The problem with science careers is sample size

Science is an attractive career for many reasons. On the surface, academics have no real boss, flexible working hours, and job-for-life stability. They spend their time poking around, collecting tidbits of data on whatever catches their eye, and self-aggrandizing to passers-by in the hallways. Sounds like a pretty enjoyable career. An undergraduate science student looking to extend her jean-wearing, coffee-guzzling days into retirement could be easily fooled into thinking this was for her (that’s right, over half the undergraduate science students at most universities are female).

As you might guess from the title of this blog, the reality is very different. In fact, the statistics are rather appalling. One in ten biologists has a professor/assistant professor position 10 years after completing her PhD. Admittedly, some of those have left science of their own volition, but many more have been driven out by a lack of opportunity. Theoretically, if everyone wants a to become an academic, a 10% success rate should mean that the best 10% of scientists get positions while the rest do something else, which isn't that different from a lot of other careers. Surely we want the best scientists to lead their own research programs. That’s the problem. I’ve seen people in that top 10% get academic jobs, and I’ve seen people in that top 10% leave science altogether. Same for the other 90%. It all comes down to a problem of iterations.

Let’s say a person can get an academic job if she publishes in one Holy Trinity journal (Cell, Science, Nature- make sure to cross yourself as you say these) during her PhD/post-doc. If a young scientist publishes a total of 4 first author papers during this time, she’s done well. The papers that make it into the Holy Trinity are there because they’re interesting. And they’re interesting because they’ve asked timely questions and gotten useful and sometimes unexpected results. Some of this comes down to outstanding experimental design and skillful execution, but in equal measures it comes down to luck. Even outstanding scientists don’t publish exclusively in the Holy Trinity. Some great ideas simply don’t pan out, or the answer to a key question was “no” rather than “yes”. Biology can’t be bent to the experimenter’s desires. The answer doesn’t change the quality of the work, but it changes the interest factor and therefore the impact factor of the resulting paper. That “yes” or “no” answer often comes at the end of a body of work, when the scientist has already invested 2-3 years in the project, is running out of time and money and needs to publish or perish. Out of 10 great ideas, perhaps 1 or 2 will result in a Holy Trinity paper. Ensuring that 1 in 4 early-career papers gets into a Holy Trinity journal is as much luck as it is skill. In order to gauge scientific ability instead of luckiness scientists need to have more iterations before having their CVs scrutinized. If a paper took 6 months of full-time work, an early-stage scientist could put out at least 10 before applying for independent funding. Three-month projects would give her 20. Then there would be enough data points to assess the quality of the candidate. The more data points there are, factors such as luck will play a smaller and smaller role. As scientists and statisticians, we should know this better than anyone.

Unfortunately, I can’t imagine science moving in that direction. Today’s papers have much more information in them than papers from 10 years ago. A knock-out mouse model used to be a paper in itself; now it’s Figure 1a. The amount of time it takes to do the experiments, however, has remained unchanged. A PhD still produces 1-2 papers, same for a post-doc. Time seems to be constant.