Correlation vs. Causation
Faulty Causation Fallacy
Correlation vs. Causation
Non Causa Pro Causa
Determining the nature of causation is very difficult. Sometimes a cause and effect are closely related - spatially, temporally or both - but sometimes they are not. However, humans seem to be inclined to assume that events which are closely connected either spatially or temporally are also connected causally.
This problem is commonly known as the difference between correlation and causation. Just because two events correlate (are close in time or space) does not mean that one has caused the other. The Latin term for such an error is called "non causa pro causa," which means "non-cause for the cause." It is important to try and break ourselves of this habit and become more critical of our natural inclinations in such cases.
There are a number of different ways in which correlation and causation can become confused. One is called the "Neglect of Common Cause." Also sometimes called "Joint Effect," this occurs when someone assumes that one event caused another when, in fact, they are both really effects of some third event. This third event is the "common cause" of the other two. For example:
1. Every time I eat chocolate, it gives me acne.
The speaker above observes a strong correlation between eating chocolate and suffering from acne, drawing the conclusion that the former causes the latter. What is ignored, however, is the possibility that both are caused by something else - perhaps this person suffers from anxiety and stress. The stress causes him to eat chocolate, but at the same time causes acne to break out. This will lead to the two occurring very close in time, even though one isn't actually causing the other.
Many times a fallacy like the one in the statement above occurs in assertions based upon statistical evidence, for example:
2. Recent studies have proven that watching too much violence on television leads to people being violent in real life.
Although it may indeed be true that viewing acts of violence can make a person more susceptible to committing violence, the above statement ignores crucial factors which might have a causal influence on both. For example, both increasing violence on television and increasing violence in society might be caused by changing economic circumstances or something else entirely. Thus, the above cannot be regarded as a sound position until other such possible common causes have been ruled out first (which, incidentally, may have indeed been done as part of the study, but this would have to be made explicit for the statement to be valid).
Here is another, more amusing instance where such an error can be made:
3. Researchers at the Aabo Akademi found that Finns who speak the language of their Nordic neighbors were up to 25 percent less likely to fall ill than those who do not.
Should we conclude from this that learning how to speak Swedish will help improve our health? Or perhaps that it is only Finns who can receive health benefits from speaking Norwegian? Nonsense - there is nothing about the Swedish language which can make a person healthier. What we need to look for are common causes of both being multilingual and of having better health - at least in Finland.
The neglect of Common Cause can be found in many political debates as well:
4. Most drug use occurs among the poor - this is because poverty causes people to engage in risky behavior, like abusing narcotics.
Now, perhaps the above position is true or at least has a grain of truth in it, but it fails to address the possibility that a third factor is responsible for both of those listed. One might be able to effectively argue that racial discrimination makes a person feel hopeless - this, in turn, may not only increase the chances of that person being poor, but might also incline them to seek escape in drugs.
This fallacy has also found a home in quite a few religious debates:
5. Morality in this nation has worsened at the same time that adherence to traditional Christian beliefs has declined. Obviously, the latter has caused the former, so encouraging Christianity will ensure a return to traditional moral standards.
In the above example, it is assumed that the correlation between dropping standards of morality and weakening adherence to traditional Christianity means that the latter is the cause of the former. This position ignores, however, the possibility of some third event being the cause for both. Thus, for example, it may be that growing diversity in society has weakened the bonds of all traditional institutions - including both religion and moral standards. Simplistic explanations like the above make it easy to propose simplistic solutions, but they cannot be accepted until alternatives like possible common causes have been addressed.
Science, Correlation, and Causation
At this point, one might wonder what the difference is between the fallacy of correlation vs. causation and the normal process of science. After all, isn't much of science a matter of correlating various observations and constructing theories about why they occur? There is some justice in asking this question - but nevertheless, there are important differences. Let's consider this example:
6. When the sun is visible, we have daylight. When the sun is gone, we don't have daylight. Therefore, the sun is responsible for daylight.
Isn't this an example of confusing correlation and causation? Why can't we say:
7. Daylight is caused by the luminescent radiance of Apollo. It just seems like daylight is caused by the sun because Apollo habitually travels alongside it. But Apollo is the real cause of light.
Well, we certainly could say that - but what's the difference between examples #6 and #7? What can make #6 justified rather than a fallacy? There are a couple of key characteristics - the first and perhaps the most important of which is that of testable predictions.
A theory is scientific if we can use it to make testable predictions about other observations we might be able to make. Are there testable predictions we can derive from example #6? Yes: when a place my hand between my eyes and the sun, the light is blocked; when clouds move between the sun and an object, it is shrouded in darkness; and when the moon moves between our planet and the sun, there is a cone of darkness on the planet.
All of these and more predictions can be made, resulting in observations that are consistent with the idea that the sun is responsible for daylight. Are there any predictions that we can derive from example #7? No - and if we tried, we wouldn't be able to reliably test them.
This points us to a second important issue that helps us separate a valid from an invalid causal connection: do we have any viable alternatives? If #7 were genuinely and successfully tested, then we would have much less justification for confidently asserting #6. The fact of the matter is, though, that there aren't any viable alternatives to #6.
So how can we be absolutely certain that an correlation between two events indicates a causal relationship? We can't, actually - the knowledge provided by the scientific method is never absolutely certain. Science forces us to remain open to the possibility that new evidence will cause a change in what we know and believe. Science doesn't allow us to become complacent, assuming that we already know everything.
Science is, however, quite reliable. With enough information, we can justify concluding that a strong correlation between two events points to a causal relationship. When all reliable evidence points to one conclusion while no reliable evidence points to anything else, then we don't commit the fallacy of confusing correlation with causation by concluding that we have likely identified the cause of the phenomenon in question.-->