As NewsBusters has been reporting the past 24 hours here, here, and here, CBS News released a poll yesterday depicting President Bush’s declining job approval numbers that were based on a sample that was in no way reflective of the country’s population. Just hours ago, some new poll numbers were released by the Washington Post and ABC News, and they too have over-sampled Democrats and Independents.
As stated at the ABC News website:
“An increasingly unpopular war, an ethics cloud, and broad economic discontent have pushed public opinion of the Bush administration from bad to worse, infecting not only the president's ratings on political issues but his personal credentials for honesty and leadership as well.
“George W. Bush's approval ratings for handling his job, Iraq, terrorism and the economy are all at career-lows. Sixty percent of Americans disapprove of his work in office overall, a level of discontent unseen since recession chased his father from office.”
Yet, the breakdown of the respondents once again showed an over-sampling of both Democrats and Independents, and, as a result, a significant under-sampling of Republicans. As stated at the Washington Post website, the breakdown is as follows:
Democrat Republican Independent Other No op.
11/2/05 31 27 38 3 1
9/11/05 31 27 34 8 1
9/2/05 32 31 31 5 2
8/28/05 32 29 33 5 1
6/26/05 34 28 32 5 1
6/5/05 30 31 34 4 *
4/24/05 35 28 32 5 *
However, when one looks at the political leaning of the respondents, the skew becomes even more apparent:
901/904 Leaned Party:Democrat Republican Independent Other No op.
11/2/05 52 41 6 1 *
9/11/05 48 39 10 3 *
Without going through the boring statistics again, the point here is that it is not difficult for a polling agency to create a representative sample of the population. In fact, for their poll to be viewed in any serious fashion, a poll MUST represent a fair and accurate portrayal of the population, or else it is worthless. And, according to the 2004 exit polls, the current population is 37 percent Democrats, 37 percent Republicans, and 26 percent Independents. As a result, any poll that deviates from these percentages is not accurately reflecting the views of the population.
So, how does a polling agency ensure such an accurate reflection? Well, by calling more people on the lists they have. After all, there are many, many millions of registered voters who have identified party affiliation with their registration, correct? If the targeted sample size is 1202 people, there are more than enough Americans out there who are registered to vote and have telephones to ensure that they include 37 percent Democrats, 37 percent Republicans, and 26 percent Independents. Think about it -- this means only 445 Democrats, 445 Republicans, and 312 Independents. A good telemarketing division should certainly be able to accomplish this in two days.
Moreover, as we are twelve months away from a national election, there really isn’t any urgency in releasing poll results, is there? As such, wouldn’t it be more important to try to get this sample right than get the results out as fast as possible? After all, from the above numbers, this polling agency hasn’t been able to accurately replicate the 2004 exit poll numbers since at least April 24 of this year.