Beware …Correlation is not enough


correlation_vs_causation

Correlation does not imply causation

Did you know that tomatoes are extremely deadly, and that this is a well-establish fact? What, you doubt that … OK, let me lay out the clear evidence. The facts are that everybody, every single one without exception, who ate one or more tomatoes between the years 1765 and 1812 is now dead.

Convinced?

No, of course not, and that is because you can see the not so very subtle but instead rather blatantly obvious flaw there. Yes, it is a 100% factual statement, and yes there is indeed a clear correlation there, all those who ate one or more tomatoes between the years 1765 and 1812 is now truly dead, but what is also rather obvious is that the eating of tomatoes had nothing to do with their demise. It’s a rather silly example that hopefully gets the point across that a correlation is simply not the same thing as causation.

Often when presented with claims that are apparently backed up with data, it is possible to be fooled with variations of this, especially if what is being presented is confirming a view you already hold. Here is evidence for X, look, it verifies that X is true … ah, but does it, or is it just a correlation, is there in fact a better simpler explanation for that same data, one that does not verify X at all.

Lets work a real example.

Claim: Percentage of Atheists in Jail is much smaller than percentage of Atheists in US population

Hey, I like that claim, it points out that the majority of people in jail are religious and that the percentage who are not religious is far smaller than the percentage of non-religious people outside jail. For this we can then make the leap to : non-believers are less prone to being criminals than believers.

There is some evidence …

  • here are some statistics from 1997 using data supplied by the Federal Bureau of Prisons that verified that only 0.2% of the US prison population were atheists.

Oooh … that’s a bit old, do we have anything more recent? Actually, yes we do …

From that we discover that 10.6% of those in jail are non-religious (atheist, agnostic, and non-religious), but we still have a huge data skew because 32% of the general US population are non-religious people.

So does that mean we can still stick with a claim that non-religious people are more moral? (You know that I as a non-believer really want that to be true, so I have a rather strong bias here that I need to be conscious about).

There are problems with that 2012 survey, it

  • simply asked prison chaplains what they thought the religious makeup of the prisoners were, so the results can indeed be challenged as being rather dubious for that reason alone, so we don’t really have an accurate measure here at all.

Lets take it a bit further … a thought experiment only .. lets pretend that we did have a very accurate measure of the religious and non-religious affiliations of those in jail, and that it demonstrated that the proportion of those in jail who were non-religious was far smaller than the general population. Could we then go, “Ah ha, there you go, the non-religious are indeed better people”.

Actually no you can’t, what you have is still a correlation and not really a causation at all. The problem is that such raw statistics, even if accurate, need not imply causation. Within prison culture there is a distinct advantage to those that are religious, so it is quite possible for those that are non-religious outside, to put on a religious mask when inside to glean some of the advantage it brings.

No, I’m not saying that this is in fact the case, but instead simply pointing out that at best, all we really have from any such statistics for beliefs in prison is at best a correlation and cannot truly establish any causation.

Bottom line: correlation does not imply causation.

Its easy to be fooled by a correlation … especially when the result is one you want.

Leave a Reply