I love science, and I’ve long planned to stay in academia after finishing my PhD. Last summer I decided to estimate how many years I would need to do a postdoc to have a good chance of getting interviews for assistant professorships at top-50 research universities. I’d also be happy, perhaps even happier, at a research institute, e.g., the Salk Institute, HHMI Janelia Farm, and The Rowland Institute, but there are very few of those positions.
Warning: When I gave this information in the form of a talk one person told me she wanted to punch me in the face, and I was told it was terribly depressing. However, I think it is important to be realistic. This is especially true if you have a significant other that’s waiting on you to “get your career started.”
NSF claims that only 23% of PhDs land tenure-track positions within 3-5 years of finishing their PhD (of course only half of PhDs aim for academia). I wanted a more useful statistic, so I used number of publications. To figure out this number, I spent a while looking at the CV’s of professors in my fields and then estimating how many publications they had when they started their professorships. I limited myself to top-50 research universities and professors that got their jobs after the year 2000. I know this is a crude measure, but I needed to know where I stood to assess if doing a postdoc was even worthwhile.
Important: I also restricted my analysis to people with backgrounds similar to my own, so for Computer Science I looked at the CVs of people studying machine learning, robotics, and computer vision. For Neuroscience, I looked at theoretical/computational neuroscience researchers. For Cognitive Science, I looked at people with a computational focus. Here are the numbers:
Computer Science Departments:
- 22 ± 7 Total Publications
- 14 ±7 First Author Publications
- A lot of tier 1 conferences and workshops
- Need about half as many publications if graduating from top program (e.g., MIT, UC Berkeley, Stanford)
- Sample Size: 17 (machine learning, AI, robotics, and computer vision people)
Cognitive Science Departments:
- 20 ± 9 Total Publications
- 11 ± 7 First Author Publications
- Combination of conference publications (NIPS, Cognitive Science) with one or two high impact papers (Science, Nature, etc.)
- Sample Size: 10 (computational/modeling people)
- 13 ± 7 Total Publications
- 7 ± 4 First Author Publications
- Typically at least 2 papers in Neuron, Nature, Science, or Nature Neuroscience
- Sample Size: 7 (theoretical neuroscientists)
So, let’s assume that I’d like a professorship in Computer Science or Cognitive Science. I have 7 publications currently. I can probably get an interview if I’m within one standard deviation of the mean, so I should consider applying once I have 12-15 publications. If I manage to generate 2 publications per year during a postdoc (let’s be conservative), then I probably need to postdoc for about 3-4 years. This means I’d be an assistant professor around age 35-37, and then get tenure hopefully before 44.
I’m the first to admit that this is a crude way to calculate the number of years, and a lot of other factors are important. For example, if you are interested in teaching universities then the number of publications you have is probably a lot less important than your teaching. If you invent a revolutionary and super hot technology near the end of your PhD, you can also probably get a great job without doing a long postdoc.
I plan to do a follow-up post on the statistics from the faculty candidates interviewed in the Computer Science Department at UC San Diego this spring.