Success rates for NIH R01 grants have been falling fast over the last few years, to the great dismay of life scientists. Rates peaked at 32% in 1999, then fell to 23% in 2005.
In an article in Science, Elias Zerhouni blamed increasing numbers of grant applications: the number of applications for R01s increased by 39% between 1997 and 2005.
The demand-side story is only half the picture, though. While poring over some NIH data in preparation for my talk at NBER this week, I noticed an interesting thing: the number of R01s being awarded has been dropping rapidly since the doubling. Take a look:
The number of new new R01s in 2007 is down from by more than 25% from the peak in 2003. (Note that all figures before 2007 come from NIH; the 2007 number of applications has been extrapolated, and the number of awards comes from CRISP. For this discussion, I’m going to restrict my attention to new awards; something similar has happened with renewals)
The story out of NIH is that there are problems because of flat budgets. That’s true, but the overall decrease in the budget in real terms has been relatively modest: the overall budget is down 6.6% in real terms. Moreover, as of this year, NIH has stopped giving automatic inflationary increases for multi-year grants. If all grant amounts stay the same, a 0% change in the budget should have no effect on the numbers. The overall budget is down from its 2003-2004 peak, but it’s still substantially greater than it was in 2002. Basically, a little less than half of the increase in the last year of the doubling has been eroded. Why has such a modest budget decrease resulted in such a dramatic decrease in the number of awards?
Scientists are complaining that funds have been diverted from R01s into other projects. This is also true: R01s received 60% of the overall grant budget in 1997; in 2005 the R01 share was down to 52.5%. This decrease took place slowly over the course of the doubling, however, and at least as of 2005 had not resulted in any decrease in the total amount of funding for R01s. So conceivably, the number of R01s could have increased even more, but again, there is no reason for the number of awards to have decreased.
What happened, I think, is more interesting. I believe that the real problem is independent of these essentially external events: the problem is built in to the system, and a contraction would have occurred regardless of the tax and allocation issues.
Here’s the story: The doubling consisted of a series of ~15% budget increases followed by basically flat budgets. (Let’s forget about inflation for now.)
In year 1, NIH gets a huge new influx of money into the R01 program. R01s last 3-5 years, mostly closer to 5 (average = 4.3 years), so they can only modify the 20% or so of grants that roll over each year. R01 budgets did not increase dramatically, so there was money left over that could be used to expand the number of awards. Similar things happen in years 2 and 3.
Toward the middle of the doubling, the costs of the higher-budget awards created in the first couple of years start to eat up more and more of the budget surplus, so the number of awards can’t be increased, and numbers start to level out. The amounts of the awards keep increasing, though.
Finally, at the end of the doubling, things max out.
Here’s the key observation: if one does not let new R01 budgets decrease from year to year, then increasing the size of R01s in one year locks one into budget increases for the next 4 years as older, less expensive R01s roll over and are replaced with newer, more expensive R01s. Thus, an increase in the average new R01 budget in 2003 locks the NIH into budget increases through 2007. Budgets since 2003 have been flat or declining, so propagating the 2003 (and 2002 and 2001, etc) increase through successive cohorts has forced numbers of grants to be cut.
In 1997, before the doubling, there were ~5 cohorts of R01s with an average of 3244 awards in each cohort and an average budget of $208,000 per year.
In 2003, at the end of the doubling, there were 4569 new R01s with an average budget of $351,000 per year.
If the 2003 numbers were to have been sustained, we would have seen the number of awards increase by 41% and their size increase by 69%. The trouble is that the budget increased only by a factor of 2, and 1.41 * 1.69 = 2.4 > 2. Basically, the gradual expansion of the R01 budgets appears to have led to more grants being created than could be sustained.
Ideally one would adopt a pay-as-you-go kind of system, where R01 numbers and budgets are not increased in any year without enough surplus in the budget (or guaranteed budget increases) to apply a similar increase to all cohorts in future years. The budget increases in the early years of the doubling appear to have been more than the NIH needed to expand the R01 program at a sustainable level, so it grew too big. The NIH can’t save money from year to year, so it wasn’t unreasonable for them to have spent the surplus growing the program. Perhaps it would have been better to have put the extra funds into a pool of shorter, transitional grants that were clearly marked for termination after a few years?