In a previous post, I presented data on the growth of average sizes of Research Project Grants (RPGs) from FY1983 to FY2013. The growth in the average RPG size over this period exceeded inflation, even measured by the Biomedical Research and Development Price Index, BRDPI. However, I noted that the **average** may or may not be a representative measure of growth since the distribution of RPG sizes has changed with the addition of large grants on one end and small grants (R03s, R21s) on the other end. I proposed the hypothesis that the growth in average RPG size could be due to growth in the number of and size of larger awards with less due to the growth in the bulk of the RPG distribution which is made up of primarily R01s.

To address this hypothesis, I have now examined full R01 distributions for years FY2003 and FY2013. From these years, the average size of an RPG grew from $380K in annual total costs to $445K, a 17% increase in nominal value. Correcting for inflation measured by BRDPI, this corresponds to a 19% drop in buying power.

How does this RPG average compare with the R01 average? This distribution of R01 sizes for FY2003 is shown below:

In FY2003, the average annual total cost for an R01 was $338K and the median was $306K. For comparison, in FY2013, the average total cost was $430K, a 27% increase (or a 9% drop in buying power, correcting for BRDPI). In FY2013, the median total cost had grown to $360K, a more modest 18% increase (or a drop of 18% in buying power).

Note that the left side of the distribution shifted by approximately $50K from FY2003 to FY2013 whereas the right side of the distribution was essentially fixed. This observation will be discussed further below.

Examination of the distribution reveals the presence of more than 250 R01s with annual total costs exceeding $1M in FY2003. Removing these large R01s has only a small effect on the growth in R01 size with the average increasing from $326K in FY2003 to $401K in FY2013, a 23% nominal increase and the median increasing from $305K in FY2003 to $357K, a 17% nominal increase.

These data reveal that the growth in the average R01 size actually exceeded the growth in the average RPG size over this period of time whereas the growth in the median R01 size closely matched the growth in average RPG size. Thus, my hypothesis was not correct. The growth in the average RPG size corresponds to the growth in R01 sizes in many regards.

The distribution of the sizes of R01s in FY2003 is approximated as a Gaussian remarkably well.

A Gaussian fits the size distribution on the left side quite well. The fit is less good on the right side with an "excess" number of R01s with annual total costs larger than approximately $400K. Based on a comparison of the distribution of annual total costs with annual direct costs, presented in my previous post, the boundary near $400K in annual **total** costs corresponds to the boundary of $250K in annual **direct** costs, that is, the cap for modular grants.

The size distribution for FY2013 is fit by a Gaussian less well.

As noted above, the left side of the distribution shifted upwards by approximately $50K from FY2003 whereas the right side was essentially fixed at the annual total cost level corresponding to the modular cap. Consistent with the notion that R01 sizes are being limited by this cap, the actual distribution is slightly narrower than the best fit Gaussian.

The "excess" of grants above the modular cap has increased by 2.5 fold from FY2003 to FY2013. This excess corresponds to 14% of the area of the Gaussian distribution in FY2003 and 34% of that area in FY2013.

This analysis reveals at least two important points. First, the modular cap on direct costs is clearly influencing the distribution of R01 grant sizes awarded and this effect is much more pronounced in FY2013 than it was in FY2003. Second, the overall trends in the R01 grant pool matches those in the overall RPG pool so that issues related to the balance between grant sizes and stipend levels are not primarily related to the choice of average RPG size as a metric. However, the influence of the modular cap as well as the impact of increased non-stipend costs (tuition, fringe benefits) raised by commenters on previous posts must still be considered.