Monday, January 30, 2006

Poll Perspectives

[ . . ]

Readers may be aware that I was fairly focused on national polling during the 2004 Federal Election campaigns, but have since dropped my scrutiny for the most part. There are a number of reasons for this, not the least being the very different methodologies applied by polling agencies during the ‘off-season’, so to speak. For instance, polling agencies query “likely voters” during the election campaign, and pay attention to respondents who claim to have voted in the last election. Now however, the polling agencies query only ‘registered voters’, or simply ‘adults’. This skews the results, and to my mind impairs the credibility of the conclusions reached. Even so, sometimes valid conclusions can be reached by a careful study of the poll results, if sufficient details are present.

I took a look at two recent polls, once each by the LA Times and Bloomberg, and by CBS News and the New York Times. What I found tells a much different story than the media would have you believe. The headlines in the LA and NY Times do their utmost to sell the lie that Bush is failing, when the facts in their own polling speak to the contrary. I chose to detail these two polls, because no one can reasonably claim that the LA and NY Times or Bloomberg or CBS News has any interest whatsoever in advancing support for President Bush or the Republican Party. Therefore, any statement in their polling which does so must be at least as strong as stated; their pretense in print scurrying to the Left as always.

First, I would like to address the methodology of both polls. The LA Times/Bloomberg poll fails to fully disclose the nature of their respondent poll, stating only that “1,555 adults nationwide” were contacted “by telephone January 22 through 25”. While the LA Times claims that the respondent poll was “weighted slightly to conform with” Census numbers by “sex, race, age, education, and region”, they also admit that “certain subgroups” may have results which skew the error margin “somewhat higher.” Just how badly the poll misses the target, one can only guess. It should be noted from the LA Times’ statements however, that they did not weight their polling to match known party identification, they did not make any effort to ensure that rural adults were properly contacted, or that the unemployed and low-level employed would not be over-represented. Yet the nature of their methodology (all-telephone polling, with three of the four days during the work week) strongly implies a bias against professionals and traditional family structures, and hints at heavy dependency on the urban, the under-employed, the coasts, and the young (this can be done while pretending to follow Census guidelines by allowing for broader age classifications – legitimate polls limit age groups to static ranges, but newspaper-based polling often accepts a variable range, allowing them to manipulate polling by oversampling, say, 20-25 year olds and using them in place of 35-39 year olds, but lumping them all in the same class to pretend objectivity).

As for the CBS News/NY Times poll, I have mentioned before that while I find their conclusions unreasonable many times, I still respect the poll for relatively transparent reporting of its respondent pool. The CBS/NYT poll also contacted its respondents by telephone, but for over five days instead of four, including the weekend. The CBS/NYT poll did not reveal specifics on the demographics of its respondents, but weights its poll according to the 2000 Census. A key indicator, which CBS/NYT has always supplied, is the hard numbers of party identification. The unweighted tally for the poll showed 1.229 adults, with 469 identifying themselves as “Independent” (yes, that seems a bit high to me), 388 calling themselves “Democrats”, and 372 calling themselves Republicans. After weighting, the poll decreased the Independents’ weight to 456 effective respondents out of 1,228, and decreased the Republican weight to 360, while increasing the Democrat weight to 412. Obviously I find such weighting absurd in the face of known results from the 2004 and 2002 elections, but at least with the CBS/NYT poll, they are up-front about telling you how they played with the numbers. The reader can understand the lean the poll has from the beginning.

With this in mind, we can now examine the meat of the polls. I skipped the summary presented by the papers because the screed and lying was predictable. I say ‘lying’, because the facts in the poll results themselves tell a story worth hearing. In the LA Times/Bloomberg poll, for instance, despite the clear bias of the poll’s sponsors, the actual results bear up rather well for Bush and the GOP. On page 13, for example, Bush rolls in at 46% overall for “honest and trustworthy”, but that’s better than the Republicans’ 38%, which is better than the 36% claimed by Democrats. Note that Independents gave the GOP in Congress a 38 favorable to 34 unfavorable nod, while tagging the Democrats with a 31 favorable/41 unfavorable slap. That alone proves that the LAT heavily weighted the polling to avoid a steep lean against the Democrats. Or to put it this way, let’s see the numbers as the LAT printed them:

Group.....ALL.....DEM.....IND.....REP
Dem Fav.....36......54......31......18
Dem Unfav...45......27......41......70

Rep Fav.....38......16......38......71
Rep Unfav...44......67......34......19


If we simply balance the three groups evenly, here’s the new result:

Group.....ALL.....DEM.....IND.....REP
Dem Fav.....34......54......31......18
Dem Unfav...46......27......41......70

Rep Fav.....42......16......38......71
Rep Unfav...40......67......34......19


That is, a balance response moves a 2-point advantage for the GOP up to an 8-point advantage. Well, well.

Even though the poll is clearly biased against Bush and the GOP, on page 14 we see that overall response to the question of whether George W. Bush’s polices “have made the country more secure” is Yes by a 52 to 21 point response. Even Democrats (34% to 33%) have grudgingly agreed to this. If this were weighted to correct the bias, the support would be even more pronounced.

On page 15, we see that even this slanted poll admits that President Bush is better trusted than the Democrats in Congress to protect the nation against terrorism, by a 45 to 32 margin overall. 55% of the people overall think the “economy these days is doing very well or fairly well”, and 61% think of their “personal finances as very secure or fairly secure”. On page 16, we find that overall, 58% of the respondents are more secure financially, or about as secure, as when Bush became President. Bye-bye, mandate for class warfare!

Well, there’s always the PATRIOT Act. Surely people hate that, right? Overall, 59% “agree with those who want to reauthorize provisions” of the Act. Hmm, well that’s interesting.

Well, what about that NSA domestic spying thing? While the issue is controversial, 52% of the total respondents said that even if “a congressional investigation finds that George W. Bush broke the law”, they would say his action was “not an impeachable offense”.

As to Iraq, we find on page 21 that everyone agrees that the U.S. is doing more to win than the insurgents are. Even Liberals agree on that point. On page 22, we find that overall, even in this biased poll the consensus is that the war in Iraq is “part of the war against terrorism”, 51% to 46%. Again, a balanced weighting would show an even stronger support for this fact. But the clear message from this poll, is that the President and Republicans are so strong, that even after skewing the weights to over-count the Left, the “summary” writers had to ignore the facts in their own poll in order to pretend that Republicans or the President are doing poorly. The evidence says otherwise, and strongly.

As for the Voice of Dan Rather, one other thing I like about the CBS/NYT is their internal consistency. As a result, a reader can track progress in the public mood by noticing movement within the poll. Even though the poll skews Left, it does so consistently enough that the reader can compensate. And that makes it interesting to note, that on page 1 of the poll, CBS/NYT admits that Bush’s Job Approval numbers are higher than their January 9th, December 7th, or October polls. On the opinion about whether Bush has “strong qualities of leadership”, his present numbers in the new CBS/NYT poll are higher than in December or September of last year. Support for his actions regarding Katrina are the highest CBS/NYT tracked on the question. And CBS/NYT was forced to admit that 61% of respondents think his NSA authorizations were intended primarily to fight terrorism. His approval numbers on that question have climbed 4 points in 3 weeks.

CBS/NYT was forced to admit that the American public “support the idea of monitoring the communications of those the government is suspicious of”, by a 68% to 28% count.

CBS/NYT wanted to sell that people are not especially expecting the economy to get better, but that’s in some part due to the fact that 57% of the people already think it’s in good shape. And 61% think it’s staying the same or getting better.

Of course, both polls played the “push poll” game, by asking questions with emotional weight while hiding important information from the respondents. For instance, they asked whether people felt there were more terrorist incidents now, without citing the statistics showing the actual conditions. They asked about how people felt the economy was doing, without pointing to low unemployment and inflation, and strong GDP growth over Bush’s time in office. But it is still interesting, for all of that, to notice how these enemies of George W. Bush, however inadvertently, are documenting the measure of his success and the stability of his support.

No comments: