Why does the number of publications for computer science and economics PhD students differ by so much?

Question

As a CS PhD student, I'm often surprised at how few papers econ students have when they go into the job market. Even students at top universities typically have 3-4 papers, mostly still unpublished, whereas it is not uncommon for CS students to have more than 10 papers by the time they graduate -- some CS undergraduates already have 3-4 papers or more. As this blog post confirms:

CS candidate typically have many papers while economics students typically have only a few polished papers.

I can see a few reasons that contribute to this, but I don't think they satisfactorily explain the phenomenon.

• Econ students often spend the first year only focusing on classes and qual exam. However, a 1-year difference in a 5-6-year program is not that much, and CS has classes and qual exam of its own, though admittedly more relaxed.

• The publication venue for econ is journals whereas CS publishes in conferences. However, top CS conferences usually have acceptance rates of 15-30% and people spend a lot of effort polishing their papers before the deadline, so I'm not sure how much more polished econ papers can be, whatever "polished" means. Moreover, many CS conference papers are turned into journal papers without major additional contribution.

• The econ job market places a great deal of importance on a single "job market paper", so there is less incentive to work on other papers. However, wouldn't other papers to go along with the job market paper help toward a better profile? Even if econ students don't want to start big projects, there are journals such as Economics Letters that accept shorter papers, whose website even says explicitly: "especially young researchers and advanced graduate students are encouraged to submit their articles."

As I said, I don't think the reasons I described satisfactorily explain this huge difference in number. What am I missing?

Answer 1

I don't believe what you say is univesally true of all CS areas; it is certainly not true of mine. But more to your point, based on my conversations with colleagues in economics, one factor (possibly of several) is definitely the submission-to-publication delay: A conference turnaround is about three months, but a journal can be much longer, especially if major revisions are involved.

Here's some insight on the subject (published a while ago).

Answer 2

Complementing the answer from the CS perspective, here are some thoughts from the economics one. My experience is a bit dated, mostly in theoretical microeconomics, and for candidates at the better ranked schools.

1. There has been increasing pressure to not have a 5+ year graduation rate, instead, to finish in 4 or at the latest 5 years. The pressure ranges from gentle nudging, hard funding limits, to termination. In the 90s, my program had a number of 8th year students (and longer); around 2000, I've seen actual evictions (or attempts to evict) happening over progress considered insufficiently fast. This obviously limits your ability to produce a large numbers of papers.

2. Except in unusual cases, the amount of coursework in the first year makes research impossible; and the one in the second, unlikely. The first quarter of my second year, I got by on 3-5 hours of sleep per night (no hyperbole). Students are likely working on smaller projects during the summer of their first year, but those rarely lead to useful papers. Considering that you need a strong job market paper about 9 months prior to graduating, this leaves 15 months (to 4 year graduation), or 27 months (to 5 year), which isn't too rich to have the polished paper you need.

3. Theoretical work in microeconomics is mathematical to the extent that few people are ready to engage in contemporary research prior to having taken certain elective classes in their second year. It can happen, but most take extensive preparation in stochastics, or algebraic topology, or similar. While the level of mathematical proficiency doesn't compare to a math PhD, it arguably puts the degree closer to one (or probably theoretical CS, or such), in which the number of papers is low too.

4. A paper in Economic Letters is fairly useless on the job market when competing for the good jobs. It's quality, not quantity that counts; and the quality journals (for this sub field) include Econometrica, AER, JET, and such. You are likely to be measured by the quality of the journal of your first publication; if it's in the above, you are likely a rising star, if not, it might be seen as "That's the best we can expect from you after hiring you?". While I know of exceptions (e.g., a job market star publishing his first (joint) paper in a second-tier journal as it grew out of his Masters thesis), it isn't common. He got his faculty position for his unpublished work-in-progress job market paper (later published in Econometrica).

5. The attitude to value quality over quantity extends to tenure decisions. The just mentioned economist has 21 publications in 17 years as faculty (and 4 as a graduate student). This is exactly one per year over his academic career as an economist (if we count graduate school); and he's tenured at a top 3 institution. While there are extremely prolific economists, quantity isn't generally the way to success at the top.

6. Further to 4., competition to publish in the mentioned top journals, their quality standards, and long submission turn-around time, mean that there is nearly no chance to have a publication that matters (those friends who eventually published their job market papers in Econometrica did so several years after graduating - in the two cases coming to mind, 3 or 4 years after, from memory).

7. There is (or at least used to be) also close to no (no?) market to present papers-in-progress prior to being on the job market, except for any internal forum. Speakers invited to give talks at another department are faculty elsewhere, and you attend the big conferences to listen and meet people, but rarely if ever to show your work (if you even do attend). This is only a comment on something stated elsewhere here.