At least in academia. Drugmonkey has a good discussion, but this is the basic problem:
The blogpost then goes on to calculate a Cumulative Investigator Rate which is basically how many PIs get funded over a 5 year interval out of those who wish to be funded. In 2003 it was 43% and this declined to 31% in 2015. This was for RPGs. If you limit to R01 only, the CIR goes from 45% to 34% over this interval of time. For R21s, the CIR was at 20% in 2003 and is down around 11% for 2015. Newsbreak: Funding rates for R21s are terrible, despite what you would imagine should be the case for this mechanism.
I don’t know if I agree with Drugmonkey’s solution, but there needs to be a recognition that the biomedical workforce–at least the academic portion of it–has a real problem. At the very least, when discussing programs to ‘change science’, we need to explicitly state who gets culled.