A cause is commonly defined as “something that brings about an effect or a result.” But what does “brings about” mean? Isn’t that just a synonym for “cause”? Maybe a cause is just when one event always (or usually?) is associated with another. Or maybe a cause is some invisible property of the first event, which is somehow transmitted to the second. The 20th-century philosopher Bertrand Russell, gave up any effort to define causation, calling it a “relic of a bygone age.”
Perhaps a working definition is that a cause is an event which, 1) when it occurs, usually leads to a particular consequence. 2). When the cause does not occur, then neither does the consequence and 3) there is an evidence-based theory that links the cause with its consequence. With such a definition, causation can sometimes be objectively tested with experiments.
“Correlation Is Not Causation”
While causation is a qualitative feature of our universe, correlation is purely quantitative. It’s a number that measures the extent to which two events tend to occur together, in a way that cannot be explained just by chance. Of course, just because two things happen to occur at the same time does not mean that one caused the other. The correlation could be spurious, i.e., without any aspect of causality. Consider this helpful chart:
A good example of a spurious correlation, unless you have an evidence-based theory that could explain tech spending and suicides.SCREENSHOT FROM HTTP://TYLERVIGEN.COM/SPURIOUS-CORRELATIONS RETRIEVED SEPTEMBER 17, 2018
There are two ways that a correlation between two events can be spurious. First, the two events could be caused by a third. For example, increased equality may be associated with a stronger economy. One could imagine a theory that that increased rewards to “job creators” are incentivizing them to invest and grow the economy. Or perhaps more likely: the increased inequality and the stronger economy are both a common response to lower tax rates, especially on the rich:
Another way that correlation can be confused with causation is when there’s a confounding variable, that is, a completely different, true cause. Let’s say that sunspot activity increases solar radiation and is associated with higher global temperatures. The science is that human activity has a much, much greater impact on Earth temperatures. Human activity is the confounding variable in this case.
Experiments Can Establish Causality
So how can we be reasonably confident that one thing causes another? Ideally, we can run an experiment. For example, we can change the default 401(k) deferral policy so that new employees automatically participate, unless they opt-out, instead of making the default that they don’t participate (opt-in). If the participation rate increases (and it usually does), and nothing else relevant has changed, then we can conclude from such an observational study that the policy change caused the change in deferral rates.
Nile blue hydro chloride in water at daylight and different pH-values. From left to right: pH 0, pH 4, pH 7, pH 10, pH 14. Image: Armin Kübelbeck.
Sometimes, experiments are not feasible (or ethical.) In these cases, we might be able to find a “natural experiment”. For example, maybe we can find two very similar companies, one with an opt-out policy and one with an opt-in policy. If the only relevant difference is the policy, we’d have evidence that it’s the policy that caused any observed difference in deferral rates.
A rigorous understanding of causation and its application is critical to good decision-making in business, investing and personal life because deciding is all about forecasting the outcomes of different choices, each of which will cause different future events. Careful experiments can help us identify causes and make predictions even if the true nature of what a cause really is can be elusive. Now, if you’ll excuse me, it’s time for me to get back to my video game.