Consider this abstract:
We investigate women's underrepresentation among holders of commercialized patents: only 5.5% of holders of such patents are female. Using the National Survey of College Graduates 2003, we find only 7% of the gap is accounted for by women's lower probability of holding any science or engineering degree, because women with such a degree are scarcely more likely to patent than women without. Differences among those without a science or engineering degree account for 15%, while 78% is accounted for by differences among those with a science or engineering degree. For the latter group, we find that women's underrepresentation in engineering and in jobs involving development and design explain much of the gap; closing it would increase U.S. GDP per capita by 2.7%. [Emphasis added.]
The concluding assertion is obviously false. This is simply because patents are not distributed randomly. The system is opt-in, which introduces self-selection bias (which may possibly be particularly influential at the corporate level), and patents have to be approved.
The latter qualification is especially important, because inventing new things and creating meaningful innovations generally requires a high degree of intelligence and knowledge. Since men have the flatter distribution curve when it comes to intelligence, it stands to reason that men should naturally receive more patents because they have more of the elite intelligence that generally correlates with invention and innovation.
Incidentally, this also renders the conclusion void as well, since it is predicated on the assumption that patents are essentially distributed randomly. Since they are not, and since women are not in a position to perform equally in regards to innovation, it is illogical to assert that closing the engineering and R&D gap will solve the patent inequality problem, and correspondingly increase GDP.