Can correlation show causality?
Posted on: March 11, 2012
Correlation and causation are often misused and muddled up, especially by people who are not as familiar with statistics. This is because causality and correlations is not the same thing, and if two variables have a correlation it does not mean that they have causality. The mix-ups can result in variables being said to cause one another, when in fact they just have a similar relationship.
Correlation between two variables is a number that is used to describe the strength of the relationship (http://www.socialresearchmethods.net/kb/statcorr.php). It is a statistic that enables people to describe the direction, indicated by whether the value is positive or negative; and magnitude of a relationship, with 1.0 indicating a very strong consistency and 0.0 meaning a very weak relationship between the two variables. So for example would be the relationship between ice cream sales and sun cream sales, both of the variables increase and decrease around the same times in the year, so they have a correlation. Whereas causality is described as the relationship between two variables, when one variable causes another variable to change in some way, and researchers investigate to see if changes in the variable would result in another variable also changing (http://psychology.about.com/od/researchmethods/ss/expdesintro_4.htm).
Correlation does not signify causality; this is for a number of reasons. One of the problems is that with assuming correlations are equivalent to one variable causing another is that is impossible to say which variable causes another. A study was done that looked into the relationship between depression and the smoking of marijuana in adolescents (http://www.aboutjuveniledelinquents.com/AJD/which-came-first-marijuana-use-or-depression.htm). It was found that there is a correlation between the two; however it is incorrect to state that one caused the other. The study did not show that smoking marijuana resulted in depression, or that people who were suffering from depression smoked marijuana. The two variables could be caused by an entirely different factor, and only have a correlation because of it. In the same way that is would be incorrect to say that buying ice cream made you buy sun cream, or vice versa. Both ice cream and sun cream buying are influenced by the seasons and one does not cause the other.
So in conclusion, correlation does not signify causality however it may be useful for determining the cause of the variables or what causes them to change.
4 Responses to "Can correlation show causality?"
March 14, 2012 at 9:15 pm
You’ve stated the points really clearly 🙂
You’re right, it is sometimes very easy to find a correlation between two variables, however establishing causation is hard. For example, Boyatzis, Matillo & Nesbitt (1995) found that if children were exposed to the children’s TV show ‘Power-Rangers’ they were more likely to show overt aggression in comparison to those children that did not watch ‘Power-Rangers’. This study reflects a relationship between watching ‘Power-Rangers’ and aggression, however it cannot assume that ‘Power-Rangers’ cause aggression. This is reflective of the point you made above about how correlation is not necessarily causation. For example, the aggression found in the children could be due to parenting-styles, vicarious-learning or even the personality of the child and this could have led them to choose Power-Rangers as a appropriate TV show for them to watch. In this sense, the aggression could have caused the watching of Power-Rangers.
Also, when looking at correlations and causation, we must consider if we can ever find causation. In an attempt to find causation, would we need to conduct our study in a lab, where it would be unlikely for any extraneous variable to affect the outcome? Then, would the results from this be generalisable and have ecological validity?
Overall, I think you make a really good argument about causation and correlation, and I also think that correlation cannot show causation in all circumstances.
🙂
March 14, 2012 at 9:19 pm
A clear background on the topic of correlation and quite an easily understood way in which you describe it. One thing to consider is the third variable so correlation does not take into account outstanding variables which may increase or decrease the probability of a behaviour. So for example your ice cream example, the two things correlate but how about there being more ice cream vans around when its hot etc. These variables may also contribute to higher sales of ice cream. However i do agree with you that we cannot assume which variable causes the other and although correlation describes the relationship it does lack the ability to state which directs the other variable.
chocolateraisons 
March 14, 2012 at 4:31 pm
Very good blog, clearly outlined, gave enough detail but in a concise manner. Indeed Gardner(2000) said that correlation can be used to imply causation. The good thing about correlation is it can aid and direct further research but should never be confused with causation.
There is, unfortunately a problem in research where correlation is misinterpreted as causation Kassel(2000) describes a paper in which the author concludes that they show smoking causes stress when actually they show a correlation between the two.
References:
Gardner, 2000: http://psycnet.apa.org/journals/cap/41/1/10/
Kassal, 2000: http://psycnet.apa.org/psycinfo/2000-02781-014