'Response bias' may have misled polls

Overall, the majority of polling organizations were wrong about support for Mrs. Clinton compared to that for Mr. Trump. One explanation is that the polling organizations intentionally skewed their surveying techniques in order to stimulate a "bandwagon effect for Mrs. Clinton and a "lost cause" effect for Mr. Trump. While such an explanation is possible, a less onerous explanation is also possible.

A random sample obtained from a population will be a good representation of that population subject to a predictable amount of variability; however, for this to be true in the case of public opinion, two things are required.

First, those respondents selected by the "random selection" machinery must be willing to participate. Otherwise, this introduces what is said to be a selection bias into the results, where only those willing to participate will offer their opinions.

The second requirement is that those respondents selected must tell the truth. If they do not, this introduces what is said to be a response bias.

Now, while it is possible to explain the error in terms of a selection bias, supposing the different polling organizations used different selection techniques, we would expect some of the polls to have correctly gauged the support for Mr. Trump. On the other hand, if there was some systematic reason for not telling the truth in the population, we would find all of the surveys having the same response bias.

Given the air of social disapproval associated with Mr. Trump's bombastic personality, it is not inconceivable that many supporters of Mr. Trump may have been unwilling to express their support for fear of being judged ill, and such a response bias could explain the widespread error in the polling. Unfortunately, such a response bias is perhaps impossible to eliminate.

Robert H. Bruhl

Hoffman Estates