Which expenses would that be that are genuinely avoided while maintaining feature parity, rather than merely avoiding the feature that is bought by those expenses in the traditional system?
For example: Of course not having to worry about chargebacks and related fraud makes Bitcoin potentially cheaper than the existing credit card system. On the other hand, fraud protection is a useful feature, and adding it on top of Bitcoin removes any cost advantage that Bitcoin might have started out with.
This example provides a template that, as far as I can see, applies to every so-called cost saving aspect of Bitcoin (except possibly the fact that Bitcoin might make the whole banking system more competitive; that would undoubtedly be a good thing).
These are only the advantages compared to national transfers, when looking at fees and regulations in the world of international transactions, the advantages are even greater.
Is that really the case? I'm quite sympathetic to the view that it might be, either presently or in the long run (an arms race where attackers and defenders are on the same footing can't get cheaper due to improved technology), but there are expenses in the current system that bitcoin avoids (and a lot that it doesn't, for sure). I've seen various analyses making various claims, and I'm really not sure what the answer is.