Outliers or Out-and-out liars?

Many years ago at the University of Texas at Arlington, I took a senior level course in Public Opinion research.  It was one of the most fascinating classes that I took while in college, because I learned that public opinion research is subject to many unintended errors that can produce a result that falls outside of expectations or is not in line with other similar polls.  In the trade, these are called outliers.

I also learned that the opinion data may be "weighted" so as better to reflect the population as a whole.  For example, if we are trying to survey a group of people about their opinions on colored contact lenses, and we want to sample an equitable distribution of different eye colors, we would first have to determine what the percentages of each common eye color is in the population.  For the sake of argument, let us say that 15 percent of the population has blue eyes, 15 percent green and 60 percent brown.  Unfortunately, when you finished your survey, you found that you had more than 15 percent of your respondents with blue eyes and fewer than 60 percent with brown eyes.  You then apply a process called "weighting" in which the opinions of the people with brown eyes are given more value or importance than those of the blue eyed people in order to conform with the statistical distribution.  Unfortunately, if your assumptions or facts about the actual population distribution of the various eye colors is wrong, then your weighting is wrong, and your final analysis is likely to produce an outlier.

On July 21st of this year, according to the Real Clear Politics average of polls, President Obama's job approval numbers went "underwater".  That is to say that those who disapprove of his job performance outnumber those who approve of it. This has held steady to today, some days reaching as much as a five to six point spread.  Repeated polls from Fox News, the AP, Reuters, CNN and USA Today have indicated that the job disapproval has surpassed the job approval ratings, at least to some extent.

Then, in the last two weeks have come three separate polls from Time Magazine, Newsweek Magazine and CBS Television that suddenly show President Obama's job approval rating to be back "above water" as much as four percentage points.  I found this particularly interesting since the daily polls run by Gallup and Rasmussen Reports still show the president to be "underwater" by as much as six percentage points.

So, I decided to look into what we poll wonks call "the internals".  Those things like the breakdown of the sample including assumptions made about the distribution of the population between Republicans, Democrats and Independents.  It was interesting what I found.

The CBS poll in particular "weighted" the results to over sample persons self identified as Democrats by 1.1 percent and Independents by two percentage points and under sampling self identified Republicans by 2.7 percent.  This assumes that the political landscape of the United States has not changed since the Presidential Election in 2008.

This totally flies in the face of current polling that shows the political divide less than 3 percent.

Although I have been unable to access the "internals" of the Time or Newsweek polls, I have no doubt that their results were "weighted" in the same manner.

Did they make an honest mistake in "weighting" the numbers with outdated population distributions?

Did they deliberately "weight" the numbers to get the result they wanted?

Neither Time Magazine and Newsweek Magazine are know to be friends to Republicans any more than CBS Television.  In fact, quite the opposite.

I can't believe that these are the same people who moan and groan about how Americans don't trust the mainstream media any more.  Do we need a better reason why?