Fontana et al consider a novel dataset on inventions from the annual awards of R&D magazine dating back to 1963. Every year, R&D magazine gives awards to the top 100 most important new, marketable inventions of the year. This data can be said to improve on patent data alone because it restricts attention to “important” innovations only. The standard results on patented innovation apply: decreasing US dominance, a divide between Schumpeterian “entrepreneurial industries” and industries dominated by a small set of firms, and the increasing importance of collaborative and noncorporate (i.e., government and university) innovation. Importantly, as in the seminal paper by Cohen, Nelson and Walsh (2000), the vast majority of important innovation is not patented, except in a few industries such as petrochemicals. This would seem to add evidence to the Levine/Boldrin argument that formal IP protection is relatively unimportant for innovative activity.
http://hermes-ir.lib.hit-u.ac.jp/rs/bitstream/10086/18005/1/070iirWP09_09.pdf
A Fine Theorem 