Please Stop Misrepresenting Statistics: Correlation is not causation.

Repeat after me: Correlation is not causation. No, really: Correlation is not causation. It may have something causal to it, or it may not. It might have something informative, or it may not.

Oh my goodness, I can barely stomach watching the news, or watching social media posts…the statistical fallacies are basically palpable.


Repeat after me: Correlation is not causation. No, really: Correlation is not causation. It may have something causal to it, or it may not. It might have something informative, or it may not.

A disparity in a statistic does not have causal power because mere statistical correlations are not inherently causal at all!

A true statistic does indicate something, but it is not obvious what that something is.

Case 1: 100% of serial killers drink water. Should we infer that water leads to murders, murderers, etc? Why not?

Case 2: If you and Jeff Bezos were in a waiting room for something, the average net worth for those in the at waiting room would be around 50 billion dollars, assuming you are worth around 0 billion. Should we infer that you do, or should have 50 billion dollars? No. But why not?

Case 3: 100% of the top cartel family is wealthy (this cartel derives its income from prostitution, violence, drug trafficking, etc).  Are we justified in concluding that a member in the top cartel derives money from nefarious means? If yes, then why?

Case 4: 30% of group A are incarcerated. This is 10% more than some other class B. Should we infer that there was a miscarriage of justice? If yes, then why?

To talk about these issues meaningfully, we have to distinguish two types of cause: Agent causes, and natural causes (non-agent causes). Agent causes have to do with the choices of free agents (humans with free will). The other type of cause does not have a direct bearing on our choices (at least not in any obvious, direct, and explanatorily rich way). If we make these distinctions, then we can easily tackle these four cases.

In case 1, the difference between serial killers and non-serial killers has relevant relation to whether they drink water (it has to do with a set of actions, pursuits, etc, of different agents, choosing different paths in life).

In case 2, the average wealth of 50 billion dollars has no bearing on your because wealth has nothing to averages. Rather, it would need further information about previous choices, investments, property, etc. Importantly, there is no obvious insight into mere averages, and we are guilty of hasty generalization if we try to sneak in any other ideas without justification.

In case 3, the wealth of a given cartel member is stipulated to be ill-gotten because of the historical provenance of wealth creation. In other words, the fact that money was ill-gotten has nothing do with the numbers and percentage, it has to do with the quality of free will actions on behalf of the agents. The numbers themselves provide no helpful information…because….mere statistical correlations are not necessarily causal, nor even explanatory, without further information.

In case 4, we are not justified in concluding anything about group A or group B at all because none of the relevant information is included. It says nothing about the choices, the laws, the process by which they were incarcerated. The idea that they should be equal in every way is astonishingly naive.

I have many ideas on why some get misled by statistics, but this is a long enough post. Let’s make the world less crazy with fewer fallacies. Please. Seriously. Please.

As a last-minute qualification, the wrinkle is that some correlations might indicate something causal. However, the issue here is that there is a good reason to not jump to conclusions. For instance, taking cyanide does lead to death. But this simple point here is that we are not justified in making an automatic jump between correlation and causation.

For other posts on voting, social media, logic, citizenship, and philosophy.

For a more in-depth look at statistical fallacies go here.

For a good video introduction to statistical fallacies go here, in the context of the social sciences, Jonathan Haidt (start towards the end if you only want the correlation/causation discussion).

Related Articles:

Other Articles from CommonPhilosopher:

Digiprove sealCopyright secured by Digiprove © 2019 Steve Byerly

Author: TheCommoner

I am a philosopher that is interested in what makes life worth living, what is worth pursuing, and how we can learn from the past. I believe that good philosophy benefits everyone and that there should be philosophers that present philosophy to those outside of the academy.