January 19, 2006

Straight ticket voting in CD22

My first thought when I saw the Chronicle poll showing DeLay at 22% was, "Wait, I bet that more than 22% voted straight ticket Republican in 2004 in CD22. If so, then the Chronicle's poll is kinda nonsense."

I was right. It turns out that 28% of the 2004 CD votes cast were straight ticket Republican votes, while 17% were straight ticket Democratic votes. So, unless 2006 is different than 2004, DeLay beats 22% based on straight ticket votes alone.**

In other words, the Chronicle's poll simply makes no sense on this issue.

Straight ticket R D
Brazoria 10793 5468
Fort Bend 38132 22806
Galveston 4963 8517
Harris 20989 9594
Total 74877 46385

Total votes R D*
Brazoria 18159 11728
Fort Bend 58444 46151
Galveston 9193 12377
Harris 64590 41778
Total 150,386 112,034

[The original version of this post had this data presented nicely in tables. However, Blogger was having some technical issues dealing with the tables, so I had to go to the ugly presentation above.]

There may be other conclusions that can be drawn from this data. If you think of any, let me know.

* In this case, the D is Richard Morrison. The R is Tom DeLay.

1. If you think that normal straight-ticket Republican voters will want to not vote for DeLay so badly that they don't vote straight ticket, then this might not be true. I frankly doubt this to be true. Voting is a habit, and it's very difficult to break people of their voting habits.
2. If presidential election cycles feature much higher straight ticket voters than off-presidential elections (or if 2004 was simply anomalously high in straight ticket GOP percentage). I don't have any data on this, but I doubt that there is much difference.

The Chronicle responds...

There's no news on this race, as far as I can find. So let's keep going with what seems to be growing as a story...

The Chronicle is continuing to stand completely behind a flawed poll that has been widely criticized by bloggers and two national pollsters. Today, National Journal's Blogometer noticed the story.

Chronicle reader rep James Campbell has posted on the controversy, but in a way that suggests that he really doesn't understand the problems and flaws of the poll (link via BlogHouston).

Meanwhile, LST publishes an email that conservative radio talker Edd Hendee sent in response to an email from Dr. Murray. I'm glad Dr. Murray is responding to Hendee, because both Dr. Stein, Dr. Murray, and Kristen Mack have ignored my emails.

I'm not sure why the Chronicle continues to stand by parts of a poll that are simply meaningless. For the Chronicle's sake, let's hope they issue a correction soon.

January 18, 2006

Chron/Stein/Murray survey, one more time

My two previous analyses of the survey (here and here) were a little rushed, so I wanted to outline more clearly the main problems with the Chron/Stein/Murray poll.

Primary problems with the Chronicle's survey:

1. They identified Steve Stockman as "a former Republican Congressman" in the general election question. This strongly overstates Stockman's general election support in a way that hurts DeLay.

2. Given such a small sample size for the primary, the margin of error for the primary results becomes very large (Benzion reports that Dr. Hill says it would be about 9%).

3. 38% of registered voters of registered voters said they were "certain" or "likely" to vote in the Republican primary. In 2004, only 0.65% of registered voters actually voted in the GOP primary in CD22. What a huge discrepancy between who votes and who was surveyed.
The second and third points combined render the reporting on the survey's primary results meaningless.

It's quite clear that the Chronicle's poll doesn't ask the people who actually vote in Republican primaries in CD22. In fact, by surveying 38% of registered voters, it's quite likely that the Chron/Stein/Murray survey is asking people who may have very different opinions than the people who actually vote in Republican primaries. When you combine that with a high margin of error, the reported primary results have no statistical significance. They're worthless.

Neither Dr. Stein or Dr. Murray has replied to my emails. Nor has the Chronicle attempted to defend its survey. They ought to do so, instead of accepting the survey on blind faith. The Chronicle, as the only major newspaper that serves this district, has a strong responsibility to get things right, especially if they are going to feature this on the Sunday paper.

The Chronicle owes its readers a correction over the primary results from the survey. Those results simply cannot be counted on to be accurate with any degree of confidence.

More on the poll

Mark Elam, who has written more than a few polls for Rep. Ron Paul, wrote in a comment to my original post:

I was polled in this survey and identified myself as conservative and Republican and was NOT asked who I would vote for in the primary...only in the general. Something is very suspicious about the whole thing. And yes, in doing a truly scientific survey, you would not identify the other candidates as "former congressman." That skews the results by giving the respondent information they might not otherwise have and by giving DeLay's opponents pretigious titles.

People talking about the Chronicle's DeLay/CD22 poll -- including on national cable

** David Benzion at Lone Star Times breaks the news that Dr. David Hill will be on "Special Report with Brit Hume" tonight on Fox News Channel at 5pm Texas time to talk about the poll. Hill's interview has already been filmed, and Benzion was present for the interview, so he previews it.

Dr. Hill has a very good national reputation as a pollster, so it will be very interesting to see how he views the piece.

** Benzion also prints an email from Chronicle opinion editor James Gibbons, in which Gibbons appears to accept on blind faith assurances that the Chronicle's poll was reasonable. Given the number of people who are providing reasonable criticisms of the poll, it seems to me that the Chronicle editors ought to look into the poll themselves.

** National pollster Chris Wilson of Wilson Research Strategies criticized the survey's methodology and resulting coverage on KSEV yesterday.

** Sedosi thinks the Chronicle relies too much on Dr. Stein as their "house expert."

** Chris Elam (whose blog has been newly christened Texas Safety Forum) has had a few posts on this poll since even before the results were released.

Have I missed anyone? If you've written on this poll, let me know.

UPDATE: It wasn't on Special Report with Brit Hume tonight. I don't know why.

January 16, 2006

Quick, rough thoughts on the subsamples made available

I'm in a bit of a hurry, so please forgive me if this isn't pretty (even for a blogger). However, here are the quick, rough thoughts I've jotted down after perusing the new crosstabs for the subsample of the Chronicle's poll:

1. They asked 560 people (the entire poll sample) whether they would vote in the GOP primary. They then screened by those people who said they were "likely" or "certain" to vote in the the Republican primary. (they didn't ask if they had voted in previous primaries, but whether they planned to...generally, this is not a very good screen, because people are much more likely to intend to vote than to actually vote)

2. 213 of the 560 sample said that they plan to vote in the Republican primary. This is 38%. 38% of REGISTERED voters sampled said they were likely to vote in a GOP primary? In 2004, 0.65% of registered voters voted in the Republican primary.

3. 50.2/39.9 was DeLay's favorable/unfavorable ratio among the 213 subset. 90.6/8.4 was President Bush's favorable/unfavorable.

4. 38.5 DeLay/ 1.4 Baig/ 5.6 Campbell/ 2.3 Fjetland was the primary numbers among the 213 subset. 52% didn't know/refused. That's high.

5. Among the 213 subsample (remember, these are likely GOP primary voters according to the poll) 39.4 DeLay/7.5 Lampson/ 14.6 Stockman/ 38.5 none/don't know/refused

6. Among the 213 subsample, when asked how they voted in 2004: 58.2 DeLay/9.9 Morrison/ 3.8 someone else/ 28.1 don't know refused.

7. 213 is a very small subset sample size. When you combine this with a relatively weak screen, it degrades the reliability of the numbers reported.
Big picture points worth further emphasis:
1. 38% of registered voters said they plan to vote in the GOP primary. In 2004, 0.65% actually voted in the GOP primary. That is a large discrepancy. This is a very weak screen.

2. Only 58% of likely voters in an upcoming GOP primary voted for DeLay against Morrison in 2004? Really?

3. The primary numbers combine a small sample size with a weak screen. This is not a recipe for a reliable poll. I continue to have a hard time taking the primary numbers seriously.
I have lots of respect for both Murray and Stein, but I think point #1 of my further emphasized points is very important.

These are just quick thoughts. Perhaps after further thought, I'll change my mind on some ponts. Since these are very quick, rough thoughts, I'm interested in further input.

DeLay vs World gets results!

[If you don't get the title, it's a joke based on something Daniel Drezner often used to write.]

The Chronicle has posted more crosstabs which Professors Murray and Stein used to calculate the primary results.

I haven't yet been able to really examine the crosstabs, but my initial note is that the sample size is only 149.

[This post will be updated tonight or tomorrow morning as time allows.]

QUICK UPDATE: The Chronicle, Stein, and Murray should be commended for making these available.

Reax to Chron poll

Kristen Mack gets reactions to the Chronicle's Sunday poll.

January 15, 2006

More on the Chronicle poll

I've put in emails to Professors Stein and Murray as well as Chronicle political writer Kristen Mack, asking them where the 39% figure came from.

It's quite clear that Stein and Murray must have provided something else to the Chronicle and other news outlets which indicated the 39%, but the 39% simply doesn't appear and isn't calculable from the data provided.