Political Polling Essay, Research Paper
The web site I chose to critique explains political polling and the statistical process behind the polling numbers. This was of particular interest to me, because I like many other Americans have never been asked nor know of anyone that has been asked a political question used in political polling. In short, the site explains what the numbers actually mean. The site uses a fictional mayoral election to explain the numbers. Also, it discusses margin of error and how it affects the polls. Next, the site goes on to clarify confidence. Finally, elucidation on what can go wrong in the polling process is discussed.
In the fictional case Republican Stephanie Higgins was running against two-term incumbent, Democrat Webster Fletcher. Once Higgins formally announced she was running for mayor, the polls showed that Fletcher would win with 56% of the vote. However, six months before the election the mayoral race was tightening up. The polls showed that Fletcher continued to have a slight lead of 3%, with a margin of error of +/- 5%. The site raised the question of whether or not Fletcher actually had a lead if the margin of error is +/- 5%. Three months before the election Higgins took a 15% lead in the polls with same margin of error. Oddly enough, the day before the election some polls showed Higgins winning by a landslide. As expected it was a close election. The lead changed hands several times. However, Higgins pulled ahead and won the election in its last moments.
To begin the site discusses the importance of the random sampling used in political polling. In class we defined a random sample as, everything in the sample stands the same chance of being selected at any point and any time. A great example was given on the site. If a doctor wants to figure out a patients white blood cell count, the doctor doesn t drain out all the patient s blood and count the white blood cells. The doctor randomly samples the patient s blood by pricking their finger and counting the white blood cells in that sample. This will give the doctor an accurate idea about the patient s total white blood cell count. However, the article is quick to point out, that if we prick the skin on any part of the body to draw out a drop of blood, we can be certain that the drop has the same properties as the rest of the blood. However, this is not the case when dealing with human opinion about political candidates. For instance if you only select an area that consist mainly of a specific race or average annual salary for political polling, you are leaving out a severe amount of the population. This is why it is necessary for pollsters to create random samples that accurately represent a cross section of the entire population.
Margin of error is the next issue tackled by the site. Margin of error is one of the most difficult concepts to understand. Of course, we can understand margin of error due to our in class work with confidence intervals. The margin of error can be computed two different ways. The first being to multiply a z or t score by the standard deviation of a population or sample, divided by the square root of the number of units in that particular sample or population. This method is usually used in compute the mean of a sampling distribution. The second method is used to help estimate the purport ion of a population by multiplying a z score by the square root of the average sample of a population, multiplied by one minus the average sample of a population, divided by the number of samples used in that particular sample. The site explains this process a whole lot easier. For example, in the begin of the fictional mayoral election, Fletcher holds 56% of the vote with a +/- 5% margin of error. This would mean that if the election were to be held that day, Fletcher would receive anywhere from 51% to 61% of the vote according to the polls. One key factor associated with margin of error is the greater number of samples, or in this case the more people pollster s talk to, the smaller the margin of error will be.
The site pays particular attention to confidence intervals. Sighting different times when confidence intervals are used. For instance, the evening newsman might say they are 95% or 80% confident in their results. So, what does this mean, what is a confidence interval and how is it computed. A confidence interval is the range of numbers that lye within the two ends of the aforementioned margin of error. The interval reflects the calculated average and those numbers that might fall outside the average. When a pollster states they are 90% confident, this means: If the poll was repeated 100 times, in 90 of the polls the answers given by people would be the same. Obviously, 10 of the polls would get different answers. It s important to remember there can never be 100% confidence.
Finally, many factors that can keep a political poll from being a perfect gauge of how all people think, feel or behave is discussed. Many things can go wrong in all testing that uses confidence intervals. Dealing solely with the topic of political polling, most polls are conducted by telephone. If we randomly select names from phone book, we are leaving out individuals who have unlisted or no phone numbers at all. Also, research shows, most phones are answered by older people or women. Only selecting people from these two individual groups leaves out a considerable amount of the population. This causes a bias within the polling. A final and most significant problem with political polling in the United States is that only about 50% of registered voters actually go to the polls and cast a vote. Therefore, only half of the sampled population will matter when taking in account their political views. However, all of the people polled will be measured when estimating political figures.