Nicholas Cage and Swimming Pools

Here’s another story about statistics—this time, it might leave a U.S. Supreme Court Justice feeling a little uncomfortable.

Correlation does not mean causation. My statistics professor emphasized this principle during our third lecture, pausing for dramatic effect and enunciating every syllable as though delivering a sacred truth. I’d already encountered it as the Nicholas Cage and Swimming Pools Paradox, immortalized in a meme graph showing a bizarre correlation between the number of Cage’s movies and swimming pool drownings.

A few months ago, Professor Shafer from the University of Edinburgh visited us to discuss the Lucy Letby case. Letby received 14 life sentences in the UK for the murder of seven newborns and the attempted murder of six others. Her conviction rested on circumstantial evidence—chiefly, a shift roster overlaid with hospital mortality statistics.

Statisticians, including Professor Shafer, have raised the possibility that the court conflated correlation with causation. The fact that Letby was on duty during these deaths doesn’t necessarily mean she caused them.

It appears statisticians themselves can also fall into this trap.

In August 2020, four professors from different U.S. universities published a paper titled “How the Race of Doctors and Patients Affects Inequality in Neonatal Mortality.” The authors analyzed 1.8 million births in Florida from 1992 to 2015 and found that Black newborns had a 50% higher mortality rate when cared for by White nurses compared to Black nurses. This finding was particularly striking given that Black nurses make up just 5.2% of the workforce in the U.S.

The study gained significant traction, as it was published three months after George Floyd’s murder by a White police officer, which ignited the Black Lives Matter movement.

Its popularity even reached the U.S. Supreme Court. Justice Ketanji Brown Jackson cited the study in a dissenting opinion when the Court, in 2023, struck down affirmative action in university admissions, ruling that considering race in admissions decisions is unconstitutional.

Justice Jackson vehemently opposed the decision, penning a scathing dissent in which she accused her conservative colleagues of behaving like ostriches burying their heads in the sand. She argued that while the Constitution may not mention racial divides, such divides undeniably exist in society. One of her key examples of systemic discrimination was the study showing disparities in neonatal care.

But in September 2023, a follow-up study was published in the same journal. The new research revisited the same dataset but introduced additional factors into the analysis. It concluded that the earlier study was flawed: neonatal mortality is more closely linked to the infant’s weight than to the race of the caregiver. When weight was accounted for in the statistical model, racial disparities in mortality rates became negligible.

It’s too early to say whether the original paper will be retracted; the new study was published just two months ago.

Trusting statistics, it seems, is no straightforward matter.