Question: Find two polls from 2007, one of which offers weak support for its conclusion, the other of which offers strong support for its conclusion. Answer the following questions about the polls:
- What is it that is being sampled?
The survey is to find out how many people will vote for the particular democratic candidate if election were called. There are a total of 134,000 registered voters in North Carolina which represents the population.. Out of this population 46% are democrats. There are 10,000 registered voters in Pennsylvania which is equally shared between the two parties. It is assumed that the samples of 676 and 462 were chosen from these populations randomly. How random the sample was is always difficult to predict. This is because the following factors always influence opinion polls:
- The sampling error. That is how sure are you that that there is at least 95% chance that including more people in the sample would not change the sample size by more than the sample error. Let us look at the sample error in both samples. North Carolina is 3.7% sample error and Pennsylvania is 4.6%. Statistically we assume this is a sample chosen randomly, that is a voter has 50 to 50 chance of being chosen for the poll. We shall soon see that this is not always the case in an opinion poll. It is not uncommon to find that other social factors like age group, social status in the town, economic status, education, color / race, sex play a big roll in the selection of the sample as well. This therefore means the sample is not random. Suppose these polls are taken at night in social places. This therefore excludes women and school going children who could be majority.
2) Is the sample size large enough to represent the population?
The answer is yes it is. If we assume the population is normally distributed, then any sample size of more that 34 cases is good enough. The samples are 676 and 462 which are much more than the lowest values of 34. Stasticians say the higher the sample size the better but for large populations the difference in sample size does not bring much change. North Carolina had a bigger sample than Pennsylvania and yet the opinion polls in Pennsylvania has predicted the polls better than North Carolina. The true polls are in order of strength Obama, Clinton and Edwards. The predictions are Edwards, Clinton and Obama for North Carolina; Clinton, Obama and Edwards for Pennsylvania, implying the Pennsylvania prediction which has lower sample size is better.
3) What type of people were included in the sample?
This is always the most difficult factor to look at and could have resulted in the inaccuracy in the sample. As said above several factors come into play during the selection of the sample and it is difficult to ascertain the extent to which the individual factors affect the outcome. Such factors include:
- Age of voter
- Economic status
- Formal Education
- Career at the moment
- Marital status
Statistically when there are too many dependant variables, then prediction is not easy.
- What type of sample was chosen?
As explained earlier the sample is assumed to be randomly chosen.but with some stratification. For example before you were takento vote factors such as:
- Age group, sex, social status, race were considered
- The party you belong to
Assumptions that social attitudes about race and gender would influence the outcome were grossly misjudged.
5) Target Population?
We assume that only registered voters by 2007 were included in the sample. As more voters were registered, the register kept changing. It was not very clear which party would get more new registered voters
6. Poll Rating:
North Carolina Very Weak: Nothing they predicted happened. Things were the opposite of their prediction
Pennsylvania: Moderate because number two and one interchanged their positions, but three was accurately predicted.
Find below data obtained from the statistics given by the Wikipedia about presidential opinion and actual polls in U.S, please treat is as attached document.
- Fay Lomax Cook: Navigation Public Opinion Polls, Policy and Future
- Binghamton University Libraries Virtual Reference Collection
- Lisa Harrison: Political Research An Introduction
- George H. Hallup: Public Opinion Polls Research Guide