Debunking Epiphenomena
Checking the sequence of events can challenge a causal story, but earlier occurrence alone does not establish causation. Granger analysis asks whether past values of one variable improve forecasts of another, given the information included in the model.
THE IDEA
Debunking Epiphenomena
Checking the sequence of events can help challenge a proposed causal story. Clive Granger, whom I regard as a well-deserved “Nobel” in Economics, developed a way to test predictive relationships in time-series data. A “Granger cause” supplies information about later values of another variable beyond the past information already included in the model. That is more specific than merely observing which event comes first.
Studying changes rather than levels can sometimes help avoid misleading associations, but it is not a universal remedy. A finding that A helps predict B does not prove “A causes B”: common influences, omitted variables and the model’s assumptions still matter. Sequence can challenge a claim that reverses cause and effect, but it does not automatically warrant the conclusion “the statement that B causes A is wrong”. Absence of predictive evidence is not proof that no causal relationship exists.
The important difference between theory and practice lies precisely in the detection of the sequence of events and retaining the sequence in memory. If life is lived forward but remembered backward, as Kierkegaard observed, then books exacerbate this effect—our own memories, learning, and instinct have sequences in them. Someone standing today looking at events without having lived them would be inclined to develop illusions of causality, mostly from being mixed-up by the sequence of events. In real life, in spite of all the biases, we do not have the same number of asynchronies that appear to the student of history. Nasty history, full of lies, full of biases!
For one example of a trick for debunking causality: I am not even dead yet, but am already seeing distortions about my work. Authors theorize about some ancestry of my ideas, as if people read books then developed ideas, not wondering whether perhaps it is the other way around; people look for books that support their mental program. So one journalist (Anatole Kaletsky) saw the influence of Benoît Mandelbrot on my book Fooled by Randomness, published in 2001 when I did not know who Mandelbrot was. It is simple: the journalist noticed similarities of thought in one type of domain, and seniority of age, and immediately drew the false inference. He did not consider that like-minded people are inclined to hang together and that such intellectual similarity caused the relationship rather than the reverse. This makes me suspicious of the master-pupil relationships we read about in cultural history: about all the people that have been called my pupils have been my pupils because we were like-minded.
When to use it
Debunking claims that a new policy caused economic growth by checking whether the policy preceded the growth in time-series data.
Analyzing whether a training program caused improved employee performance by comparing performance changes before and after the program, while considering other factors.
When evaluating a treatment, check whether improvement preceded its administration and what other explanations are possible. Timing can challenge a causal story, but an individual before-and-after sequence does not establish treatment effectiveness.
FROM THEORY TO EVERYDAY LIFE
In practice
Mandelbrot Influence Misattribution
Journalist Anatole Kaletsky assumed Benoît Mandelbrot influenced Nassim Taleb's book Fooled by Randomness because of similarities in thought and Mandelbrot's seniority. However, Taleb did not know Mandelbrot at the time of writing; the intellectual similarity caused the relationship, not reverse influence.
A MOMENT TO REFLECT
Put it to work.
How can you use the sequence of events to check whether a commonly assumed cause-effect relationship is actually an epiphenomenon?
Recall a time you believed that A caused B based on correlation; how would applying Granger's method of analyzing changes and sequences change your conclusion?
In what ways does the backward-looking nature of history (as Kierkegaard noted) contribute to illusions of causality in your own learning or reading?
CONTINUE THE CONVERSATION
Source texts
- Antifragile (Nassim Nicholas Taleb)