University of California Berkley Admissions

A case of gender bias became an issue for the University of California Berkeley during the 1973 academic year. The case caused a lot of controversy for the school as well as potential lawsuits. But, it wasn’t until a statistician was hired to dive deeper into the data and unveil what was really going on.

Here we can see the initial issue when people looked at the numbers from a surface-level perspective.

From a gender point of view, it is clear that the proportion of admitted female students were lower than admitted male students. Although the proportions are helpful to see a broad view of admission status of students, it does not give enough details such as the sample size or what department the students applied for.

With this information, let’s take a deeper dive into UCB’s admission data to see for ourselves what actually happened here.

From the graph above, it is evident that the admission rates are relatively the same for each gender when broken out by department level. By looking at this, there is not a gender discrimination problem at UCB. It is actually quite the opposite. It is very clear that different departments accept a different proportion of students. For example, Department A accepts more students than Department F. This concept also contributes to the situation.

How can it be that looking at UCB’s rate intitially, there seemed to be a problem? Well, this issue can be explained by the phenomenon called Simpson’s paradox, which is when a trend presents itself in different groups of data, but changes when the data is aggregated. Simpson’s paradox is actually a common situation that takes place in data. That is why is is extremely important to look deeper than the surface level.

We can take a step further by observing the number of applicants in each department.

With this information, women tend to apply for Departments that do not accept as many students. Also, the sample of female students is significantly lower than male students. Therefore, each acceptance and rejection holds more weight for females due to smaller sample size. This concept along with a majority of women applications going to departments who accept a smaller number of overall students contributes to the issue we saw in the beginning. Simpson’s paradox is at work!

Thankfully, there are tools out there to help us discover situations like these. Without them, the University of California Berkeley would have had a serious lawsuit on there hands because people were looking at the data to broadly. With data analytics, we are about to tell a true story about what happened with UCB admissions in the year 1973.

Works Cited

Fenn, Shirin. “Simpson’s Paradox and Interpreting Data.” Towards Data Science, Medium, 10 Dec. 2019, https://towardsdatascience.com/simpsons-paradox-and-interpreting-data-6a0443516765.