As many times as I have read polls in news stories, I have never even thought about the reliability of the polls. If it was from a news source that I trusted, I trusted the poll. I feel that this is a problem many people have when reading stories, they trust the source and therefore trust the poll. After doing this reading, I realize that it is up to us to find out if the poll is reliable or not.

There are many different kind of polls being used today, including one of the most commonly used, the sampling poll. The sampling poll has many different forms and strategies to come up with the sample. Sampling polls take smaller portions of the population who are intended to represent the larger population. In the book A Journalist’s Guide to Public Opinion Polls by Sheldon R. Gawaiser and G. Evans Witt, they discuss the different ways to acquire the sample, why it is important, the sampling error, and how to avoid getting a high sampling error. In other words, the many ways to get a sample and what to do with them.

Gawiser and Witt start off by emphasizing how important it is to be scientific while conducting a sampling poll. “If the sampling methods are not scientific, the results could be completely misleading.” They compare it to a doctor taking a blood sample to test it, rather than all of your blood, or tasting a portion of soup instead of the whole pot. This is intended to get the point across that sampling polls are/can be accurate.

The samples are very important. Not just any sample will do. The sample chosen has to represent the larger population as accurately as possible. This means that you cannot just walk into the mall or go to a retirement home and poll the people walking through the mall or living in the retirement home. That would only represent shoppers, or the elderly, and not the entire population. Also, you much get a large enough sample if the large population is big. In other words, you cannot sample ten people if the larger population is one-thousand. That would not be a good representation of the thousand people and there would be a large sampling error. The larger the sample out of the population, the smaller the sampling error. Most good polls have a sampling error of plus or minus 3%.

Sometimes, though, when conducting a poll that is supposed to represent a huge population, like say an entire state, it is hard to personally poll enough people to cut down on the sampling error. Probability sampling is one way of doing this. This is when each person has a “nonzero probability of inclusion in the sample.” This is also called a random sample. This means that each person has an equal chance of being a part of the sample. This kind of sampling can work for two reasons: first, each person has an equal chance of being selected. Secondly, each choice is made independent of the others. You could also assign each person a number from 1 to n. N is the total number of the population. This can be done by computer as well. Next, you must randomly select a middle number and the increment that each person will be selected. For example, I will start at 4, and then pick every tenth number after, so I would have 4, 14, 24, etc.

This can get pretty expensive, so to cut down on costs, some have turned to cluster sampling. “In a cluster sample, sets of members of the population are first selected, and then the individual units are selected from the chosen sets.” Gawiser and Witt have cited Sudman for the four rules of cluster sampling: 1. Each unit in the population must belong to one and only one cluster. 2. The number of elements in each cluster must be known or reasonably estimable. 3. The clusters must be small enough that it is cheaper to use them than take a random sample. 4. The clusters should be chosen to reduce the increase in sampling error caused by the use of clusters. Although cluster sampling can be convenient, there is a risk of increased sampling error.

Telephone sampling is another way to take a poll. Disregarding the risk of excluding those who do not have a household telephone, the ten digit telephone numbers are randomly selected by telephone companies to make sure these numbers exist. Once a call is made, it is important to realize who you are speaking to. If not, teenagers may be the group most highly represented. This can be done by asking how many adults live in the household and then how many female adults, or by asking who has the most recent birthday.

All of these factors play into the reliability of the poll. If you cannot identify where the poll came from, how many people were interviewed, and how the people were chosen, it is probably a good idea to disregard the poll as reliable.

Rosalind Sixbey

## Leave a Reply