Yahoo had a couple of blurbs on the failure of political polling. It seems all the political analysts were very surprised with the Tuesday election results. If one believed Huffington Post, the Democrats were not only going to keep control of the Senate but pick up a few more seats there and in the House. Surprise, surprise! My university degree is in psychology and I have takes a number of courses in metrics, statistics, method design, and experimental methods. So when a writer who has very little or no knowledge of these concepts expresses dismay that somehow the polls were fooled or some such other reason, I chalk it up to basic ignorance. Indeed, in most of the social sciences there is a basic ignorance of test and measurement. Even the Social Psychologists don’t seem capable of getting it right. And it is not just preconceived ideas of how the outcome should be.
House said it correctly, “People lie.” When psychology first started as a science, the idea was that the investigator would report of whatever phenomania was under study. Look at a sheet of red paper and report what you see sort of self reporting was the order of the day. The problem is that a metric is scrutinized by two ideas. Does the metric measure what it purports to measure, or is it a valid measure. If I want to measure height, width, depth, does my measuring device accurately measure those features? The second is the reliability of the metric. That is, if my measure of height, width, and depth is valid I should get the same results each time I use it. The whole point of valid experiment is that any one who uses the same methods and metrics should be able to reproduce the results. If that is not the case then the experiment is deemed unreliable. When it comes to experimentation a third point comes into play. That is, is the experiment run without error induced into the result. For those in the physical arena the classic example is the measurement of the electrical energy of an electron. Such a particle is so small that almost any attempt to measure its energy will induce error into the measurement. This also applies to the social sciences of which psychology is a member. The expectations of the experimenter may influence the results. Finally, the forth pillar is choosing the participants by random form a valid sampling of the population.
Next we should consider the method of polling. For these political polls used to predict the results of any particular election issue or candidate, very often the questions that are asked are constructed such that they may influence the answer. Just like the old question, “Do you still beat your wife?” The implication is that you have, in the past beat your wife. How you ask the question can determine what result you want to show. I remember the old joke about two monks, each from a different order, who met one day and were comparing the differences in their orders. The first, a Dominican, indicated that his order did not allow him to have a glass of sherry while saying his office. The other monk, a Jesuit was surprised to hear that for he always had a glass of sherry while saying his office. The Dominican ask, “How did you manage that?” The jesuit replied that it in the way he asked the question. He said that he asked his superior if he could say his office while having a glass of sherry. In the classes of Educational Psychology I learned how to construct tests with questions that would best question what students had learned. The art is ask and teach at the same time. Multiple choice tests are difficult to construct as far as measurement goes. An essay question can evoke quite a bit of information but sometimes leaves the question of interpretation. True/false or yes/no questions are the least reliable since the student has a 50/50 chance of getting it right.
Political polling has always been more art than science. It is the combination of what questions you ask of how and if the answers are truly representative of the population as a whole. Remember that while not every citizen is a registered voter and not everyone who is votes, the problem comes when these individuals are included in the sampling. If you ask someone who is not registered to vote who they are going to vote for, are you getting a true representation of the whole? And if you ask someone who is registered but seldom votes, is their answer valid. And one must try to choose a fair sampling of those who claim to belong to a political party and those who claim they are independent. If one registers as independent are you truly independent or do you bring along with you the baggage of the political party you once belonged to? The problem with political polling is that it tries to short-cut the process. This is why focus groups often lead others astray. One of the best examples is when Ford Motor Company did extensive polling and focus group in the design of the Edsel. This was suppose to be the best vehicle design since sliced bread. It was a flop and production was discontinued after two years. People lie.
Lastly, if you are a social media presence, be it news or political commentary, and your readership is dependent on your agenda, do you conduct honest attempts at polling? If you are the Huffington Post or any other progressive and liberal organization with a large readership, do you conduct an honest pool and report the results that do not agree with your agenda? The answer is no. The reason why is that any evidence that shows individuals that they may be wrong in their thinking tends to affect your paycheck. This is why the political and social medias that purport to report the news of events are almost always dishonest.