Simpson's Paradox

In 1973, UC Berkeley faced potential legal action due to perceived gender bias in its graduate school admissions. While the overall admission rates showed a discrepancy, with 44% of male applicants admitted compared to 35% of women, a deeper examination by a hired statistician revealed unexpected findings.

Contrary to initial assumptions, the analysis unveiled that the admissions process actually favoured women over men. This revelation stemmed from a closer look at departmental admission rates, where it was discovered that women had higher acceptance rates than men in 4 out of 6 departments.

Further investigation revealed that women tended to apply to more competitive departments (C, D, E, and F), while fewer applied to departments A and B. This disparity in departmental preferences contributed significantly to the observed gender-based differences in overall admission rates.

This case underscores the importance of caution in interpreting data and highlights the phenomenon known as Simpson's Paradox. It serves as a reminder that without careful analysis and consideration of all relevant factors, data can easily be misinterpreted or manipulated to support misleading conclusions.

More on Simpson's Paradox here.