Correlation and Causality: More than meets the eye
Our evolution has manifested itself largely in our ability to find patterns and trends with very little information. On the one hand, it has allowed us to develop ways of deciding quickly without having complete information (heuristics). For example, when we drive a car and make hundreds of decisions automatically.
On the other hand, it helps us to reduce uncertainty on complex issues (or for which we have no information) by preventing us from consuming ourselves to look for elaborate explanations without having sufficient data. For example, when we used to blame divine entities for phenomena we did not understand (such as lightning).
But apart from the obvious advantages, this trend of ours also brings some limitations. One of them is that we tend to consider that if one thing increases at the same time as the other (correlation), then it is because the increase in one causes the increase in the other (causality).
It is true that when there is a correlation between factors A and B it may mean that cause B. However, this is often not the case. Even when this chance seems obvious to us.
This happens with positive or negative correlations (when A increases B decreases and vice versa).
Group cohesion and performance
In one football team, enquiries were made to gauge group cohesion over a season. It was found that when group cohesion was high, performance also tended to be high (positive correlation). So we can see that group cohesion increases income, right?
Not necessarily. There are also a number of investigations suggesting other possibilities as to the sense of causality. But let us look at possible explanations:
- Group cohesion improves performance.
- Performance improves group cohesion.
- Both of the above hypotheses are true and there is a circular effect (positive feedback).
- There is another variable which, when it increases, improves both cohesion and performance. For example, the hours of training.
- There are several factors which improve both variables. For example, the hours of training, the number of internships, collective training, group meetings.
- It was a correlation caused by casuality. Without major changes in the other variables, in the following seasons the two dimensions had no significant correlations.
Correlation without causality
This fragility makes us very susceptible to manipulation and to the internalization of “false truths”. Pre-conceived or widely spread ideas accentuate the effect and make us fall into the bias of confirmation.
We can give several examples:
- • A brand which associates the performance of an athlete with the use of a given sporting material when other factors such as training are probably more decisive.
- • A politician who states that the country’s good economic performance was the result of his policies when other factors were more decisive (e.g. regional growth, previous reforms, economic factors, etc.).
- • Someone who proposes products or therapeutic practices that we perceive as effective but which, when scrutinised through good quality scientific research, show no result.
- • A leader of a xenophobic movement that justifies with genetic differences the greater number of crimes committed by representatives of a given ethnic group. Often, however, the differences result from very specific social problems within that group or its subgroup, or simply these crimes occur mainly in social strata where that ethnicity is most represented.
- • An activist who associates small numbers, of one ethnic or gender group, in a well-paid activity with widespread discrimination without recognizing the different characteristics of that group, such as simply the relative share of those interested in those functions.
In fact, determining causality is more complex than it seems, and it is not always possible to do it exactly. Reality usually presents more complex relationships than we tend to perceive.
In order not to fall into this error of reasoning it is important to avoid hasty conclusions even if they seem logical and always to put forward other possible hypotheses. Only in this way can we avoid being deceived by others, or only by our intuitions.
Correlation does not mean Causality!