From my anecdotal observations, it has been a long-standing effort in the free software community to de-emphasize the monetary impact and to bring to the front the political impact. While I don't consider it a mistake, I also think that there is absolutely no shame in bringing economic factors into view.
The final test would be to do an actual TCO study, which is very hard to do, given that most programs are just too different in their free and non-free incarnations. For example, libreoffice is not a free version of M$ Office: it has different features, different requirements, different compatibility relation with other software, and different uses. On the organizational level we find that a typical company uses many pieces of software, some of them free and others non-free, and they all work with each other in some way, and none can be easily swapped out with a free or non-free alternative.
All of these issues can be overcome, however, and in the meanwhile, it is easy to argue that we should expect free software to be cheaper (in TCO terms) than non-free software.
The development should come out to be cheaper, feature for feature. There are many ways in which the cost of developing non-free software balloons with no benefit to the users.
First, non-free software vendors are constantly tempted to develop "anti-features" (they call them features, of course, but they are basically malware). The most successful of them do it with probability close to 100%. They have teams of people dedicated to reducing functionality (for example, tiered OS offerings), breaking compatibility (even with their own older software) to force upgrade, and inserting spyware. Last but not least, they have people whose only purpose is to make software "sexy", so that it can be sold to chaps, even though it is stuffed with malware. Enter graphical interface over-design and marketing expenses. All of this takes real money out of the development budget, and sooner or later the costs are passed on to the users.
Second, they treat the source code as a trade secret, and consequently have to spend money to provide a fitting level of security, starting with physical access to the production hardware, and ending with checking one's credentials every time the code is accessed. As non-free software projects get bigger, they have to take the trust factor into account, so they prevent most programmers from accessing most of the code. They introduce even more expensive access control at this point, and most of their programmers are less efficient than they could have been, because they are prevented from understanding how the software works. And the debugging has to be done in the house: unlike free software projects, they are unable to crowd-source it, which would allow to shift some of the cost onto the early adopters and volunteers.
Third, we don't have to limit ourselves to just the direct cost of developing the software. What about the cost of educating the very programmers who develop the software? These people expect to be paid enough to match the investment they've made into education, so reducing the cost of educating programmers should lower the TCO as well. The impact is hard to quantify here, but I am willing to bet that making people understand how a free OS works (understand it enough to write great software for it) is cheaper than making them understand how a non-free OS works. Why would we expect anything else when non-free vendors spend money to prevent free education? As the guardians of the trade secrets, they have the de facto monopoly on "teaching" people how to write around their dumb OS. So we should expect them to charge the monopoly prices. Compare it with a free OS situation, where you can go to the free market and find thousands upon thousands of people who understand how a free OS works, all in competition for your educational dollar.
Want a feature added? Why would you expect a monopolist developer to do it cheaper?
Want to transition away from a piece of software or to replace it with a counterpart? Why would you expect a non-free software vendor not to write anti-features specifically designed to make this as painful and costly as possible?
This list just goes on. And what do we have on the other side of this scale? Where exactly do we waste money when we develop free software? May be I am just lacking imagination, but I really cannot point out a single factor that makes non-free software cheaper to develop feature for feature. And once again, this rant should not be taken as a claim that free software is cheaper. I am arguing that it should be cheaper for many very good reasons, and that a careful statistical study is urgently needed to drive this point home. Freedom is great, but a simple consumer advocacy argument cannot possibly hurt. All non-free software is a rip-off and a scam, and I am convinced we should be able to prove it.