The Polls

A number of people can not believe that Clinton is getting such high approval ratings and have taken to criticising the polls themselves. On various call-in shows, Iíve heard people claim that the polls are fixed, that the polls study the wrong people, that 1200 people can not possibly give an adequate picture of the population, etc. I heard one man point out that 60% of those polled did not know what impeachment meant even after it had occurred. He wondered how accurate a poll could be when the subjects did not understand the question. The mathematics supporting polls is indisputable but I doubt that most people know what information polls give. I took statistics in college and I find it difficult to explain to my daughters the math behind polls. A poll does not say that it gives a reflection of the total population within 3% of the real value. It says that its answers are within 3% of the real value 98% of the time or some other combination of the values determined by the number or respondents. For instance, you could be within 1% of the real answer 80% of the time; but the values that are most useful are those in the range of being with in 5% or less of the real value 95% or more of the time. These values require about 1200 random respondents; more respondents are not worth the increase in accuracy.

One problem with polls is predicting the future. I remember the pollsters missing Reaganís í80 election, the Republican victories of í94, and the democratic gains in í98. Two weeks before the í98 election, I heard Zogby predict 17 new Republican representatives. The Republicans lost 5 seats in the house. The pollsters could make the argument that their predictions helped the underdogs make the proper decisions but I have my doubts. The problem is identifying who will vote. From the interviews Iíve heard, the pollsterís methods are vary crude. They simply ask if the subject is going to vote or if he voted in the past. The polling agencies could consult with psychologists or someone who would ask more oblique questions to identify voters. I can imagine another way to go about it: Break down the population into groups and then determine what percent of those groups actually vote. Instead of identifying likely voters, the pollsters could multiply each respondent's answer by the percentage of his group that voted.

The most difficult task for the pollster is getting a truly random sample. In telephone polls, for instance, they reach only those people who have telephones and answer them. I would never pick up the phone for a pollster. Itís not that I wouldnít participate in a poll; itís just that I would assume that the pollster was a salesman disguising his intentions.

I thought of a method for ensuring greater randomness and I donít think they use it. They should study the results of a poll and if they find that one of the answers is not within the confidence interval (I think thatís what they call it) for a known characteristic of the population, they should throw the poll out. For instance, if they find that 40% of the respondents are over 60, they should throw out the poll if that finding is not within 5% (or some other figure they could determine) of the actual population. Of course, this would be expensive since Iím sure all polls show a lack of randomness.

A better solution would be to poll a few hundred extra people, break down the sample into groups by age, sex, or whatever and have a computer choose the 1200 or so respondents who would give a sample that fit the known values of the populations. A computer program could look at all possible samples of the respondents that fit the known values and then average them. It would be interesting to see if this method produced results that were significantly (and mean significantly in the technical sense) different from the greater poll.

Back to Clinton's approval. Clinton's approval rating has been higher for the entire year after the scandal broke than at any time in the previous year. It's obvious that at least 10% of the population who did not think Clinton was doing a good job now believe that he is doing a good job after he was proven to be a lecher, liar, perjurer, and justice obstructor. If he remains president (and I think the odds are 9 to 1 he will), I predict that his approval rating will be under 60% in 6 months (unless he creates another scandal). Regarding that 1 in 10 chance that he will be kicked out of office: it could be that the Senators will all of a sudden and unexpectedly find him guilty. They could be hiding their intentions now so that Clinton won't start dropping A-bombs on someone or something worse to remain in office.

Thank God the man has offended the military forces. If not for that, he'd probably provoke an emergency and declare himself PRESIDENT-FOR-LIFE.