Recently, co-bloggers Scott Sumner and Kevin Corcoran engaged in a fascinating dialogue about causation, coincidence, and identity (check out Scott’s post here and Kevin’s posts here, here, and here). I’d like to throw my hat into the ring with some readings and reflections that might pique the interest of fellow thinkers.
Central to both Sumner’s and Corcoran’s arguments is the concept of coincidence: the occurrence of two events simultaneously without any clear causal link. Coincidences are not just intriguing anecdotes; they are surprisingly common. Consider these two lighthearted examples:
- On July 14, I bowled my best two games ever, scoring 161 and 157. Even my third game, at 118, surpassed my average of 110. Interestingly, it was also the first time in months that I had a $5 bill in my pocket. Could it be that the $5 bill was the secret ingredient to my newfound bowling prowess? Surely, it can’t be a mere coincidence!
- On July 3, the Boston Red Sox made a visit to Donald Trump at the White House. Following this meeting, they embarked on a remarkable winning streak, clinching ten consecutive victories and climbing from the bottom of the rankings to just a few games behind the division-leading Toronto Blue Jays. Did Trump sprinkle some magical pixie dust on the Sox? It can’t just be a coincidence!
Of course, these examples are playful in nature. Anyone asserting that a five-dollar bill or mere proximity to Trump caused these events would likely be met with skepticism. Indeed, counterexamples abound; the Washington Nationals visited Trump in 2019 after their World Series win, only to suffer a losing streak ever since. It seems improbable that Trump had any impact on either the Red Sox’s success or the Nationals’ misfortunes.
To discern causation from coincidence, a robust theoretical framework is essential. A well-tested theory acts as a lens through which we can differentiate between mere coincidences and genuine causal relationships. Poorly constructed theories lead to the blurring of lines between the two, resulting in misunderstanding.
However, it’s crucial to acknowledge that even rigorously tested theories can ultimately be proven incorrect. Take the miasma theory, for instance. It dominated medical thinking for centuries, backed by evidence that seemed to correlate bad air with the onset of disease. Yet, upon closer investigation, it was revealed that the “bad air” was often a byproduct of the very diseases it was thought to cause. John Snow proposed that specific illnesses stemmed not from miasma but from something else entirely (though he passed away before the germ theory was fully established). For those interested, I highly recommend The Ghost Map by Steven Johnson for a deeper dive into this topic.
Determining causation is a complex endeavor. Judea Pearl, a prominent statistician at UCLA, has authored several works that delve into causation from a statistical perspective. His technical book, Causality, is a challenging read, even for seasoned statisticians. For the rest of us, Pearl offers a more approachable text: The Book of Why, where he outlines the evolution of thought regarding causation and our current understanding. The takeaway? We are still grappling with the intricacies of determining when two phenomena are truly causal. Every model we create rests on certain assumptions, some of which may be quite tenuous.
This brings me to my concluding thought: the phrase “It can’t be a coincidence!” ranks among the least scientific assertions one can make. Not only is it frequently uttered by conspiracy theorists and misguided thinkers trying to validate their flimsy notions, but it also embodies a level of certainty that is fundamentally unattainable. Coincidences occur all the time, and the probability that a causal link is coincidental should always be acknowledged. Even claims of statistical significance, such as “P<0.05,” hinge on probabilities that are influenced by the aforementioned assumptions. Those who assert certainty often do so because they lack the theoretical grounding and evidence to substantiate their claims.
In light of the assumptions necessary to establish causation, we ought to embrace a degree of humility and admit, “It’s possible I am wrong.”
