Why Statistics are Bullshit
Wow, this is all very confusing and contradictory!
Hopefully you get the point we’re trying to make. Statistics are often regarded as the ultimate truth, when in fact, most statistics are bullshit. We’re using strong language for sure, but we really want to drive home out point.
Statistics can be used and misused in any number of ways by a person or entity to bolster their argument, to get clicks, to win over voters, to make their product seem more effective than it really is.
Why Can’t Statistics be Trusted?
Let’s look at some of the reasons why statistics can’t be trusted.
1. Sampling Bias
Sampling bias occurs when the selected sample group does not accurately represent the entire population. That leads to researchers inaccurately assuming that a large group has the same traits as the smaller sample size.
Perhaps the most famous example of sampling bias occurred during the United States’ presidential election of 1948. On election night, telephone surveys were conducted by the newspaper The Chicago Tribune so they could attempt to predict before anyone else who would win the race between Thomas Dewey and Harry Truman.
Unfortunately, the pollsters didn’t consider the fact that telephone technology was relatively new and expensive. The vast majority of voters who owned telephones were likely to be wealthy.
Wealthy voters were more likely to vote for Dewey, while lower-class voters, who did not own telephones, were mostly Truman supporters. Therefore, anyone who took part in The Tribune’s poll was much more likely to be wealthy and a Dewey supporter. The opinions of the Truman supporters were not captured.
Unaware of their sampling bias, the Chicago Tribune thought Dewey would win in a landslide and ran the headline “Dewey Defeats Truman” before the election results were finalized. Of course, that wasn’t correct, and the newspaper was humiliated.
2. Small Sample Size
Sometimes politicians, companies, and other institutions use small sample sizes to quickly bolster their point without having to conduct more thorough research.
Very small sample sizes can be used to make insignificant differences seem much more important. For example, a company that’s promoting a new weight loss drug may tout a statistic in their headlines: 75% of users lost weight with our product!
That sounds impressive at first glance. But what if you then learn the study only included eight people and six of them lost weight? Then, the results don’t sound too impressive. Those six people could have lost weight for reasons that had nothing to do with the drug.
Interestingly, there are times when small sample sizes are called for. Doctors who are testing experimental drugs will often choose to conduct trials with just a handful of patients. If those trials prove successful, then the doctor may increase the sample size. However, a responsible doctor won’t publish the results of the early trials with a small sample size. Again, a “responsible doctor” won’t publish the early trial results, but we all know how things really work.
3. Cherry Picking
Cherry picking refers to choosing data or statistics that support a pre-determined narrative while suppressing other data that may contradict the preferred narrative.
One of the most infamous examples of cherry-picking in the history of the United States was the research conducted by the tobacco industry about the link between cigarette smoking and lung cancer.
In 1964, Philip Morris’s public statement was “We don’t accept the idea that there are harmful agents in tobacco.” They backed up this claim with cherry-picked data from their own research.
They went on to deny the addictiveness of cigarettes and the harmful effects of second-hand smoke. Of course the tobacco industry’s lies were eventually exposed.
Here’s another example of cherry-picked statistics. You probably have a hand sanitizer or hand soap in your home right now that claims to “kill 99.99% of bacteria.” That sounds great, right?
What the product’s manufacturer doesn’t tell you is that they only tested the product against specific types of bacteria in laboratory settings. That means the product is 99.99% effective in those exact settings. There’s no guarantee that the product will be as effective in your home or on your hands.
And let’s look back at the weight-loss drug example. Assume the manufacturer did distribute their drug to a statistically significant sized group of 1,000 people. Let’s further assume that after 12 weeks on the drug, 800 participants lost an average of 1.4 pounds.
Technically speaking, 80% of people lost weight while on the drug. That’s the statistic that the company will use as the basis of its marketing campaign. However, they won’t mention how little weight was lost, or how long it took the participants to lose that weight. In effect, they are cherry picking the information they want to share.
4. Correlation vs. Causation
Another phrase you’ll often hear around statistics is “correlation vs. causation.” This means that correlation between two sets of data does not always mean that one data set influenced or caused the other data set.
For example, you could find a statistic that says in the months when ice cream sales increased, so did shark attacks. Does that mean sales of ice cream caused sharks to attack people? Of course not. However, ice cream sales are more likely to increase in the summer months. That’s also when people are more likely to go to the beach, which means their chances of being attacked by a shark are higher.
Here’s another amusing real-world example. In 1980 researcher Robert Matthews discovered a high correlation between the number of storks that lived in various European countries and the number of human babies being born in those countries. The more storks, the more babies were born each year. He chose to look at these two statistics to play on the myth that storks deliver babies.
Obviously, storks don’t deliver babies, but Matthews wrote about a third variable: the geographic size of the country. Large countries have larger populations, which logically leads to higher birth rates among humans. Larger European countries are more likely to have natural spaces where storks like to breed and nest, hence more storks.
5. Average vs. Median
You must ask yourself if you’re looking at the median of a set of numbers or the average of those numbers. Those terms can sometimes be used interchangeably but they are not the same.
The median number is the number right in the middle of the set. Half of the numbers are lower than the median number and half of the numbers are higher than the median number.
The average number, on the other hand, is the total of each number added up, then divided by the number of items in the list.
Take a look at these 7 annual incomes of Americans:
With these numbers, someone could report that the median American’s salary is $103,000. Or someone could report that the average American’s salary is closer to $350,000.
There is quite a difference between those two numbers and each tells a different story. Based on which number is reported, you may feel that financially speaking, you’re doing better than most Americans. You could also feel that you’re worse off if the higher number is reported.
Statistics Can Be Deceiving
“There are three kinds of lies: Lies, damned lies, and statistics.”
Which one of the two men actually uttered the famous quote doesn’t really matter, because the meaning behind it rings truer today than ever before.
Statistics can be manipulated in a million different ways. Anyone can use a statistic to sell a product, persuade others to agree with their point-of-view, or to forward their political agenda.
You should be suspicious of any statistic you read or hear about. Consider who’s providing the statistic and what the real motivation might be for publishing it.
One Statistic You Can Trust
Despite everything we’ve just said about statistics, we’d like to offer at least one statistic you can trust: Empire Resume has a 100% satisfaction guarantee!
Our professional resume writing service experts will work with you to present you in the best possible light and significantly increase your chances of success in landing that interview.
If you don’t receive an interview within 30 days after following our job-search guidance, we’ll revise your resume for free. If you still don’t land an interview within 60 days, we’ll revise your resume again at no charge. In fact, we will keep working with you until you land an interview.
We’re standing by ready help you. Contact us today at 800-690-4085 or email@example.com.
Dr. Phillip Gold is President/CEO of Empire Resume and has vast experience writing resumes for both professionals and servicemembers transitioning from the military into civilian roles. He served as a Captain in the U.S. Air Force and was responsible for leading nuclear missile security. Phillip is a Certified Professional Resume Writer and holds a BA in Communications from The Ohio State University, an MS in Instructional Technology, an MBA in Finance, and a PhD in Finance.