Friday, January 11, 2008

Driven poll events: What if the poll is not accurate?

Opinion Polls represent a kind of summarized information about society ´s mood, concerning some specific issues. From the economic point of view, it represents the reduction in transaction cost, which authorities can count on in their decision making process .Thus, if anyone want to discuss abortion, divorce, environment, political preferences, regulation and the like ,opinion polls show a convergence path about them, to reduce the chance of mistake for any policy design.
However, what is the real relevance of this opinion poll?. Well it depends upon both the quality of the sample, and expert´ s judgement to explain its results. Let take an example: you want to know how people would react , in extremes case such as an alert of tsunami. You guess that between stay in a shelter or running away, at leas 90% would run away , whereas other expert might think that 50% would spilt between running and seeking for shelter. You decide to do some research about it, so you ask randomly to 25 people about what the would do in case of a tsunami: Assume that 10 people of this sample cares about tsunamis ,and 70 % say that they would run away from it, while 30 % say they would seek some shelter around. Both answers imply a public policy requirement either to build up proper shelters ,or to set a proper signalling mechanism for those who would run away. But is that it?. From the public policy point of view, it is important to use efficiently scarce resources, therefore it is a better approach to improve the information about what it is at stake: If you spend money on the basis of poll results, it might end up to be a waste money, because we do not know the probability of occurrence of each one of such event. The next question is , Taking a sample of those ten people ,what is the probability that seven of them, would run away in case of a tsunami threats?. Applying statistical tools, the probability would be 0,0574. What about staying at a shelter? .The probability would be 0,117. Therefore, it is more feasible for people to look for a shelter, than to run away ( 5,7% probability in the former, against 11,7% in the later) . Thus, the correct policy option should more be in favour of building up shelters , instead of concentrating on signalling equipment . Quite the opposite, if the decision would have been based on the first poll result.!. Beware that this conclusion is sensitive to the sample size. It means that whether 8 instead of 7 people run away ,the probability changes.-
Something similar it might happen in a political election: Two candidates go on a very tight race, although one of them have a slight lead. Is that enough to conclude that the one with the slight lead ,would win? .Recent experience shows it might be not. The result depends upon the probability for voters, to keep up to their preferences. Thus , it can happen that the probability of keeping their vote in favour of the one who is leading the polls, is lower than the probability to change their votes, of the other candidate´ s supporters. It means that the one who is leading the poll , has more volatility in his voters preferences, than the one who is behind .In such a case , if you rightly consider the probability scenario as a complement , you would have more information to take a better decision.-
Most of the time, opinion polls show the event in absolute yes/no terms. However, modern times offers citizens a lot of more options to look upon for additional information , which at the same time allow them to shape in a better way their preferences, affecting deeper than expected the probability of changing them. In the information society , instantaneous event make preferences more unstable.-