More On the New House Polls

The twenty-seven new House polls from earlier today were, generally speaking, good news for Democrats. However, looking a little further into the poll, the numbers are even better. From the press release on the poll (emphasis in original):
Results from the first round of polling indicate the following:
  • Control of the House is largely held in the hands of Caucasian Americans. In all these swing districts, the power of white voters in these races is magnified. Based on the demographic modeling for each district surveyed, across the 30,000 interviews, 94 percent of those surveyed were white, 3 percent were Hispanic (mostly in New Mexico 's 1st District) and 3 percent were African-American.
OK, that right there is crazy. 94% of the electorate is not going to be white--not even close. The idea that minorities will only make up 6% of the electorate anywhere, much less in twenty-seven different districts, is just bullshit. According to Tech Politics, the median percentage of minority residents in these districts is 11.2% (CT-02), yet we are supposed to believe that only 6% of the votes in these districts will be minorities. If Republicans are losing the House in polls where basically everyone is white, then they are really, really up a creek. Republicans have consistently lost heavily among every minority group, especially African-Americans, who will make up a lot more than 3% of the electorate in these twenty-seven districts. More form the release:
  • The president's job approval rating is having a big effect. Overall, across all 30 districts, 38 percent of those surveyed approve and 55 percent disapprove of President Bush's job performance. This clearly may affect the Republicans' ability to hold these contested seats.

  • Both parties need to shore up swing voters. Since at the present time the two parties are holding their bases, each needs to shore up weak supporters, half of whom are not members of their party and most of whom are women.
So, a lot can change. And if Bush is at 38% approval in these districts, which tend to lean Republican, then, well, no wonder Republicans who sucked up to him in 2002 are now running away from him. Flip-floppers. Here is one more important tidbit:
The party identification of those truly undecided breaks down as follows:
  • Forty-one percent are Democrats.
  • Thirty-five percent are Republicans.
  • Twenty-four percent are independent or other.
  • The majority are male.
The Partisan breakdown among the undecided also bodes well for Democrats. Combine that with minorities being significantly underrepresented, and this poll is even better news for Democrats than previously believed. Based on what we have seen today, I feel comfortable concluding that if the election were held tomorrow, Democrats would take the House.



Display:


I think that problem has been covered before (none / 0)

I recall polls from Cleaver/Patterson in 2004 where it looked to be reasonably close, because the polling didn't get enough people in Kansas City.

As for NM-1, the problem there sorta goes back to such things as Hispanic not being a race on the Census and all that. So basically there's a good segment of people who put Hispanic/Latino origin and Caucasian.

I'm gonna go totally crazy soon and try to figure out which candidates have the strongest support (dividing strong support from their total).

So keep an eye out there.


by RBH on Wed Sep 06, 2006 at 05:02:04 PM EST

Re: I think that problem has been covered before (none / 0)

Oh yeah, the strength thing may also show which candidates are strong with their base, and screwed with everybody else.


by RBH on Wed Sep 06, 2006 at 05:06:08 PM EST
[ Parent ]

Re: I think that problem has been covered before (none / 0)

NM-1 is also complicated because Heather Wilson is, unfortunately, very good at running for office. It is going to be a knockdown, drag out battle. (Worked in the district in 2004.)


Can It Happen Here?
by janinsanfran on Wed Sep 06, 2006 at 08:27:55 PM EST
[ Parent ]

Re: More On the New House Polls (none / 0)

There is a clear link between the percentage of minority voters and the percentage of voters choosing Democrats. The National Election Survey (linked to MyDD) lists the percentage of white voters and the percentage of different minority groups.  In 2004, 75% in 2002 (a strong Republican year) and 72% in 1998 (a democratic year, of sorts).  The historically bad 1994 results were coupled with non-Hispanic whites making up an abnormally large 78% of the vote. Going back to 1948, the percentage of white voters never reached 94%.  It has been declining steadily over the years, particularly if smoothing techniques are applied.


by David Kowalski on Wed Sep 06, 2006 at 05:53:14 PM EST

Re: More On the New House Polls (none / 0)

Probably a sampling problem of small populations within the larger population. When you are sampling a population and certain variables in that population are at the tails (i.e. small numbers of people, double i.e. minorities) of the distribution, then there is a chance that you will undersample those small-number variables.

So they probably weighted the poll to reflect party id and geography and gender and what-not, but forgot (or chose not) to weight it according to race as well. Thus 22% of the voting population gets undersampled down to 6% because they were starting with smaller numbers.


by adamterando on Wed Sep 06, 2006 at 06:19:09 PM EST
[ Parent ]

Re: More On the New House Polls (none / 0)

I don't get it -- doesn't the fact that the party ID numbers tilt Dem in districts that should have more Republicans suggest that might be tilting the polls in our favour? Maybe voter ID is shifting, but that tends to shift more slowly than polls.


by thesleepthief on Wed Sep 06, 2006 at 05:55:10 PM EST

Held tomorrow (none / 0)

The real problem is that the election will not be held tomorrow.  

The Republicans have a lot of time to raise terror threat levels, foil bogus plots, divert planes for fake reasons, use Cabinet officials to campaign, announce a bullshit tax cut, make it illegal to vote while black, and of course, unleash whatever mudslinging ad campaigns they have been holding back for October.

We have a lot of work to do.  We have to build margins that won't melt away despite the dirty tricks.


by freedc on Wed Sep 06, 2006 at 06:04:39 PM EST

House Forecast Page (none / 0)

Chris, you have a typo at the top of the House forecast page:

PVI was created by the Cook Political Report and is reprting here courtesy of the Cook Political Report

Reprinted, maybe? You might also want to put a link to the Cook site.


by DavidNYC on Wed Sep 06, 2006 at 06:11:26 PM EST

Majority of males as undecided is bizarre (none / 0)

Historically it's the other way. I've looked at thousands of poll cross tabs and a tiny fraction have more male undecideds than female.

I'll take it as a very good sign, that white women who have wavered between us and them since 9/11 are fed up and booked firmly with us. However, I would be very careful to assume the undecideds will break our way, if they are majority male. There is probably a significant group who typically lean Republican but now are disillusioned and listing themselves as undecided. I know the numbers say 41-35 party ID edge but I'll trust my instincts over that. If the majority are male we can't expect much, if any, net among undecideds. You know damn well the African American males aren't undecided.

Also, regarding minorities it's critical we inspire the lower income and lower educated Hispanics to show up. That didn't happen in the last midterm in 2002. Hispanics who voted that year were significantly higher in income and education than the Hispanic electorate as a whole. Hispanics who identify themselves as independents are actually likely to vote Democratic, but they stayed home in 2002 also. The GOP gains among Hispanics since 2000 are partially Hispanic males who drifted Republican due to national security concerns, but it looks like more a case of their natural Hispanic base getting to the polls in higher percentage than ours.


by jagakid on Wed Sep 06, 2006 at 06:13:28 PM EST

Re: Majority of males as undecided is bizarre (none / 0)

if stoller's diary is accurate it seems like not all is being done to get them to the polls.


by bruh21 on Wed Sep 06, 2006 at 06:17:18 PM EST
[ Parent ]

Re: Majority of males as undecided is bizarre (none / 0)

Right you are. If those "undecideds" are white males, don't expect a good break to us.


Can It Happen Here?
by janinsanfran on Wed Sep 06, 2006 at 08:30:57 PM EST
[ Parent ]

Florida Democratic Turnout..... (none / 0)

was expected to be low.  There was really nothing for Florida Democrats to be excited about.  Our governor's race was not very sexy and had practically no paid advertising until the last week or so, and neither man is very well known state wide.  Nelson had no opposition either.  And I don't think there were any competitive House primaries on the Democratic side either.  All of that coupled with heavy rain in the three big Democratic counties in South Florida (Dade, Broward and Palm Beach) led to a very poor turnout.  The state of Florida doesn't help matters with a Primary the day after Labor Day.  You must remember Jeb Bush runs this state now.  I'm pretty sure Florida Democrats will be out in force in November to make sure Nelson keeps his seat, to keep the unctuous Charlie Crist out of the Governor's mansion and of course to win back some House seats.


by fred on Wed Sep 06, 2006 at 06:43:19 PM EST

Good News for the GOP (none / 0)

I'm struck by how few seats the Democrats actually have outside the margin of error. Moreover, Cook's polling indicates that the pick up in seats would be 13....two short of a majority in the House.

However, Chris is right to smell that something is fishy with the 94% Caucasian number. I suspect it is because (as others have mentioned) there may be no accounting for Latinos as being anything but Caucasian. It's not surprising that in some of these districts the population of blacks and Asian-Americans might be low enough to justify the 94% number.

But I'd also venture that in districts like VA-02 you probably have a lot of blacks who are barred from voting because of convictions on drug charges solely by virtue of being younger. You see, the black electorate is actually pretty old...and part of the reason it doesn't get replaced quickly is that blacks stopped having children in the 80s and 90s at the pace they did earlier in the century. But it's also clear that disenfranchisment through drug convictions has been enormously effective.

Therefore it's not that the white population is magically growing across America...it's that the electorate in real terms in shrinking, but that younger voters tend to be much more Caucasian than they were 20 years ago. Given Kerry's giant edge with young people, this seems to make even more sense: if population growth is coming from disenfrachised parts of society (either because of criminal law or immigration law) it's only natural for whites to have a near monopoly on votes in these swings districts.

This is just more evidence that America is set for a decline in population and a host of political and economic concerns heretofore never seen in the US before, but relevant nonetheless.


by risenmessiah on Wed Sep 06, 2006 at 07:12:30 PM EST

Re: Good News for the GOP (none / 0)

When it comes to pickups.

Remember that in late October 1994, not one seat held by a Democratic incumbent was listed as leaning R. And then 33 incumbents lost.


by RBH on Wed Sep 06, 2006 at 07:26:12 PM EST
[ Parent ]

Source? (none / 0)

Good point. But what's the source? I know that in 1994 the polling was no stronger for the Republicans than it is now for the Democrats.


by elrod on Thu Sep 07, 2006 at 02:09:07 AM EST
[ Parent ]

Re: Source? (none / 0)

It's a bit of a reach though to use 1994 as a comparison.

Don't forget that the Republicans have been very consistent in gerrymandering districts over the past 12 years. The other thing is that most of the socioeconomic upheaval has not been within states but between the coasts and the heartland. A few states are having significant realignments like Virginia but otherwise...it's a like Rahm says...they are picking a lock.

Oh and don't forget the "HillaryCare" fallout in 1994. By extension, Bush would have to push a policy that the Republican Congress would have to repudiate directly not simply abandon. As of yet...that hasn't happened.


by risenmessiah on Thu Sep 07, 2006 at 02:43:27 AM EST
[ Parent ]

A Closer Look At Minorities' Percentage (none / 0)

Look at the unweighted average, the minority share is even higher at 15.14%.

This breaks down as follows:

African-American: 4.08%   
Asian-American:   2.15%
American Indian:  0.59%
Hispanic:         7.09%

However, the GOP seats have almost twice as many minorities:

GOP TOT:          16.41%
African-American:  4.45
Asian-American:    2.19
American Indian:   0.65
Hispanic:          7.85

DEM TOT:           8.78
African-American:  2.24
Asian-American:    1.94
American Indian:   0.28
Hispanic:          3.28

It does indeed look like this pool understates Dem strength.


by Paul Rosenberg on Wed Sep 06, 2006 at 07:32:50 PM EST

Re: A Closer Look At Minorities' Percentage (none / 0)

The demographic info they're using matches the U.S. Census info - It's not based on the poll.


by WADem on Wed Sep 06, 2006 at 07:38:23 PM EST
[ Parent ]

Race Percent depends on district (none / 0)

If 30,000 people in their polling came out to be 94% white, it looks to me like they surveyed primarily in districts that are more white than the populace in general.

On the other hand, the Colorado Districts (CO-07 and CO-04) with which I am familiar are 31% and 21% minority. So that makes me question the 94% white fiture as well.

Where are you getting the data for your averages, from the Constituent Dynamics or from the census data for the districts.


There's more of us than there is of them.
by MetaData on Wed Sep 06, 2006 at 08:36:50 PM EST
[ Parent ]

Re: Race Percent depends on district (none / 0)

The Constituent Dynamics data IS census data. 94% white is the actual census number for the 30 districts that are in the Majority Watch project.


by WADem on Thu Sep 07, 2006 at 01:26:17 PM EST
[ Parent ]

Well, not in the two Colorado districts. (none / 0)

I looked at CO-04 & CO-07, districts I know fairly well and for which I have census data in a spread sheet.

Constituent Dynamics shows CO-04 POLLING as 6% Hispanic and 1% black, whereas the census shows the district as 17% Hispanic 1% black. CO-04 POLLING shows 8% hispanic and 5% black, while census shows 20% Hispanic and 6% black.

Are they weighting for registered or likely voters?


There's more of us than there is of them.
by MetaData on Thu Sep 07, 2006 at 03:53:10 PM EST
[ Parent ]

More digging (none / 0)

The CD methodology statement explains it:

In each of 30 contested U.S. House Congressional districts (CDs), a representative sample of about 1,000 voters were selected from voter lists provided by each state. The interviews were conducted August 27-30, 2006. Results were weighted to represent the likely electorate by demographic factors such as age, sex, race and geographic location in the CD. No more than one time in 20 should chance variations in the sample cause the results to vary by more than 3 percentage points from the answers that would be obtained if all voters in a CD were polled.

Hispnics are perhaps 50% under-represented due to demographic factors, citizenship, and because neither of the two main parties really seems to be serving their needs.


There's more of us than there is of them.
by MetaData on Thu Sep 07, 2006 at 03:57:58 PM EST
[ Parent ]

Re: Well, not in the two Colorado districts. (none / 0)

The numbers on the left column are census data of all adults in the district. The numbers on the right are census data showing the percentage of people who voted in the 2002 election - the most likely voters.


by furniture on Thu Sep 07, 2006 at 11:54:44 PM EST
[ Parent ]


You are not logged in.

In order to post a comment, you must be logged in. If you have a member account, please log in to comment.

If not, you can make an account right here. It's quick and free.