The Pace of NSF Funded Research

Recently on Facebook I came across a note by Chris Erdmann that some handy folks at Harvard put together statistics on (nearly) every astronomy paper from 1995 to present that was funded through an NSF AST grant. This seemed like a really interesting dataset, especially for a young (read: financially uncertain) research such as myself.
So parsing through all 29,042 papers listed, here are two interesting things I've learned...

1. A typical AST grant produces < 10 papers

This is a simple histogram of the number of papers each unique Grant Number produced. Many people have only produced 1 paper with a grant, but the average is about 8.75 (and the median is 3).
Distribution of number of papers per grant, with a
mean of 8.75 papers per grant (blue dashed line)

2. A grant has its peak output of papers at 3.1 years 

This is a more intriguing figure to me. I've plotted (Year Published - Year Grant Awarded) as a function of Year Grant Awarded for all 29K papers, and then binned it up with pixels. You can clearly see the peak productivity between 2-4 years. I've marked the mean (solid orange) and stdddev (dashed orange) lines for each year.

"nsfastgrantbib" data from the Astronomy Dataverse


Following this second figure, we expect that the grants from 2009 onward have only produced maybe half of their useful (publishable) results, despite many with dwindling funding.

Likewise, it's very encouraging that the grants awarded today will still be producing usable science well into the next decade!

Finally, we are reassured that Astronomy is an exciting and fast-paced field, and that continued strong funding is required to preserve this fact.

Data Reference
Christopher Erdmann; Louise Rubin, "Compiled List of NSF Grants to ADS Records from 1995 to August 2012", hdl:10904/10152 V2 [Version]


  1. Might want to consider normalizing by grant duration.

  2. Or by grant award amount! What's the $ per paper? And do larger grants have more papers per $ than small grants? Or is it the other way around?

  3. We purposely left out grant $ in our dataset thinking it would be too controversial. We have it though. Perhaps we should release it in the next version but after consulting our ADS friends.

    1. I for one would *love* to see the grant $ data when it becomes available! Would be fascinating too see the typical "cost" per paper published.

      With this data, however, one can go and find the *most* productive grant, and the awarded funding for it. Perhaps I'll do some follow-up analysis on this.

    2. Hi James. We're working on it. We'll post the new dataset to the Astronomers Facebook page soon. BTW, I know I am late responding but it is great to see that this was picked up by the Atlantic Wire.


  4. From a writer. . .and sort of a jerky thing to say, but I can't help myself. It should be "its" not "it's," which means "it is."


Inappropriate comments, advertisements, or spam will be removed.
Posts older than 2 weeks have moderated comments.
(Anonymous commenting disabled due to increasing spam)