typically

Pure science may not be a typical case, though, because the people who control the funds don't really have any idea whether the work they are funding is ultimately going to turn out productive or not. The work involved is far from routine and basically a jump into the unknown.

I get the risk of fraud and nepotism, but in some other situations (Bell Labs etc.), "choose very good people and let them improvise within certain limits of a budget" turned out to be very efficient. The devil is in the "choose very good people" detail.