Image by Author

Statistics should clarify the world, but sometimes numbers reveal something deeply counterintuitive. These five paradoxes show how data can reverse expectations, how probability intuitions fail, and why careful analysis matters.

1. Simpson’s Paradox: When Groups Tell the Opposite Story

A treatment can work better for men and women separately, yet appear worse when combined. This reversal happens when a hidden variable creates unbalanced groups.

The 1973 UC Berkeley graduate admissions data seemed to show gender bias: the university admitted 44% of male applicants but only 35% of female applicants. But when researchers examined each department individually, they found women had equal or higher acceptance rates than men. Why? Women applied disproportionately to departments like English and comparative literature, which had more applicants competing for fewer spots. Men applied more often to engineering and chemistry departments, which had lower competition and higher admission rates overall.

Simpson’s Paradox shows that aggregated data can deceive when lurking variables matter.

2. The False Positive Paradox: Why Accurate Tests Are Often Wrong

A medical test that correctly identifies 99% of sick people and 99% of healthy people sounds reliable, but for rare diseases, most positive results are false alarms.

Here’s why: imagine screening 10,000 people for a disease that affects 1% of the population. That means 100 people actually have the disease, and 9,900 are healthy.

The test correctly identifies 99 of the 100 sick people, but it also incorrectly flags 99 healthy people as sick. So if you test positive, you’re one of 198 positive results (99 true + 99 false). Your actual chance of having the disease? Only 50%.

The rarer the disease, the worse this gets. For a disease affecting 0.1% of people, that same test means a positive result gives you less than a 10% chance of actually being sick. The problem isn’t the test — it’s that when the disease is rare, there are so many more healthy people to generate false alarms.

3. The Friendship Paradox: Your Friends Are More Popular Than You

On average, your friends have more friends than you do. This feels impossible, but it’s mathematically guaranteed in nearly every social network.

Think about how friendship networks work. Popular people with many friends appear on lots of friend lists, while less popular people appear on just a few. The more friends someone has, the more likely they ended up on your list in the first place.

Say you have 10 friends. Some might have 50 friends, some might have 5. When you average those numbers, the people with 50 friends pull the average up. And those highly connected people? They’re showing up in many people’s friend counts, not just yours. As long as there’s any variation in popularity, your friends’ average friend count will mathematically exceed yours.

This isn’t just a curiosity. It explains why social media can feel so deflating. Your followers really do have more followers than you, on average. It’s not you; it’s math.

4. The Inspection Paradox: Why Your Bus Is Always Late

Arrive at a bus stop at random, and you’ll wait longer than half the average interval between buses. The gap you experience is systematically longer than the average.

Here’s why: imagine buses run every 10 minutes on average, but the schedule varies. Sometimes there’s a 5-minute gap, sometimes 15 minutes. You’d think arriving randomly means you’d wait 5 minutes on average (half of 10). But you don’t.

You’re more likely to arrive during the longer gaps. A 15-minute gap gives you three times as many moments to show up compared to a 5-minute gap. So you end up in the longer gaps more often, and your average wait stretches beyond 5 minutes.

Whenever there’s variation in the schedule, your average wait will always exceed half the average interval. This is why public transit feels unreliable even when the published average is accurate. You’re not imagining it. The gaps you personally experience really are longer than the average gap between buses.

5. Berkson’s Paradox: Why Hospital Data Misleads

Two completely independent conditions can appear correlated when you only observe hospitalized patients. Selection creates phantom relationships.

Berkson discovered this in 1946 while studying hospital patients. He found that diabetes and cholecystitis (gallbladder disease) appeared negatively correlated among hospitalized patients, even though they were completely independent in the general population.

Here’s what happened: hospital admission acted as a filter. If either condition could land you in the hospital, then among hospitalized patients, having one condition made the other less likely. If someone is hospitalized for diabetes, they might not have cholecystitis. If someone is hospitalized for cholecystitis, they might not have diabetes. But plenty of people with neither condition aren’t in the hospital at all, so you never count them.

The correlation Berkson saw was real in his sample, but it was an artifact of selection. This same bias shows up whenever you’re studying a pre-selected group rather than the general population. The relationship you observe might exist only because of how people ended up in your dataset.

Conclusion

These paradoxes share a common lesson: data alone doesn’t tell the complete story. Simpson’s and Berkson’s Paradoxes show how grouping and selection distort relationships. The False Positive and Inspection Paradoxes reveal how base rates and sampling mislead intuition. The Friendship Paradox shows that network structure creates unavoidable biases.

Understanding these changes how you evaluate evidence. The next time data shows something surprising, ask: what’s the hidden variable? How was the sample selected? What’s the base rate? The numbers might not be lying, but they might be answering a different question.