#6. Of Ice Cream & Tylenol
- Barry Markovsky
- Oct 22
- 6 min read
Updated: Nov 20
I love ice cream. But for the sake of our collective health, maybe I should give it up. It’s not just because the main ingredients of our beloved frozen indulgence are known to contribute to obesity, type 2 diabetes, and heart disease. No, it’s something a lot more urgent and insidious:
Sharks!
Scientists have discovered a strong statistical link between ice cream sales and shark attacks on humans along coastal waters. For example, this graph appeared in an article by Dr. David D. Leonard, a Primary care physician with Family Physicians of Old Town Fairfax (VA).
This is real, people. Put down your spoons and exorcise the demon treat from your freezers. Same for its evil step-siblings: frozen yogurt, gelato, sorbet, sherbet, and their skinnier cousins.
Of course, this is a joke. But speaking statistically, it’s not.
Ice cream sales and shark attacks are strongly associated. It’s because both vary with air temperature. The hotter it is outside, the more people eat ice cream, and the more sharks eat beachgoers.
Here I offer a tiny, number-free statistics lesson to add to your critical thinking toolbelt. Obviously, eating ice cream doesn’t cause shark attacks. Their correlation is false, or spurious in the statistical jargon. Air temperature is considered a confound (pronounced CON-found) because it confounds (pronounced con-FOUNDS) the correlation between shark attacks and ice cream consumption. There’s certainly no direct link. At best, eating less ice cream could reduce your waistline, but it won’t reduce shark attacks.
Now that you know about spurious correlations, you may question all kinds of popular assumptions in our cultural zeitgeist. Does exposure to damp weather cause colds and flu? Do stress and spicy foods cause ulcers? Does sitting too close to the TV wreck your eyes? Does candy give kids the zoomies?
These claims, and many more, have proven to be spurious.
Have you heard the expression, “correlation does not imply causation”? I’m happy to see that it’s gaining some cultural traction. This simple message can improve your life, and possibly save it.
When it comes to explaining silly correlations, spuriousness and confounding factors aren’t the only possibilities. Sometimes it’s only a coincidence. Examples to come later.
Tylenol in the News
Robert F. Kennedy, Jr. (RFKJ) is the undisputed reigning champion of spurious correlations. The U.S. Secretary of Health and Human Services has issued warnings and policy directives that will absolutely cause unwarranted suffering and death. The only question is how much.
[Rant warning.] RFKJ has degrees in law, literature, and history. He has no formal training in science, medicine, or public health. It shows. Despite claiming to be evidence-based, his approach is intuitive and obviously politically driven. I fear the worst. Time and again, he’s proven himself to be unqualified to evaluate research. His approach involves cherry-picking studies that can be spun to appear supportive of his agenda. Truth be damned.
One of RFKJ’s most fundamental problems is that, for him, correlation does imply causation, at least when it suits his agenda.
Let’s talk Tylenol. More generally, this is about the generic drug acetaminophen and not the gigantic Tylenol brand. It’s been in the news a lot lately. Under RFKJ’s stewardship, our federal government has been blasting out health warnings about acetaminophen with the urgency of tornado warnings in Kansas. I’m thinking especially of RFKJ’s recent pronouncement, full-throatedly endorsed by no less of a medical authority than the President of the United States. In a recent press conference, the POTUS and RFKJ lectured America about a very concerning link between pregnant people ingesting Tylenol and their children being diagnosed with autism.
With rhetoric like, “If you’re pregnant, DON’T TAKE TYLENOL!” and policy actions surrounding this warning, it’s crystal clear what they wanted the public to believe: Tylenol, and its generics and store brands, cause autism.
Tornado warnings exist because tornadoes are proven to cause death and destruction. The evidence behind the Tylenol warnings is far less well established. In fact, it’s the opposite. It’s not that research has never found a correlation. It’s that the correlation tends to disappear in more stringent tests.
Better tests require large samples, numerous measures of potential spurious factors, and other statistical controls. Importantly, one of the most recent tests was probably the most definitive.
The project included over 2.4 million children in Sweden. Using the technical scientific jargon, this is known as “a ginormous study.” One component examined sibling pairs where only one was exposed to Tylenol during their mother’s pregnancy. Researchers found no difference in the incidence of autism between the exposed and non-exposed siblings. This was important because siblings share genetics, environmental factors, and family health practices, all of which can contribute to spurious correlations. Non-exposed siblings formed a natural control group.
Correlation and Cause
Finding correlations is easy. Establishing causes is hard, especially if you can’t administer treatments and placebos in clinical trials. The only alternative is the nearly impossible task of identifying and statistically controlling for every possible spurious factor.
Suppose you want to chew on the hypothesis that ice cream consumption causes shark attacks. Your statistical method can control for lots of possible factors, which will let you see whether ice cream still correlates despite eliminating the potential confounds. Your list includes, say, jaw size (the shark’s), age of victim, latitude & longitude of the attack, and exposure to the movie “Jaws.” Fine, but if you fail to control for air temperature, you’ll still find the same old spurious correlation, but this time feel that you’ve now got even stronger case that it’s the ice cream’s fault.
Dr. Mikhail Varshavski is a well-credentialed family physician and stellar science communicator. In contrast to RFK and DJT, he has a sound grasp of scientific research methods and the medical literatures pertinent to the topics he discusses on his YouTube channel, “Dr. Mike.” Responding to the RFKJ-Trump press conference, he took a critical look at their claims and the research they cited to support them. I quote the following from Dr. Mike’s conclusion.
When you nix nuance, and mix that messaging with partisan politics, people lose trust in us, actual doctors. It also pushes some to avoid Tylenol when they really need it. Will people actually get hurt as a result of [RFKJ’s and Trump’s] press conference? Sure. I actually don’t think that’s unreasonable to expect.
Tylenol has proven to be a safe and effective way to control fever, for the pregnant and non-pregnant alike. Not only will it reduce discomfort for pregnant women, but good research shows that fever can do serious harm to fetuses. Dr. Mike touched on why our administration might ignore such research and knowingly promote harming mothers and fetuses:
Truly, this is part of a bigger pattern. RFK and other politicians have built their careers on overstating scientific uncertainty, especially with autism. First vaccines, now Tylenol. Same playbook. Highlight just enough data to sound convincing, then exaggerate the conclusions to look like he’s saving the day, even if it means twisting the science. It plays well to score political points, but it’s a disaster for pregnant patients who need honest, balanced guidance. It’s even a bigger disaster for us who want to find the actual reason for increased autism rates.
He alluded to a study published in 1998 by Andrew Wakefield and colleagues linking childhood MMR vaccines to autism. It has since been thoroughly debunked and withdrawn from the journal—a rare occurrence, but warranted in this case, where it was uncovered that the Wakeful study was fraudulent. Unfortunately, the article helped spark the anti-vax social movement which, among other problems, has led directly to dangerous measles outbreaks in geographical areas where anti-vaccine sentiments run highest.
Not to detract from the seriousness of the issues, I’ll leave you with some random facts—emphasis on “random.” These are actual pairs of highly correlated variables measured annually. They were culled by automated online searches of publicly available databases. The vast majority of correlations tested yielded nothing interesting, but just by chance, a few fun ones popped out.
1. Per capita margarine consumption; Maine’s divorce rate.
2. Deaths by venomous spiders; winning word length at the national spelling bee.
3. Cheese consumption; death by bed sheet suffocation.
4. Swimming pool drownings; new Nicholas Cage films.
5. Age of Miss America; murders by steam, hot vapors, and hot objects.
Could any of these claims be legitimate? No doubt we can get RFKJ’s staff to look into them.
Wonderful explanation of spurious correlations and confounding variables. Great humor too. Good read. Thank you.