Gender Inequality: The Simpsons Paradox within UCB’s Admission Data

When viewing UC Berkeley’s admission data, a certain major inequality pops out. Specifically, the data appears to show that women are admitted much less often than men on average. This clearly is not ideal, especially if it is in any way misogynistic/malicious but any ethical quandaries regarding this can be gotten rid of when more properly examining the data.

Apparently, this data representation/portrayal falls under a common statistical anomaly named the Simpson’s paradox after the widely beloved show. Essentially, the data shows something far different than reality while viewed on the surface level due to any number of hidden variables. When these variables have been accounted for, the data often either corrects itself, even to the point that it reverses. This is very much the case for UCB’s admissions data. When accounting for more than just gender and admissions quantities, we can see the data correct itself to the point that women are even admitted more often than men on a percentage level. The data point that reveals this is the department of which the genders more often apply too. Women on average tend to apply to departments that are ultimately harder to be admitted to, while men on average tend to apply to the easier departments. This leads to a surplus of men and a lesser number of women despite the equal and greater admission rate. Ultimately, when viewing data with such problematic statistical tendencies, it is of utmost importance to search for any hidden or lurking variables.

Figure 1

Figure 1 shows the fundamental problem shown above: women seem to be admitted far less on average then men. This is because of the very surface level nature of this representation.

Figure 2

Upon viewing Figure 2, the first major discrepancy discussed becomes apparent. Females tend to lean towards the harder classes on the far right while males tend to go for easier ones on the far left. When viewed with this extra variable, the fact that females are actually more likely to be admitted for any given class can be seen.

Figure 3

Finally, by examining figure 3, a more direct view of the admittance rate for any given department can be seen. This only further hones in on the concepts touched on above.

Citations:

Tom Grigg. “Simpson’s Paradox and Interpreting Data.” Website/Blog, Towards Data Science, Dec 9, 2018, https://towardsdatascience.com/simpsons-paradox-and-interpreting-data-6a0443516765.