13.2: Association and Causation

The stress induced by living with discrimination has been of major concern to health advocates for many years. There has been increasing concern that exposure to discrimination may be associated with mental health issues. One study carried out a preliminary evaluation of how mental health is associated with perceived discrimination in the United States (Kessler et al. 1999). The focus of the study was on exposure and mental health effects of perceived discrimination based on gender, race or ethnicity, age, and socio-economic status. The study was based on large-scale national survey. While discussing the results of the study the authors' state:

However, as noted in the introduction, caution is needed in interpreting these associations in causal terms. Unmeasured common causes and differential propensity to interpret experienced stresses as due to discrimination could introduce bias into these relationships.

It is noteworthy that the authors of the study never say that exposure to discrimination is causing the differences observed in the study. Additionally, in the quote above the authors state that the results should not be interpreted in causal terms, and that “unmeasured common causes could also explain the differences.” Why are the authors of the research being so careful, and what does all this mean?

To help understand why the authors are being so cautious about their conclusions here let us consider a much simpler and more common situation. Suppose that you have gone to study and your favorite coffee shop. This shop also sells sandwiches and other small snacks. On the wall of the eclectically decorated shop a small cuckoo clock chimes out the noon hour. Shortly afterwards, the shop becomes crowded with people buying lunch. You have observed this on an almost daily basis for the past few years. The question is, are people showing up to the coffee shop because the cuckoo clock chimed, or did they show up because it is noon and it is time for lunch?

Of course, the real reason people show up is that it is lunchtime. The clock on the wall also happens to chime because it is lunch time. That is, the time of the day is the reason that people are showing up to get lunch. Certainly, if we set the clock ahead and noon chimed at 9:00am then we would not expect people to show up and eat lunch just because the clock chimed. Moreover, if the clock had wound down and the chime did not happen then we would not expect that there would be no lunch rush simply because the clock did not chime. The two things that we observe, people showing up for lunch and the clock chiming, are both because of the actual time. The fact that it is noontime is said to be a cause of both the clock chiming and the coffee shop being crowded.

Definition: Cause

A cause of an event is one that directly contributes to the occurrence of the event.

A cause is part of what is known as a cause-and-effect system or process. When the first event occurs there is a direct effect which produces the other event. There may be more than one cause for an event. For example, the coffee shop may be closed on Sundays, and hence the noontime hour approaching may not bring a rush of people expecting lunch. Hence, the day of the week is also another cause for the lunchtime rush.

However, if we did not know any better, and we only observed the clock and the people rushing in, we might be tempted to draw the conclusion that there the clock causes the lunch rush to arrive. The chiming of the clock certainly happens in time precedence with the lunchtime rush. If we watched this day after day, if the clock was working properly and it was not Sunday, we would see that the chiming of clock would be followed the lunch rush. But the clock is not causing the lunch rush, what we are observing is two events that are a result of the time. That is, both events are the result of a common cause. The two events that we observe are said to be associated.

Definition: Association

Two variables are associated if the values of one of the variables tends to occur with values of the other variable.

The difference between the two definitions is very important. If there is a causal relationship between two variables, then that means that the first variable directly affects the value of the other variable. Hence, if the first variable is changed then the second variable will respond with a change as well. In the case of the coffee shop, when the time changes the customers show up in response to it being lunch time. If two variables are associated, then changing one of the variables does not mean that the other will change. In this case, changing the time on the clock will not cause customers to show up to the coffee shop.

Researchers would usually like to try to establish a causal relationship, showing that one condition is causing the other condition to occur. When a causal relationship had been established then the research can suggest specific changes that would affect the outcome. If only an association is established then the researchers can suggest that a change may affect the outcome, but there is no guarantee that it will.

In the example of the research on discrimination and mental health, it would be very useful if a causal relationship was established. While the stress for discrimination is not something that can be mandated to change, the information from the study could suggest health interventions that could address the problems of stress and hopefully help those who experience discrimination. Such interventions are still suggestive if only an association is established, but the conclusion is much weaker, and it is not certain in this case that the interventions would be helpful.

As one can imagine, showing that there is a causal relationship between two variables is much more difficult than establishing an association. The quote from the paper on discrimination and mental health hints that more work would need to be done to establish a causal relationship. The next section will discuss what is needed to establish an association, and then we will tackle what is required to establish a causal relationship.