Logical Fallacy of the Day: cum hoc ergo propter hoc

One of the most oft encountered logical fallacies, particularly when quoting the outcome of studies or statistics, is that of confusing correlation of variables with causation; cum hoc ergo propter hoc (latin, with this, therefore, because of this). This fallacy is the root of irrational beliefs, fears and biases in otherwise intelligent and informed people.

This came to light earlier this week in a forum where some people who would generally not be considered bigoted where having a “truthful” discussion about whay they thought of black people. Statistics, such as the rate of black men being incarcerated vs. that for white men, were quoted and the crowd all pointed saying, “See, they’re more likely to go to jail; the stastics don’t lie, they’re just bad people.” This was from a bunch of people who don’t want to be racist, but they found it troubling to agree with the statistics, which are true, agree with their own experiences (bad neighborhoods == black neighborhoods) and yet not see blacks as being inherently predisposed to crime.

The root of their misunderstanding lies in the logical fallacy of assuming that correlation proves causation. The most important thing to realize when reading studies or statistics is that the correlation of two variables does not mean that they are anything but related. I recently read of a study done by the US Army which found that the best predictor of whether or not someone would ever be involved in a motorcycle accident was the number of tattoos they had. Does that mean that tatoos cause motorcycle accidents or cause poeple to be stupid? No. All it means is that people with lots of tattoos tend to have motorcycles and ride them more than those without. Perhaps the root cause is that they’re more inclined to take risk, but quite obviously, tattoos don’t cause motorcycle riders to crash.

To get a better understanding of how to interpret studies, I highly recommend reading Freakonomics by Steven Levitt. His book examines a number of different situations and a huge amount of data using techniques like regression analysis to find the root causes in otherwise random data. For the race problem, it basically boils down to the fact that criminal tendencies are not well correlated with race but rather with low education and low income. Drawing on that, it’s quite easy to understand how blacks are still associated with being criminals and easy to see just how long it is going to take for the remanants of a racist society to be removed from our civilization.