Microsoft Research seem very evaluatable (they routinely sell their work to product tems, and things like the algorithms driving the ad network from bing came out of MSR) in terms of the bottom line, and Google's research is so integrated with the usual engineering it's hard to know which is which (and hence I doubt it fares any worse than the rest of google in terms of contributing directly to the bottom line).
Oh I agree, I don't know how you would go about it, and I'm sure it's not easy or it would have been done. Talking with someone who has actual experience in both research and ROI studies leads me to believe however that these companies have a pretty good idea of their research ROI, but that it is merely a hard sell to shareholders due to the time delay.
I bet R&D departments benefit from economies of scale -- 5 PhDs can likely produce much more than one PhD at 5 companies, and a solid team attracts top talent -- which means the aforementioned big shots are the likeliest to produce meaningful research. Still, I would give a lot to work at a something like a "research startup."