You may have already heard this surprising statistic that might seem puzzling to the uninitiated: every summer, ice cream consumption and drowning rates rise in tandem, as if eating an ice cream cone could make you drown... This surprising "duality" often sparks jokes, but it mainly illustrates a classic pitfall of the human mind: confusing correlation with causation.
Heat, ice cream, and swimming
When temperatures rise persistently, our bodies seek to release excess heat. Two simple reflexes kick in:
- Cooling off with a treat: Ice cream or sorbet provide instant palate cooling and gustatory pleasure. Between 12 p.m. and 6 p.m., ice cream sales skyrocket at markets, beaches, and snack bars.
- Jumping into water: Swimming in the sea, rivers, or pools becomes a natural remedy against the heat. But the more people swim, the more opportunities there are for slips, fatigue, or encountering currents—and thus the risk of drowning accidents.
These two behaviors aren't linked by a cause-and-effect relationship: it's not the "zest" of frozen lemon that makes you lose your balance, nor the promise of a swim that triggers your craving for pistachio vanilla. They simply share the same driver: summer heat.
Why the correlation?
A confounding factor occurs when two variables evolve together under the influence of a third element. Here, the sun and rising temperatures play this role: they increase both the demand for frozen desserts and the number of swimmers exposed to aquatic risks.
This phenomenon is far from anecdotal: it reminds us that our brains automatically seek meaning in numerical coincidences, with a tendency to weave cause-and-effect links when observing two parallel trends.
Mathematical explanation
To measure the strength of a statistical relationship between two variables, we use Pearson's correlation coefficient, denoted as r. It's calculated as follows:
r = Cov(X, Y) / (σₓ · σᵧ)
For example:
-
X = ice cream consumption (in liters or tons)
-
Y = number of drownings (accident count)
-
Cov(X, Y) = covariance (measure of how X and Y vary together)
-
σₓ and
σᵧ = standard deviations of X and Y (measure of their individual dispersion)
The coefficient r ranges between –1 and +1:
- r close to +1 indicates a strong positive correlation (both variables increase together).
- r close to –1 signals a strong negative correlation (one increases, the other decreases).
- r close to 0 means no linear correlation.
In this case, we might have r > +0.80, showing a very strong positive correlation. But beware: correlation says nothing about causation. To prove that X causes Y, we'd need to isolate all other possible variables—which is impossible here, as the true common cause remains the heat.
Conclusion
The parallel between ice cream sales spikes and rising drownings is an excellent example of a "spurious correlation": two phenomena that occur together without one causing the other. It's the heat that drives both enjoying a sorbet and diving into water—with its risks. Understanding this helps maintain critical thinking when faced with graphs and statistics, while keeping the pleasure intact: enjoy your ice cream safely, and follow swimming guidelines to avoid accidents.