I read the summary article, and skimmed the paper.
Interesting article, I'm still trying to figure out the implications.
Are they conflating design time with manufacturing cost? An open design cuts out the design time, but there is still a significant manufacturing step (unlike software).
No, I don't think so. They seem to try to limited their focus to design costs. However, in some cases, manufacturing costs, as embodied in the purchase price, are part of their calculation, as are the marginal cost of producing an alternative using 3d printing or similar.
Are people just rediscovering DIY, vs COTS? People could make all sorts of things DIY, bread, clothes etc, but the point is opportunity cost : how much is your time worth vs the cost you are saving?
A legitimate point, I think. On the one had, the seem to be focused most on areas where the COTS options aren't suitable, on the other hand, they go so far as to assume that customized versions make significant penetration into the market at expense of COTS options.
I guess there are lots of bits of lab equipment that would be cheaper if you made them, but you might end up with no time doing science. The question is which is worth more, 1 hour of doing science, or 1 hour of making syringe pumps? e.g if the science is worth $1000/hour to society, is it worth one hour saving $200 in syringe pumps?
Using a COTS component can bring compromises beyond the cost of acquisition, and those compromises will have costs associated with them. Also today, in the US at least, the labor of graduate students has a very low value.
Perhaps a lot of scientific research is fairly worthless, so it is worth reducing costs as much as possible, but a key scientific discovery could be worth billions, not millions. Quantifying the value of research is hard, as it may only be realised many years later.
Indeed, and to some degree, the scientific enterprise will always involve a lot of inefficiency, because it is (or should be) focused on exploring the unknown. As I've noted above though, I think your model is oversimplified. Selecting and integrating a COTS component can itself take a lot of time and effort at the expense of "doing science." It isn't hard to imagine situations where adapting an existing open source design and running it off on a 3D printer requires less effort, rather than more.
The key part seems to be that open designs can facilitate DIY manufacture with machines that can be run in a home/office. But what is not studied is whether that is a net benefit.
At least part of the analysis seems to be based on existing behavior, and the assumption that that behavior is economically rational. Of course, that, and many other assumptions in the paper can and should be questioned.