I think that should tell you something. The people who have the most money (and available staff) to test this sort of thing have generally opted to not make opening an account mandatory.
See, the way I see it is this. There are basically two groups of shoppers. One group doesn't see a problem with customer accounts. If a site offers the option, they'll likely sign up. But if I site doesn't offer accounts, they'll probably still shop there. It's really unlikely they're going to walk away from a site because the site DOESN'T offer account registration.
On the other hand, there is a good-sized group of customers who absolutely will NOT shop somewhere that requires an account registration.
(Of course, there may be a few people who love setting up accounts with everyone on the planet, and who won't shop anywhere they can't register. It takes all kinds, I'm told. I've just never met such a person so far. I think it's reasonable to assume that if this group exists, it is a tiny minority.)
So if you don't have user accounts at all, you've got a good shot at sales from both of the major groups of shoppers. If you make accounts optional and allow shopping as a guest, no effect on the first group, and you have at least a fighting chance of getting most of the second group. If you make accounts mandatory, still no effect on the first group, but you'll permanently lose 100% of the second group.
So, do the perceived benefits you're going to get from requiring people to fill out an account registration (whatever those benefits may be, and frankly I'm hard-pressed to think of any -- keep in mind, you've already got all their contact information thanks to their having placed an order in the first place) outweigh the potential administrative hassles, reduced customer goodwill and lost sales?
Personally, I don't think so. YMMV.