When I was a medical student, I remember doing a psych rotation on a pediatric ward. I was struck by how many of the admissions were for eating disorders. Even before then, I was aware that more than a small number of my female friends and acquaintances were suffering from some form of disordered eating. So when Austin sent me this paper, I was more than a little intrigued.
Eating disorders are currently the deadliest mental disorder in the United States, affecting an estimated 12%–25% of all college women. Previous research has found a positive correlation between sorority membership and eating disorders, but the causal link has not been ﬁrmly established. We contribute to the literature by investigating a possible causal link among sororities and diagnosed eating disorders, measurable weight outcomes, and disordered-eating behaviors using data from the American College Health Association Survey.
I wasn’t wrong in my perceptions back in school and beyond. Eating disorders are the “deadliest mental disorder” in the US, and affect up to 25% of all college women, depending on the study. Prior work has also shown there to be a correlation between belonging to a sorority and having an eating disorder. Of course, correlation doesn’t necessarily mean causation. It’s possible that something else, related both to sorority membership and to eating disorders, is the cause.
This study sought to get more at the causality link between sororities and a number of outcomes, including weight, disordered-eating, and eating disorders using the American College Health Association Survey. This included more than 144,000 students from 123 colleges in three semesters in the United States. It contains data on their health and weight-related attitudes, perceptions, and behaviors. It also shows whether they are in a sorority.
It’s not perfect. Some data on family background is limited. But it has information on some socio-economic factors as well as racial ones. And ones on other health behaviors.
Moreover, the study used propensity score matching and instrument variables, which are sophisticated techniques to try and get towards causality without an RCT. Propensity score use involves matching those in the intervention group (sorority membership) with at least one subject in the control group (non-members) who are as similar as possible on the many measurable characteristics in the dataset. This allows you to be surer that differences are because of the variable of interest.
The instrument variable chosen for this analysis was the percent of males at the same college joining a fraternity. This was assumed to be predictive of sorority membership, but have no effect on the outcomes, after controlling for other factors. Using such a variable in the analysis also helps to get closer to proving causality.
The results? It found that sorority membership did have an effect on obesity (less of it) and BMI (lower, in general). But it didn’t have an effect on being underweight. Membership did, though, have a significant effect on disordered-eating behaviors and diagnosed eating disorders, especially with behaviors like vomiting and laxative/pill misuse.
But when they implemented the more sophisticated methods, they found that it wasn’t as clear as many think. I’m going to let the authors discussion speak for itself:
Namely, in both of these exercises, we ﬁnd evidence that there are possibly no positive causal effects of sorority membership on our outcomes with the exception of BMI. That is, we can only conclude that sorority membership leads to lower BMI for those females joining a sorority, but we cannot exclude the possibility that sorority membership has no effect on sorority members’ other weight outcomes, disordered-eating behaviors, and EDs.
These results, casting doubts on the common belief that sororities are a main cause for adverse weight-related outcomes for college women, have the important policy implications that programs targeted at reducing the prevalence of these adverse outcomes might not reach their goal if they focus on sororities. Limited resources need to be allocated elsewhere.
I look forward to seeing what others think of this study.