The Real Cost of Patent Trolls (PDF)
In this academic paper from the Boston University School of Law, the authors show that patent trolls have cost defendants — mostly large technology companies who invest heavily in R&D — $500 billion from 1990 through 2010, and over the last 4 years, the cost has averaged $80 billion per year. Moreover, very little of this money ever makes it to the actual inventors, meaning that the money lost by defendants doesn’t incentivize other inventors. In short, software patents (for most of this behavior surrounds them) are simply a drag on innovation and real progress.
Actually, the paper addresses software patents & patent troll behavior; it omits discussion of any OTHER effects of software patents. Here's one beneficial effect that could be difficult to see from the outside:
Once upon a time I was the main R&D researcher for a compiler company. Pre SW-patents: Everything I did was considered a trade secret, and went unpublished (with a 15-year NDA).
Post SW-patents: either "Apply for a patent, and then publish" (unusual: patents are expensive for a small company) or "Publish fast, so nobody else can patent it out from under us."
Hence, software patents suddenly changed the equation from "everything is secret" to "every advance becomes public, either as part of a patent or as a defense against someone else's patent." Multiply this effect by lots of researchers industry wide.
As always in economics: consider that which is NOT seen as well as that which is seen.