Can correlation show causality?

Although correlation and causality sound similar, they are in fact very different things, and not intrinsically linked.  Variables can be correlated (positively or negatively), without one being responsible for the change in the other.

Correlations implies a relationship between the two, but this is different to causation.  For example, ice cream sales and aggression may both increase during summer – but this does not mean that the more ice cream people eat the more aggressive they are likely to be.  However, the increase (hopefully!) of heat could be responsible for causing both – and if a correlation were found between increased ice cream sales and the increased heat, this could be an indicator of causation, although not proof of.  This is not 100%, but can be used as a clue towards where to investigate, but this is pretty much all that correlation can conclusively be used for.



12 thoughts on “Can correlation show causality?

  1. Assuming causation from correlation is one mistake which is often made in psychological research. Concluding that one thing causes another without sufficient evidence is a definite no no and can cause research to be invalid. Anyone can pick two things that are correlated with one another (either positively or negatively) and state that one causes another but what we need is an explanation to back up the point. We need to know how and why this conclusion was made.
    The church of the flying spaghetti monster is a parody religion which theorised that the increase in global warming was due to the significant decline in pirates. Their aim was to prove that any two things that show the same correlation can be connected to one another and assumptions can be made. Scientific research shows that the reasons behind global warming are not related to the decline in pirates so we can rule that out as a cause. But with this in mind we can see how two things can be said to be linked when often they aren’t even linked at all.
    With regards to assumptions made in psychological research it’s often the case that one characteristic is caid to be the cause of another characteristic when it could in fact be the other way round. For example, upon conducting research a psychologist may come to the decision that a child’s mood affects their performance at school because the lower the child’s mood is, the lower their grades are. It would be incorrect to say that their mood causes low grades, how do they know that this is the cause? The low grades at school could be causing their low mood.
    We are all human at the end of the day and it is in our nature to make assumptions based on past experience because it’s how we go about our daily lives but when it comes to conducting research for science we must dig deeper to find an explanation for our assumption and cannot just go off our own beliefs.

  2. I would like to point out that to find a correlation between ice cream sales and increased heat does in no way indicate causation. Even though this appears to be the obvious cause to the ice cream sales you can not infer, scientifically at least, even a chance of cause from the correlation coefficient of these two co-variables, only that there is some relationship. To even begin to show scientifically that there is a chance that one has caused the other an experiments must be carried out under controlled conditions where heat (IV) is specifically manipulated and then it’s effect on a dependent variable such as salivation at the site of ice-cream is measured. From an experiment like this we may then be able to suggest that the IV has cause the DV because the rest of the environment is controlled so that we can be sure that no other factors are effecting the DV, however, even in such controlled conditions there are still many factors that may be the causing the change in the behaviour such as history effects experimenter bias and other demand characteristics. Also we may have to control the experiment so much to attempt to infer causation that we may get to far from what we are actually trying to find in the real world. In which cause we should go through the whole chain of empirical research using every type of study (observation, survey, natural exp, field exp, laboratory exp, etc.) possible in order to attempt to make an educated judgement of causation.

  3. Assuming that correlation does cause causation is a common mistake made by many individuals. One must not conclude that one variable causes another for two reasons:
    1) There may be extraneous variables present that may account for the relationship between the two variables
    2) It is possible that Factor B causes Factor A rather than Factor A causing Factor B
    Therefore, it is appropriate to conclude on a note that says there is a relationship between the two variables as opposed to stating that one variable causes another. Take a look at this example: ( Notice how, in this paper, the researcher states there is an association between alcohol use and increasing severity of domestic violence. It is clear that the researcher has not stated the direction of the relationship between the two variables and for obvious reasons too. This could be due to the fact that there may be extraneous variables present and in addition, domestic violence could cause alcohol use and alcohol use could cause domestic violence, implying that each variable could go either way. This example demonstrates that correlation does not cause causality.
    On the contrary, just because correlation does not cause causality, that does not mean that it is of no use to us. The findings of correlational research can have serious implications in the real world. McNeal & Cimbolic (1986) found a correlation between low synoptic levels of serotonin and an increase in the symptoms of depression. As a result, these findings have led to effective therapies being implemented in the treatment of depression.
    In conclusion, it is clear that one must know how the conclusion was made which correlational findings fail to provide us with. However, just because they merely suggest a relationship that does not mean it has no value to us which is why it is important to consider the correlational findings as they can have implications in the real world.

  4. Refreshingly brief blog, well written. Just because two events appear to be linked does not mean that a cause can be found and it is still a case that not many correlations are particularly rigorous in showing a connection. If correlation was really all that was needed i doubt there would be so many people wasting their time with SPSS.

  5. A wonderfully brief and to the point blog.
    It must be remembered that although correlation does not show causation it can be an indicator that further research is justified. Another thing is that if there is a absence of correlation then this is an indicator that there is an absence of causation.

  6. Interesting thing with correlation and causality is we can never be certain unless we are able to empirically deduce something from an event. Even then, if this were to be done under controlled, experimental conditions, there are always minute confounds – one could say negligible, but in addition these could have a significant effect. We have to take some reasonable conclusions though, because we cannot live our lives on our tip-toes for the risk of being wrong. 1) If the two items that are correlated seem logical, they can be assumed logical, although they’ll need further testing to validate this. Yes this is debatable in ethics, but we need something to act on, being tentative can waste time!

    1) Dowdy, S. and Wearden, S. (1983). “Statistics for Research”, Wiley. ISBN 0471086029 pp 230

  7. Correlations not only an issue with causation due to a possible common factor but also because the variables being investigated could in fact work in the opposite direction. I demonstrate this through an example often argued in the media, that increased exposure to violent video games causes an increase in aggression from the individual (Sherry, 2001). However to me it seems just as plausible that aggressive people could have a preference for violent video games. Therefore it could be just as likely that aggression leads to violent video games just as much as violent video games could lead to aggression. This indicates that due to the limited amount of detail in correlations it is impossible to know for definate the direction of interactions.

  8. Pingback: To My TA: Comments for 14/03/12 « mballen91

  9. it’s always dangerous to say that one variable causes something else becuase we can never actually be 100% certain no matter how strong the evidence appears. For example research has found a correlation between amount of television watched and the level of aggressive behaviours displayed (Black & Newman 1995). As soon as the media saw this research they concluded that watching T.V leads to children becoming aggression. However in over 1000 research studies no research has found a causation. This is becuase many other factors could play a part e.g. one reason for this trend may be that children who watch more T.V are likely to have parents who are less involved in raising them and may neglect their needs which causes the higher aggression. Because of the many variables that can be involved in a certain trend or behaviour it’s virtually impossible to imply a causation as a researcher can never be sure enough to say defintaviley Factor A causes an increase in factor B

  10. As it has been pointed out it is a very common error to confuse correlation and causation. This is because usually if one thing cause another they are usually correlated but that does not mean that if two things are correlated that one cause the other. This can be a difficult thing as with your example of increased heat and melted ice cream. Now common sense would say that the heat causes the increase in melted ice creams but that does not neccessarily is the cause. These two things are highly correlated but this can lead to incorrectly concluding that one cause the other. To assume causality we would have to get rid of every other underlying variable which is next to impossible. With correlations you are generally looking at two variables but how do you know there isn’t another common variable which is the cause which you could overlook and come to the wrong conclusion. I found a web link which explained this really well and had a lot of examples on it so if anyone is still a bit unsure then check it out:

  11. Yes correlation does NOT show causality but it can imply or hint at it.. another way you can look at it is that you can have correlation without having causation but without correlation you cannot have causation…. As Gravetter and Forzano (2009) have stated a correlation is a relationship between two variables and if one variable causes an effect on the other then they do have a relationship. Although correlation does not actually show causation, it can indicate potential causation and whats perhaps overlooked sometimes is that this is important as it can help produce further research questions (Rogers & Nicewander, 1988) thus allowing further advancement within the field… correlation not all bad.

  12. I think that correlation is often a lot more useful than people give it credit for. Even though it cannot conclusively prove anything, it is still a very important part of research. An interesting correlation can make a researcher think of investigating it further, and maybe find causation. It is also very interesting just to know that there is a relationship or link between two variables, especially if they are something that may not be immediately obvious.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s