|Reflexions on Surveys|
Guillermo Ramón Adames y Suari - PVNN
November 15, 2010
Why is it that whenever you read a survey you and the people around you never agree to the results? This question is asked to me as a statistician more often that you can imagine. Sometimes surveys are very well thought of and extremely well planned, and they give true results. Those are the best. Statistically it is extremely interesting to plan and carry through such surveys. It is very professionally satisfying. That is the way in which surveys should be carried on. But there are some flaws.
As a professional statistician, my answer will be split in five parts:
1) Most surveys concern crucial issues and most people are very deeply convinced of their ideas. Often enough people simply cannot conceive a slight variation of their personal convictions. Truly enough they often react. As an example take a political survey. Usually the “second” person is a “stupid”. How come people cannot simply “see” that the second person in the survey is completely out of his (or her) mind? It is “so obvious”. Of course you agree almost fully with the first person. You also want to show that you cannot be in completely full agreement with the first person, but almost 90%.
2) Sometimes your approach is disguised by stating: “person number one is the less worse of the two”. So he (or she) will become your “leftover” choice. Unfortunately you agree to a lesser extent than 90% but at least 55 or 60% otherwise you would choose the second person.
3) Your environment defines tendencies. If you live in a trendy area and work in a like environment, you are likely going to find people thinking like you do. Just imagine the question concerning sub-primes (money is a crucial issue like politics and religion). So you, as a banker, are likely to ask the banker next door whether he will agree in the suppression of sub-primes. But you will not certainly ask the tax payer who just lost his home and has nothing to eat. Certainly his tax money is paying the rescue of your Bank or Insurance Company, car factory or you name it, so that you and thousands of others do not lose their jobs. He was not asked whether he wanted to lose his shirt for you but he did. Usually people are so much in their own ideas that sometimes they even forget that there are “other” ideas.
4) Surely enough, sometimes surveys are invented or exaggerated. It is easier to do the second ones as you can actually measure the tendencies and calculate to which extent you can lie. Whether the lie that you are trying to “pass” or “sell” is not over done: “it can actually be possible”. The scheme is complete - you did the planning, collected the data, did your computations, and presented the graphs and summary statistics. You actually have costs and individual data. But in this case, as an invented statistical result, there is always something that does not appear coherent with reality or with the rest. If your survey concerns a political candidate, he (or she) is so fine so far. In a lie, the structure collapses at one point or other. You can present a very good fake survey and give the impression that it is a real study. Usually these studies are very short and concern a single pointed question. There is no space for argument. There are situations where you can present a fake study. As an example of fake studies you can find political personalities in “not very democratic countries”. The country simply would not survive without his leader: the best example today is North Korea or Burma (Mynamar). We have a couple of examples on this side of the planet but the idea in this article is “Surveys” and not “Politics”.
5) The last part I want to consider is marketing. Pure marketing: Either economical or political marketing. It has been measured statistically that you can influence “the undecided” to a margin of about 15-25% if you “present a solid” argument: “95% of the people use detergent X”. So your response is “95% cannot be wrong so let’s give it a try and if not I change afterwards”. In political marketing the arguments are more subtle. Out of a candidate, you stress his main one or two points. Of course candidate two does not have them and the country would collapse without these. Political psychologists know how to evaluate the momentum of the argument and the argument itself (which of course was included in the “fake” survey). After all, sometimes you simply need 25% of the undecided to ensure a victory.
Guillermo Ramón Adames y Suari is a former electoral officer of the United Nations Organization. Contact him at gui.voting(at)gmail.com.