- Deb
October 2003 The Public Opinion Polling Fraud from Z Magazine <http://www.Zmag.org> by Marc Sapir and Mickey Huff http://www.retropoll.org/polling_fraud_z-mag.htm
excerpt:
Gallup reports on their web site that after completing a poll they weight the demographics to assure correct proportions are represented. Weighting means that you multiply the results of an underrepresented group by a factor that will bring their input up to intended and expected levels. Another thing that can be done is to simply over-sample in a population that is expected to self select out of the poll. If, for example, you want to double the number of African American responses you just begin with a sample that has 24% African Americans instead of 12%. These tricks of the trade work on paper and in statistical analysis but they both fail to address the important question "why would any particular group be less likely or more likely to participate? And "are these refusers different?"
If those questions sounds familiar they should. For this is just a more specific and powerful example of the pesky problem of the 70% refusers to participate in polls - the problem that won't go away. When we take it to the level of the under-representation of ethnic groups however it is easier to see that there are probably specific socio-political and/or economic reasons why some people are more likely to participate and others to not participate. These can include issues like English language skills, fears of being monitored by race, lack of self-confidence, or poor educational background. Any of these factors or dozens of others that may have an impact on peoples' decision would invalidate the principle of a random poll sample that can be used to represent the general public. If those African Americans who agreed to participate were more wealthy or better educated than those that refused, then adjusting their input upward by a multiplier (weighting them) to provide a bigger contribution would be a charade since their views might not represent those of less educated lower socio-economic classes of African Americans. You might, for example be inappropriately magnifying the views of a tiny group of African American Republicans.
But the pretense of random samples is only part of the polling problem. ...
=====