about more than seeing what direction the numbers move. Magnitudes matter, and a rough correlation isn’t going to give magnitudes.

Further, as they point out, the actual uses of this are extremely limited. It turns out that during the 2008 Presidential election, mentions of both “Obama” and “McCain” on Twitter were correlated with Obama’s polling popularity. So really they have no way to judge how popular McCain was by Twitter, it seems.

It seems to me they’re not measuring opinions, but just excitement. Obama supporters were excited when he was doing well, and depressed when he was doing poorly. On any issue where youth and Internet use don’t distinguish the people on two sides of an issue, as in the 2008 Presidential election, this analysis will be absolutely useless.

You just can’t beat a random sample.

Possibly Related Posts

• August 29, 2011 -- Polling catastrophe for President Obama
• August 24, 2011 -- The oddity of Ron Paul
• June 8, 2010 -- Tension in Missouri

Building on the work I did on the House races, let’s see what the latest Cook Political Report ratings suggest for the Senate elections in November.

As before, I’m treading Toss-Ups as 50/50 shots, Lean R/D as 90% for the party, Likely R/D as 99% for the party, and Solid as 100%. Given that every single call from 2008 went for the party, and the toss ups went 60/40, the best I can do is work off of the conservative House percentages.

So I ran the usually 10,000 simulated elections using the latest Cook estimates. The result is that the early, conservative Cook estimate is a three seat Republican gain. Republicans need 10 for a majority, 9 for a tie that the Vice President would break for the Democrats.

For reference, in May 2008, Cook’s numbers had Democrats at +4, when of course they ended up +8. As with the House, Cook’s conservatism holds him back before the primaries, when we should expect a larger shift predicted.

Possibly Related Posts

• June 8, 2010 -- Tension in Missouri
• May 20, 2010 -- PPP vs Rasmussen: Colorado Edition
• May 12, 2010 -- Pennsylvania update

The Cook Political Report’s House projections are rather conservative. In 2008 only one flip was not marked competitive, and that was Louisiana’s 2nd CD, in which Joseph Cao upset William Jefferson. I naturally give him a pass on that seat.

What do Cook’s 2010 projections say, and what do we learn from them?

Comparing Cook’s 2008 ratings with the actual results, “Likely” seats fell 100% for their parties, which I’m going to call 99% for the sake of modeling. “Lean” seats fell 21/23, which I’m calling 90% for the model. “Toss ups” fell 18 Republican, 17 Democrat, which meshes with a 50% win probability for each side.

That it takes a 90% win chance to move a seat away from Toss Up to me suggests a high degree of conservatism. This conservatism served Cook well, too. As I said before, only one non-competitive race changed hands, and that was in the remarkable case of Cao defeating Jefferson in LA-2, a district with an overwhelming registration advantage for Democrats. But beyond that, in each category I detect no partisan bias.

Clearly, since I’ve calibrated my modeling of Cook’s categories with the 2008 results, my simulation of the 2008 House election using the Cook 2008 final ratings matches the actual results: D+21 mean and mode for the election.

So let’s turn to 2010. It’s early and many races still have contested primaries in progress, so I expect these numbers to change. But for now, my simulation has Cook projecting 2010 to be the mirror image of 2008: mean and mode R+20. Republicans want 40 for the majority.

That said, Cook gives Republicans a net 99 seat advantage in competitive seats, with Republicans holding 17 and Democrats holding 116. Clearly that R+99 outcome is not in the realm of realism, but it is currently Cook’s upper bound for Republicans this year. For comparison, the same upper bound for Democrats in the final 2008 tally was D+52.

I will be interested to see if Cook moves more races into the Toss Up category as the primaries conclude and we move closer to November.

Update: I’ve been asked to check how the Cook Political Report estimates looked at this point in the last cycle. So, I ran the simulation using the April 24, 2008 ratings. Result: D+6 mean, D+5 mode. These numbers can move sharply.

Possibly Related Posts

• July 14, 2010 -- Incumbency Matters: The Joseph Cao story
• August 24, 2010 -- Simulating the major House ratings
• August 4, 2010 -- It’s time for the newest dose of GQR Doom