This article contains commentary which reflects the author's opinion
Get The Real News Delivered To Your Inbox
“And not only the pride of intellect, but the stupidity of intellect. And, above all, the dishonesty, yes, the dishonesty of intellect. Yes, indeed, the dishonesty and trickery of intellect.”― Leo Tolstoy
Four Simple Elements to Polling
With the arrival of the political polls for the 2020 presidential race, I see a lot of people on social media calling them fake. The nature of many of these comments suggest that many people do not understand how political polls work. This article explores how legitimate scientific polls are conducted, and investigates how political polls differ. Also, “models” predicting the Wuhan Coronavirus’s future are affecting our lives. So, I also briefly address those models.
Polling or sampling is often performed in scientific studies. These polls consists of four elements. First, the pollster must decide how much confidence they want in the poll. The standard for most scientific studies and political polls is 95%. That means the probability that the results are accurate is 95%.
Second, the researcher decides the level of precision. The most common measure is ± 3%. The level of precision means that if the researcher repeats the sampling multiple times, 95% of the time, they will get results within 3%. Polls are little use in discerning smaller differences.
Third, the confidence and precision levels determine the sample size. With a large population over 5,000 e.g., a 100 million registered voters, the above parameters require a sample of 1,067 people. That number assumes only one question is to be reported e.g., who will you vote for? Since many polls want to report results by gender or age etc., we generally see polls with several thousand people. Still this may seem like a small number: for additional sample size information see Sample Size, and I will also discuss it more, later in the article.
Finally, a random sample is drawn from the population of registered voters. Random means that everyone in the population has an equal chance of being selected. If everyone does not have an equal chance of being selected, the poll is biased. This criterion is critical and often presents the greatest challenge to researchers and pollsters.
A Proper Scientific Poll Defined
A hypothetical example of a scientific poll can illustrate how the process should work. Assume a state has 100,000 Wuhan virus patients admitted to its hospitals, and researchers want to know how many patients had a dry cough. That information is available on their admission records. With the admission records the researchers could count the cough cases, or they could get a fast estimate with random sampling.
Since the population is over 5,000, they would need 1,067 random admission records. If the researchers found 90% of randomly selected records indicated cough symptoms, they could conclude that 90% ± 3% of all 100,000 cases them. At least with a 95% likelihood that they were correct. This type of sampling, sample sizes, error rates, and confidence levels have been proven by statisticians. It is important to note that often researchers do not have the option of simply counting the entire population, it maybe unknown (like the number Wuhan Coronavirus cases in China).
The Media Loves Polls
The media’s use of polls to drive up their ratings can be illustrated by an example using Fox News. On April 22, 2020, Fox News published a political poll. It showed President Donald J. Trump trailing Sleepy Joe Biden by big margins in Michigan, Pennsylvania, and by a smaller margin in Florida. One of their motivations for their poll soon became clear.
For several days following the polls release Fox News held interviews with politicians, campaign managers, and other experts to discuss the poll. Even Tucker Carlson, a Fox News superstar, devoted time to the poll in a Tucker Carlson Tonight segment. After nearly a full week of daily coverage, Fox News proclaimed there was a catch to “the” poll. They recalled that in 2016, polls showed Hillary leading in in the same states by similar amounts right up to the election.
This was so even in states Donald Trump had ended up winning. Fox News knew this was a problem when they published the poll, and yet used it to generate hours of “news”. In other words, they knowingly manufactured fake news. This is symptomatic of FOX News’ new standards since new management took over after the exit of Roger Ailes.
Americans Do Not Trust the Media
We know the mainstream media is extremely bias. This was documented early on in President Trump’s administration by a Harvard study. Like much of academia, Harvard is well known as a bastion of liberalism. I have studied at Harvard, and witnessed their single mindedness firsthand. Yet they still published a study showing President Trump was being unfairly treated by the media in an extremely biased fashion.
Fox illustrated how polls create news. With the media’s documented bias and other factors, American’s trust of the news media is at an all-time low. The media’s reduced public standing motivates them to add credibility to their polls. They use universities for that purpose, the media covers the costs and the universities execute the studies, and get featured in the news cycle. Just like that, the media has a scientific poll.
Political Polls are Different
Political polls present unique challenges that make it is worth asking if it is even possible to do one. That is, every voter must have an equal chance of being polled. Some voters are easier to contact than others e.g., we have military personnel serving overseas. Even the most basic things can create critical problems, like contacting people by telephone land lines.
Many people no longer use those phones, and people work all hours in five different time zones. Even if they are just focused on one state, the challenge is daunting. Unless everyone in the chosen population has an equal chance of selection, it is not a random poll. If a poll is not random it is not valid. The pollsters rarely address that critical issue.
Pollsters know they are getting non-random samples, and often use that as an excuse to adjust their data. To illustrate, in a hypothetical poll, assume that the raw data shows President Trump leading by a few points above a statistical tie. Good luck selling that to the mainstream media, their viewers do not want to hear it. The media would ignore the poll, they would label it an outlier and likely forgo funding future polls from the university that produced it.
A Little Bit of “Magic”
To arrive at their desired results, pollsters go through the raw data and adjust it. Assume they see that the number of Democrats and Republicans are representative of the two parties. Of course they decide Democrats are more likely to vote than Republicans, so they trim 10% of the Republicans from the study. Next, they turn to college students and recent graduates, and decide they are underrepresented, so they inflate their numbers.
Through these hypothetical adjustments, Biden now has a small but statistically meaningful lead. They continue to adjust for minorities and every identity imaginable until they have the numbers they want. Perhaps they go back and trim 12% or 8% of the republicans to meet their desired outcome. Keep in mind these are small data sets, perhaps on an Excel spreadsheet, so the pollster can see the impact of their adjustments instantaneously.
No surprises with this process. In short, they start with highly questionable set of non-random numbers and adjust them to meet the story that the media wants to report. These maneuvers are both routine and dishonest. Even if the person making the adjustments has solid evidence that certain groups are more likely to vote than other groups, it is still dishonest.
Furthermore, while the media may be clueless about these procedures, the academics are not. They know that these practices are wrong. Nevertheless, if you read through the details accompanying some polls, they will describe their adjustments in detail while other polls remain silent. Pollsters are paid to create newsworthy results, specifically ones that make President Trump look vulnerable.
Do we Need Political Polls?
Political polls have a history of being wrong. The polls leading up to the 2016 elections provide excellent evidence of their bias. That year, seemingly every day a new poll showed then candidate Trump was going to lose. In many cases losing by large margins. Thousands of polls reported slightly different numbers, but virtually all of them drew the same conclusion – Donald J. Trump would never be President.
Its illogical to think they all made honest mistakes that produced the same dishonest result. As we do not know the 2020 election outcome, we are seeing a repetition of the last Presidential election’s polling. No one has announced that they figured out what went wrong. No new standards or corrections have been implemented.
The pollsters have done nothing to earn our trust. To paraphrase Brad Parscale, President Trump’s campaign manager; there are so many adjustments to those polls it’s not worth paying any attention to any of them, or the news they generate. The political polls are meaningless, and the hours the media spends talking about them even more so. Arguably, the polls are DNC propaganda generated by the media and aimed at undermining our President’s re-election bid.
The Rise of Prophets
With the Wuhan Coronavirus came models that supposedly predict the future. Even though models change frequently, our government uses them to justify taking away our freedoms. Often model changes are counter-intuitive or flat out unbelievable. An example is a recent “model” from the University of Washington, that shows disaster if we reopen the economy.
Scientific models are nothing more than one or more expert’s estimate of the future. The word “model” adds unjustified credibility, saying “the model shows” sounds better than “our best guess is.” The estimates are the sum of a set of assumptions about future events. For Coronavirus, for example, the events are infection rate for people who return to work, death rates, and the number of people who will return to work.
Young people will be more heavily impacted. Experts quantify the impact of their assumptions and add them to get their predictions. They should be better than everyday people’s, because they know factors others might not consider. On the other hand, little is known about the Wuhan Coronavirus and that hinders assumptions of any “expert”.
We know their models have always been wrong, or else they would not constantly change. To make matters worse, the doctors say things like the earlier model did not consider mitigation. Yet President Trump had already halted flights from China before the models were made. Did they think that we would do nothing?
Converting Predictions to Models
With an Excel spreadsheet (or other software) to calculate the impact of their quantified assumptions, scientists can quickly see the impact of their forecasts. Using software to add the assumptions is what changes an estimate to a model. Models are quantified assumptions, that can be easily changed. Anyone who works with Excel has probably created a lot of models.
The Wuhan virus models have been off by large amounts. The people who put forth models are relatively knowledgeable, but they are not prophets. With a multitude of possible outcomes, predictions are generally wrong, regardless of the topic. We must ask the question of how long do we want “models” to be used as an excuse for taking away our freedoms?