As far as I recall, most of these studies, especially [2], mostly just present their findings of a correlation, along with the data and any variables that may b

As far as I recall, most of these studies, especially [2], mostly just present their findings of a correlation, along with the data and any variables that may be relevant, without trying to prove causation. They may posit some tentative theories in their analyses, but are careful not to make any conclusions on those. However, your observation about many software startups would prefer to release an MVP and iterate rather than invest a lot of upfront work and patent is also borne out by data in [1] and in other studies, such as some of Bessen and Maskin's papers. I think Malandrew's sibling comment is also a very salient point, and combined with my hunch that a vast majority of startups (or, heck, software companies in general) are not doing particularly technically innovative work, it goes a long way in explaining the data in those papers.