Chapter 5: Polls and Surveys
By Russell Varner
November 24, 2008
Polls and surveys are very important to newspapers and their stories. Sometimes, those polls and surveys will be all an article is talking about or the main part of it. It’s very easy to see why they are so important for any journalist to understand.
But, Kathleen Woodruff Wickham mentions, polls and surveys “offer only a glimpse of public opinion, and often are skewed one way or another. It is the reporter’s job to help readers understand the validity of polls and surveys they are reading about.”
So here are some tips to help you understand the polls and surveys you as a future journalist will one day be writing about.
•Polls are an estimate of public opinion on a single topic or question usually used in political circles. Surveys are also based on representative samples, but they usually include multiple questions and are used in many different social science settings.
•Cluster sampling involves just one area or region (usually defined by a ZIP code or county).
•Multistage sampling involves picking a specific geographic area, then a random sub-group, individual blocks within that sub-group and then a smaller block within that.
•Systematic random sampling is where one picks a specific number (for example, 10) and then going through a phone book, city directory or other reference book and then polling every 10th person.
•Quota sampling tries to select a sample based on known demographic characteristics. The example Wickham uses is in a poll of women who work outside the home and have school-age kids, one could create a sample that would include an equal number of women who work in offices and factories.
•Probability sampling involves “putting all of the potential subjects in a hat and drawing out a designated percentage.”
•Margin of error shows the degree of accuracy of the research based on standard norms. It’s expressed as a percentage and is based on the size of the randomly selected sample. Statisticians have also found that as the confidence level increases, so does the margin of error. It should also be included in all stories related to the polling along with an accurate interpretation of what the percentage means.
•Confidence levels are the percentage, or level, of confidence researches have in the results of their research. The formal definition is “the probability of obtaining a given result by chance”. It is usually determined in advance and falls at 90, 95 or 98 percent. Also, say there is a confidence level of 95 percent in the research. That means there is a 5 percent probability that the result occurred just by chance.
•Adjusted figures are figures that are statistically manipulated to make up for missing data, while unadjusted figures are the opposite. Adjusted figures are commonly used in the U.S. Census, which also releases unadjusted figures.
•A z score, or “standard score,” shows just how much a certain figure differs from the mean. The unit measure used is the standard deviation and z scores can be either positive or negative. The formula to find a z score is as follows:
oz score = (Raw score – mean) / standard deviation
1.) What is the difference between a poll and a survey?
–Polls are on a single topic and are usually used in political circles while surveys usually include multiple questions and are used in many different social science settings.
2.) What type of sampling do you think is used for the U.S. Census? What would be the best sampling method to use for the Census and why?
3.) A group of students took an informal survey following the Vice-Presidential debate I October. Out of 230 people, the students found that 37% said Senator Biden won, 23% said Governor Palin won, 12% said it was a tie and 28% had no response. If the students used a 95 percent confidence level, what was the margin of error?
4.) You are given a raw score of 64, a mean of 59, and a standard deviation of 2. What is the z score?