Don't believe polls

    Polls are quoted as if self-evidently valid (without considering the source or methodology), but in fact, I have come to believe that polls and opinion surveys are almost entirely worthless.  

    The biggest reason is that with the disppearance of landlines, there is now no easy way to randomly select from the population.  In addition, there has been a proliferation of polls on everything under the sun while the surveys themselve have become increasingly long and tedious.  So even if there was a way to select a random sample of the population, most of the initial sample may not respond to the survey.

    For these reasons, even though pollsters can analyze their data very precisely, the results may be entirely unrepresentative with respect to the population as a whole.  Indeed, those who respond to surveys, particularly phone surveys are probably not evenly distributed among the population as a whole.  Many of them may be high-conscientiousness members of the generations who grew up instinctively trusting people who call them on the phone and instinctively trusting polls.  On the other hand, there are probably very few people under 35 who answer surveys at all, especially phone surveys.  Futhermore, I suspect that most young children or teengers who answer polls are probably bribed in some way (free food, for example) and so for that reason their responses should be taken with a grain of salt.    

    But it is even worse than that because often surveys are badly designed: the meaning of a particular question to the designer may be entirely different from how the responders view the question.  Also, frequently surveys are not able to capture the nuance of individuals' views because most surveys are not open ended, all possible responses are proscribed in advance.  

    And those considerations do not even take into account fake surveys.  At worst, the survey may be entirely fraudulent, with numbers made up on the spot.  But even if that is not the case, it may be designed so that only certain responses are possible, or interpreted in a disingenuous way, or presented to a deliberately nonrepresentative sample of the population.  

    Therefore, the burden of proof is on the polls.  A survey can only be trusted when it has been shown that those who design and implement it are both honest and competent.  And even then, it may be that a representative sample is so difficult to find that a honest and competent surveyor has to conclude that no conclusion can be drawn.      

    If we wan to understand the modern world, let us base our analyses on something more substantial than polls.

2 comments:

  1. I agree - there are problems with opinion/ attitude surveys at every level. The only safe conclusion is to assume that they mean nothing at all - except what those reporting, want them to mean.

    ReplyDelete
    Replies
    1. Exactly. At this point, polls are just done to find "evidence" for what people already want to believe.

      Delete

The real AI agenda

    On a post  by Wm Briggs, about artificial intelligence, a commenter with the monniker "ItsAllBullshit" writes:           "...