Political Opinion Polls

In the following opinion on "Opinion Polls" by anonymous CJ, he is confining his discussion to political polls. He essentially finds opinion polls of no value, for the reasons given.

You decide whether to put any credence in opinion polls, as continually reported by the news media.

While I tend to agree with CJ, this leaves me in the uncomfortable position of assuming there is nothing to predict reasonably the outcome of an election.

                       Opinion Polls by CJ

The media are infatuated with public opinion polls and talk about them regularly.  In fact, most reporters either do not understand the limitations of opinion polls or simply do not provide the basic information for their evaluation.  Statistical theory clearly shows that a valid sample of a population will provide a good estimate of outcomes. 

In political opinion polls there are three populations from which to draw a sample:

a. Those of voting age

b. Registered voters

c. Likely voters

Conclusions drawn from sampling the first population are generally meaningless in terms of probable outcomes.

Surveys of registered voters are unlikely to provide a valid sample since only 55% of registered voters actually vote in presidential elections.

Likely voters is the population that will provide a good estimate of outcomes.  However, that population is the most difficult to identify.  In personal telephone surveys, the polling agency tries to define the likely voter population by asking interviewees about prior voting participation and assuming that the answer is the truth.  Just think about the likelihood that an interviewee will admit that he or she did not exercise the voting franchise. Surveys done by “robocalls” have even less chance of identifying likely voters.

Now think about how many times a reporter identifies the sampled population.  Instead, they trumpet the “margin of error” whatever that means.

“ Figures can lie and liars can figure”

CJ