One way we could dramatically increase the availability of funds for science: allow science to be conducted at market rates.
Currently, most science is funded under the following model. A scientist is directly employed by a host institution, typically a university. The scientist then applies for money from the government, of which 50-60% will be spent on research. The remainder will be pocketed by the host institution as "overhead".
In principle, the overhead pays for things like lab space, but in reality lab space could often be acquired for vastly less than 50% of a large grant. Scientists put up with this because it is not possible to apply for a grant without being employed by a host institution.
I propose to cut out the middleman, and give money directly to scientists. This will dramatically increase the funds available for science.
It's walled off so that larger entities cannot propose to it (somewhat contrary to "free market" principles, but whatever).
I've managed proposal reviews for SBIR, and the average quality is noticeably lower than NSF proposals I've reviewed. Of course there is significant overlap between them, but that's the average.
So I'm not sure your idea would accomplish anything new.
SBIR is mainly grants to get innovative small businesses going (which is why large businesses are not involved). It's also mostly military, as far as I know. I think SBIR is a great program, but it's not really meant for funding basic research.
But yes, something similarly structured but aimed at basic research might be a good idea.
Greater transparency also couldn't hurt. We could simply demand that universities itemize overhead. Then the PI is responsible for rejecting unneeded items (e.g. university computer lab fee if the PI has his own computer) and the NSF is responsible for rejecting items which are >10% above market.
(I managed a small SBIR category for NASA.) Despite the rhetoric, they do provide a sustainable flow of work to many organizations (sometimes referred to as "SBIR shops").
About basic vs. applied research: What I was trying to say is: The relative lack of breakthrough success for SBIR seems to imply that smaller, more nimble (market-responsive?) organizations are not the answer.
It's possible to apply for grants without being in a university, though the bar to have your application taken seriously is admittedly higher. But, if you convince people you're legit, there are a good number of "businesses" whose revenue is mostly NSF grants, e.g. Cycorp: http://www.cyc.com/
Depending on your area, I'm not sure the overhead costs come out to less, though. If you're independent, you not only have to pay for the overhead of physical equipment out of that money, but also of a bunch of salaries and benefits (including your own, lab staff, etc.). You're also probably going to have to pay your researchers more than a university pays its grad students.
Currently, most science is funded under the following model. A scientist is directly employed by a host institution, typically a university. The scientist then applies for money from the government, of which 50-60% will be spent on research. The remainder will be pocketed by the host institution as "overhead".
In principle, the overhead pays for things like lab space, but in reality lab space could often be acquired for vastly less than 50% of a large grant. Scientists put up with this because it is not possible to apply for a grant without being employed by a host institution.
I propose to cut out the middleman, and give money directly to scientists. This will dramatically increase the funds available for science.