Rylan Schaeffer

Kernel Papers

1 May 2022

How Survivorship Bias Creates PhD-Professor Market Inefficiencies

by Rylan Schaeffer

Finding your PhD advisor is perhaps the most critical decision you’ll make in graduate school. A good advisor can set you up to transform the world, and a bad advisor can drive you to insanity. Consequently, I thought I might share a lesson I learned while rotating through 3 supervisors in the first year of my PhD in how survivorship bias creates market inefficiencies in the PhD-Professor pairing market.

Disclaimer: In many situations, I fell short of being a model student. This reflection is not about placing blame, but rather how future PhD students might avoid the mistakes I made in finding a supervisor.

Before starting my PhD rotations, I was strongly recommended to speak with previous students to get a sense of what each advisor is like, how to develop a productive relationship with them, etc. I made sure to do that, speaking to an excess number of current and graduated students (probably ~10-20). I would ask oftentimes blunt questions about incompatibility, asking what type of person works poorly with the professor, probing what the worst outcome of working with that professor has been. I am very appreciative that current and former students honestly described to me what they felt are the pain points of working with their respective professors, and in almost all cases, I found their descriptions to be correct.

However, upon rotating, I discovered that the students’ perspectives were often warped! Much to my chagrin, I discovered that key pain points had been omitted. Why, in the many conversations, did these additional details not arise?

Survivorship bias. A student’s relationship with a professor is akin to dating (albeit where one party holds financial control and commands the other party), and students have different dealbreakers. Of the students that remain in a lab after rotating, these students are less likely to notice mannerisms that others might find disagreeable and more likely to downweigh such mannerisms in significance. Consequently, these students offer highly biased views.

In order to obtain an unbiased view, a new PhD student should also speak with previous students that worked with the professor and ultimately decided not to continue. However, to the best of my knowledge, no such list exists (or certainly isn’t public), and as such, an inefficiency exists in the PhD-professor dating market that needn’t exist. In retrospect, I wish I had asked lab members for names of people who hadn’t joined the lab and had spoken with those few people far more extensively. This also suggests a desirable new service: ratemyadvisor.com

tags: graduate-school