Some research I read recently suggested that it's not software that's the exception, it's chemical and pharmaceuticals (and even there it's not a clear win, nor is it based on high costs, but rather the specificity of patents in those fields creating clearly delineated property rights).
In every other industry surveyed the patents may be felt necessary under current conditions but aren't considered useful in the sense you describe for rewarding innovation (e.g. defensive patents to stop others from blocking access to your own innovations).
I believe the research was cited in Eric von Hippel's Democratizing Innovation, ah here we are from, Chapter 6,
" The real-world value of patent protection has been studied for more than
40 years. Various researchers have found that, with a few exceptions, inno-
vators do not think that patents are very useful either for excluding imita-
tors or for capturing royalties in most industries. (Fields generally cited as
exceptions are pharmaceuticals, chemicals, and chemical processes, where
patents do enable markets for technical information (Arora et al. 2001).)
Most respondents also say that the availability of patent protection does not
induce them to invest more in research and development than they would
if patent protection did not exist."
Various researchers have found that, with a few exceptions, innovators do not think that patents are very useful either for excluding imitators or for capturing royalties in most industries
By this, do they mean that imitators are typically able to operate despite patent protection or that there aren't attempts to imitate? The latter does not really support the concluusion that patents are not useful/effective, since a lack of imitators is something we should expect if they were (though it also is not proof that they are). The former suggests some problem with the legal situation surrounding patents, since it is exactly what patent law is meant to prevent.
In every other industry surveyed the patents may be felt necessary under current conditions but aren't considered useful in the sense you describe for rewarding innovation (e.g. defensive patents to stop others from blocking access to your own innovations).
I believe the research was cited in Eric von Hippel's Democratizing Innovation, ah here we are from, Chapter 6,
" The real-world value of patent protection has been studied for more than 40 years. Various researchers have found that, with a few exceptions, inno- vators do not think that patents are very useful either for excluding imita- tors or for capturing royalties in most industries. (Fields generally cited as exceptions are pharmaceuticals, chemicals, and chemical processes, where patents do enable markets for technical information (Arora et al. 2001).) Most respondents also say that the availability of patent protection does not induce them to invest more in research and development than they would if patent protection did not exist."
link to the full (and very good) book: http://web.mit.edu/evhippel/www/democ1.htm