« Easily manage photos & videos on iPhones & iPads | Main | An aside: Every Conference Call Ever »
Friday
Jun202014

NSF Astronomy grant sizes

A news item in today's Science contains this statement:

the average size of an NSF astronomy grant, roughly $325,000 a year over 3 years, is almost 30% larger than it was a decade ago.

A colleague pointed out to me that that number ($325K/year) seemed high and was probably a typo. Without commenting on the news item itself (that NSF AST is implementing a "voluntary" limit of one proposal per year as a PI or Co-PI), let's look at some data. (Bottom line: yeah, I'm pretty sure it's a typo and should have read roughly $325,000 over 3 years, giving more like ~$108K/year.)

NSF Awards are searchable. There's a limit of 3000 records returned per search, so I broke up Astronomy awards into <1990, 1990-2004, and >=2005 and downloaded each as a CSV file.

After combining them into a single file, I read it in with a little python:

Now, we're ready to play with the data with a few caveats:

  • Recall that buried in these data are everything from a few thousand dollars for a summer student to some $400Mil for ALMA construction.
  • Also, the data field we're looking at in the CSV file is called AwardedAmountToDate. Sometimes NSF funds a 3 year project all at once. Sometimes they do it year-by-year. It depends on lots of factors. So, the data should be reasonably decent for completed awards ('expired' in NSF lingo). But, for currently active awards there will be some noise. (e.g. Dr. X was awarded a $300K grant for 3 years starting 2014-01-01 that was funded all at once, while Dr. Y was awarded a $300K grant for 3 years starting 2014-01-01 that received its first $50K on 2014-01-01, but has been promised another $50K in September and then the remainder in 2015 and 2016. While for Dr. X and Dr. Y these are nearly effectively the same grants ($300K over 3 years), in the publicly available data we're using here those show up quite differently.)

First, let's look at the per-year funding on grants:

The mean is extremely noisy due to big projects (like ALMA at something like $400Mil), but the median tells a consistent story, having crept up toward about $100K/year. (Recall that the data should be viewed as incomplete for the most recent few years, back to 2011 or so.) You can also see the effect of the ARAA stimulus money in 2009-2010. But, there's no (fair) way to claim that typical NSF Astronomy awards are $325K/year, without cherry picking data. More likely it was a typo. The median per year per grant looks like it's crept up to about $100K/year or a little more depending on how you extrapolate into the 'noisy' data of the past few years. Most grants are 3 years. You can do the math.

Finally, let's look at total award per grant to confirm that:

Yup. Nothing too wacky going on here.


Code to generate the first plot:

Code to generate the second plot: