Political junkies obsess over politics and use numbers taken from election polls to express their own personal opinions on which candidates should win and which issues ought to prevail. They never tire of this chatter because this is their calling and life’s work. Divided as they often are along political and ideological lines, it’s inevitable that these junkies will select from the many polls available which ones best support their opinions.
If you are tired of hearing who’s ahead and who’s behind, what matters and what doesn’t, you’re not alone. There are even polls taken on election coverage fatigue!
Polls are taken to find out what people are thinking, are doing, and what they intend to do in the future. Questions that are important to organizations that cannot not afford to enumerate the entire population are addressed instead to a sample of persons thought to represent that population. Questions such as Do you think marijuana should be legal for recreational use? Do you use a face mask at all times while you are out in public? Do you drink wine on a daily basis? Do you think that 16 and 17 year-olds should have the right to vote?
To be representative of the relevant population, the sample must be carefully drawn to eliminate any bias in the responses to the questions. For example, knowingly and intentionally over-sampling persons who actively use marijuana recreationally does not produce answers that are representative of the entire U.S. population because large numbers of Americans do not use marijuana. To be representative of all Americans the sample would have to be drawn randomly from the entire population.
Even then, the estimates based on a sample and derived from the answers given by the persons selected for the sample are subject to error precisely because the entire population was not enumerated. In general, the size of that error depends on the number of persons included in the sample. The smaller the size of the sample, the larger the error. If the margin of error is +/- 4.5 percentage points, and 47 percent indicate that they will vote for BIDEN, 44 percent say they intend to vote for TRUMP, with the remainder responding UNDECIDED, anyone using those estimates cannot claim that there were more BIDEN votes than TRUMP votes because the difference between 47 and 44 is smaller than the margin of error and therefore is not statistically significant.
There are only two valid conclusions that can be drawn with confidence under these conditions. First, there is no difference between BIDEN votes and TRUMP votes. Second, TRUMP votes were not greater than BIDEN votes.
A serious problem of bias in the survey findings arises whenever the questions have been stated in a way that prompts a specific response. Take for instance this question: Do you think that 16 and 17 year-olds, many of whom have a license to drive, should have the right to vote? The “many of whom have a license to drive” may lead a respondent to answer YES because if they are mature enough to drive 16 and 17 year-olds should be smart enough to vote.
Other issues involving the respondents raise doubts about the accuracy of the election polling results. Do they understand the meaning of the questions being asked before they give their answers? Do they answer those questions truthfully? Do respondents change their minds between the time they were polled and the time comes for them to cast their vote?
Other issues involve the integrity of the enumerators and the persons handling the raw data. Do the enumerators record the responses accurately? What do they do when a respondent answers some of the questions but not all? Do they plug in an answer based on what other respondents have replied? Do they leave it blank? Do they fill in answers of their own when they cannot contact the persons included in the sample? This last practice is known among pollsters as data fabrication.
Are the raw data handled by one person with no one else following behind to validate that the data were handled accurately. Before the results are released to the public does anyone fact check the press release? And perhaps most important of all, are survey results bent in ways to please the party or organization that’s paying for the polling?
To comment on the polling results regarding political candidates or public issues, cable news programs often invite persons who are openly and directly associated with the Republican party, the Democrat party, or in the name of fairness and balance strategists from both parties are invited to comment at the same time. Too often those invited partisans spin the results in an attempt to persuade the American public that their candidate is better qualified than the other’s and their party’s position on public issues is right and the other party’s is wrong.
At times, their discourse becomes ugly with accusatory words like “racist,” “handmaiden,” and “liar” and loaded terms such as “climate denier,” “dog whistle,” and useful idiot.” When that happens, you may be feeling election coverage fatigue. Don’t get angry, get even. Switch channels or turn off your TV, then call together the members of your family around the kitchen table and decide how you intend to vote.
Edward J. O’Boyle is a Senior Research Associate with Mayo Research Institute. He has offices in West Monroe, Lake Charles and New Orleans. He can be reached at 381-4002 or email@example.com.