Sunday, July 23, 2006

More Fun With Poll Math

Perhaps the most amusing thing I have noticed this week from Survey USA’s State-By-State Job Approval poll, is the way Liberals want to throw out anything which won’t fit the bumper sticker. Blacks must not only hate President Bush, they insist, it must be pervasive throughout the nation. And the same for Hispanics, they demand, there cannot be a state, much less many of them, where the President is respected, or worse, admired?

There are two important messages coming from these polls. The first is that there are many states where minorities like and trust President Bush. It’s simply foolish to pretend that support or rejection would be nation-wide, yet that is exactly the reaction from Liberals – they cannot accept what even simple common sense would tell them.

The second message comes from those self-righteous prigs who on the one hand want to ignore Bush’s support in certain states, because they dispute the sample size, yet they expect to be able to cling to results presented which use equally small respondent samples. A few examples:

The Associated Press (AP-Ipsos) has locked President Bush in the mid-thirties for Job Approval for quite a while, and claimed 11% Black and 13% Hispanic response portions of a total respondent pool of 1000 adults. Those people trumpeting the results must, in the interest of objectivity, admit that 110 respondents is far from the same statistical accuracy of one thousand adults, nor is 130 much better. My own results had the virtue, at least, or cross-supporting separate states, something not evident in the Ap-Ipsos internal data.

Gallup is the most reliable nationwide pollster, but even they admit they weight by nominal demographics, which means that a standard respondent pool means significantly less reliable results from minorities.

But it did give me an idea. One thing which turned out to be different than it seemed, was that Survey USA asserts that despite the apparent control by local television stations in managing the direct polling, the stations were merely the sponsors of the poll, which Survey USA handled in every location.

Presuming then, that the methodology is consistent throughout all fifty states, and taking into account the other weighting factors based on prevalent assumptions on religion and political allegiance, I found that applying the reported results by Black or Hispanic responses and tallying them up together gave the following overall nation-wide result:

Blacks: 2,826 responding
426 Approve (15.0%)
2,343 Disapprove (82.8%)
61 Not Sure (2.2%)

Hispanics: 2,064 responding
722 Approve (35.0%)
1,311 Disapprove (63.5%)
74 Not Sure (1.5%)


At least this gives a statistically valid reference point for consideration. The prior comments made should also be considered to demonstrate the regional and demographic vectors which affect the specific topography of the results.

1 comment:

SurveyUSA said...

DJ, just to clarify:

SurveyUSA's monthly tracking polls are run by SurveyUSA, using questions written by SurveyUSA. While we use the voice of our clients' anchors to conduct our research, our clients do not conduct the interviews themselves.