Causation, Comparative, Correlation...'C' words for all!

Gosh, Susanne. You weren't lying when you said this blog would be a toughie. I think the reason I find it so difficult is because my natural inclination is to infer causation when I'm reading a comparative, or especially correlative study. While reading a few articles to prepare for my lit review (and for this blog entry), I've had to catch myself not assuming that one variable automatically causes the other. I don't know what innate in all human being that makes us want to do that...but something is and it has got to be stopped.

In a study that examined family instability, school context, and the academic careers of adolescents, Cavanaugh and Fomby desired to discover a correlation between family instability (changes in the family structure any time during the child's life up to the present point) and the type of courses a student takes in high school. These researchers used data from the National Longitudinal Study of Adolescent Health and a research study from UNC to form a list of participants and parents. The parents were then given an in-home interview to self-report the family structure changes. Academic transcript were used as data for the type of coursework track the student was on, specifically focusing on the math trajectory. In addition, the school context was evaluated by means of the proportion of families within the school who had experienced at least one family structure transition.

Results showed multiple correlations between the three variables, but for this blog, I'd like to zero in on the correlation that was found between school-level family instability and the math course trajectories. The findings indicate that adolescents in schools with low exposure to family instability were more likely to be on time (paced) with the ninth grade math course enrollment. By the end of high school, the majority of students from these school contexts (84%) completed college-prep math courses. Like I mentioned above, it is hard for me (and I'm sure others) to not take that finding and draw an assumption that families that exhibit stability cause adolescents to be a higher level trajectory in school courses, particularly math.

This automatic inference is dangerous and potentially misleading. There are a number of other variables that need to be considered (average SES of the families in the school, parental education, teachers in the high schools and feeder schools--to name a few) because they may have a factor in the causation. Additionally, who's to say that the student's trajectory isn't causing the family instability (ok, maybe this isn't the best study to illustrate this point, but another danger in making a causation inference is the uncertainty of the direction of the possible causation). Implying causation in a correlational study is far from the purpose of it, which is just to determine if a relationship exists. Inferring causation then not only is ignorant, but minimizes the beauty that is research methods.

Maybe one day researchers can unite and change the license plate to say...NO CAUSATION FROM CORRELATION!