By Dr. Emmet (John) Fritch
Faculty Member, Reverse Logistics Management, American Military University
As the November elections daily draw nearer, Americans will receive increasing numbers of phone calls from pollsters wanting to know how they will vote.
Historically, U.S. telephones were “landline” devices, and phone numbers were accessible from printed telephone directories. Opinion pollsters relied on inferential statistics to test a sample of people and assumed the sample represented a large proportion of the population. As its name implies, inferential statistics processes attempt to reflect many people’s opinions from a much smaller number of participants in the sample poll.
Get started on your Homeland Security degree at American Military University. |
However, inferential statistics processes were challenging. Acquiring an acceptable sample set was difficult. For example, not all people in rural areas had phones. Today, there are new challenges to polling as a result of far fewer landline telephones and a huge increase in mobile phones, as well as a scarcity of published directories.
A Shift to Random Digital Dialing Sampling Is a More Reliable Method of Polling
A shift to random digit dialing (RDD) sampling provides a more reliable method for increasing the chances that all people in the target population have an equal opportunity of participation. However, meeting the criteria for sample selection is still difficult.
The Differences Between Non-Probability Polls, Probability Polls and Inferential Statistics Opinion Polls
According to The American Association for Public Opinion, non-probability and probability polls are the most common methods of acquiring opinions. Non-probability polls receive people’s opinions, but they are not generalizable beyond the participating samples; non-probability polls represent the opinions of the participants only. Inferential statistics allow probability polls to generalize results to large populations from sample polling.
Many pollsters administer non-probability opinion polls to avoid the difficulty of generalizing samples. As a result, there is confusion when poll results are released, and full disclosure of the methods used is absent.
Non-probability polls are accurate only at the time they are administered. They reflect only the participants’ opinions and cannot be generalized to participants outside the sample. Probability polls rely on rules for inferential statistics to ensure reliability.
Inferential statistics is the favored method today. For the reasons cited above, pollsters developed RDD, which allow sample participants’ randomization to meet the statistical test for sample selection. RDD, however, fails to meet the standard for inferential statistics even though some pollsters will try to convince us that sample poll results reflect the whole population’s opinion.
Results of probability polls are suspect. Below are a few of the reasons probability polls are not reliable. By the nature of the way the data is collected, polling responses cannot be statistically significant. For example:
- According to G*Power HHU, two factors to consider are sample power and error rate. When sample sizes are established, understanding the elements is essential.
- All people in the sample target must have the same chance of being selected. Since polls are mostly conducted by phone, this is not possible. Many people do not answer polling calls (those often annoying robocalls); others no longer have same phone number used in previous polling; and some may be deceased.
- RDD ensures the random selection of phone numbers but does not guarantee that the participants selected have an equal chance of being included.
Pollsters attempt to use statistical methods to justify probability results, but they fail to meet the standards for statistics. Proper methodology ensures that samples represent population opinions.
Sherri L. Jackson summarized the use of telephone opinion surveys in her book, Research Methods and Statistics: A Critical Thinking Approach. She reminds us that as telephones became more established in society, telemarketing calls motivated people to increase the use of answering machines, caller ID blocking, and other techniques to prevent unwanted calls.
Random-digit dialing tied to landline phone number formats became less likely to include participants using mobile phones. She also warns of telephone opinion surveys’ tendency toward bias when questioning participants.
According to the Pew Research Center, opinion polls are capable of helping us understand potential outcomes. To do that, the center advises comparing results to government-funded surveys; the Pew Research Center reports that its poll research closely aligns with government-funded survey results.
Another method to establish the credibility of opinion polls is to compare the actual results of current polls to past surveys. People should also be aware of the sample methods pollsters use. Polls not conforming to statistical principles should be suspect.
About the Author
Emmet (John) Fritch is an Associate Professor of Business at American Military University. Dr. Fritch earned his Ph.D. at Northcentral University. He earned a Master of Science degree in technology management from Pepperdine University. Dr. Fritch has over 20 years of experience in global supply chain management.