Look Closely at Your Data

Watch out for Simpson’s Paradox!  Edward Simpson described this phenomenon in 1951.  It is often associated with an analysis of potential gender discrimination at University of California, Berkeley (UC Berkeley).  In the 1970s, there was a concern about gender discrimination with regard to student acceptance.  At an aggregate level, it appeared as though there may be an issue.

However, once the data was further broken down into groups and their acceptance rates to investigate the situation further, it turned out that the gender bias was for women instead of against them in four departments, while there was no bias in the other departments.

What had led to the data coming across as the reverse of what was actually happening was due to Simpson’s Paradox as a result of most women actually applying to departments with lower acceptance rates in comparison to men. This situation tells us to look at data from various angles and break it down further if possible to avoid something like Simpson’s Paradox situation arising from any hidden variables in the data. And not only to just look at charts and numbers when making decisions, but also to take the time to disaggregate the data as required.