When looking at UC Berkeley’s admission for their graduate school in 1973, a concerning trend was noticed and caught before a lawsuit could take place. This cause for concern was due to possible gender discrimination due to the higher percentage of males being admitted to thge school compared to females. The figure below shows that when looking at the admission status based on gender, more men were accepted compared to women. To avoid such a lawsuit, UC Berkeley hired Peter Bickel, a statistician, to take a closer look at the data.

What Bickel found was unexpected. There was a gender bias in the admittance to the UC Berkeley graduate school, although it was not towards the men, it was towards the women. This gender bias towards women was apparent in four of the six departments because the majority of women applied to departments that accepted a smaller percentage of students overall. The original data in the figure above did not include the different department acceptance rates, but only focused on the overall acceptance rate of the school. The figure below includes each department and their admittance rate. It can clearly been seen that more women were accepted in departments A, B, D, and F.

What Beckel came across is known as called Simpson’s Paradox. This paradox occurs when data is looked at in small groups, which may lead to the date being interpreted incorrectly until the missing variable is added. This missing variable, or lurking variable, is added to the data resulting in the initial findings of the small data set to be false. The UC Berkeley admissions situation is a great example of Simposn’s Paradox. Without the departments included in the data, it easily looked like gender discrimination against women was taking place. When Bickel added the departments in the data, which was the lurking variable, gender discrimination against women was quickly dismissed. The truth in the data was discovered once the mssing variable was present.

Works Cited:

Grigg, Tom. “Simpson’s Paradox and Interpreting Data.” Medium, Towards Data Science, 8 Jan. 2019, towardsdatascience.com/simpsons-paradox-and-interpreting-data-6a0443516765.