This hypothesis is testable and, from personal experience that I regrettably can't show you, the data do not support it. There is no reason you should take my word for this -- try it for a week or two on a site you control.
Increasing user investment in the interaction is one way to cut down on abandonment -- for example, if you're doing comments, asking for a name/email address BEFORE they write the comment will decrease the number of comments you receive markedly, but asking it afterwards will decrease it very little, because people have emotional investment in seeing those two paragraphs they just wrote actually show up.
I'd say it's also a way to stack up on churn. Treat me like that and I may go through your damned procedure to at least not have wasted my last 5 minutes for nothing. But at the same time you have just lost all my sympathy which might influence buying decisions and definately influences loyality.
Always look at the "lifetime value" of your customers, a simple conversion split-test is only half of the story.
Yes, making accounts is a pain that drives users away. But you get round by being clever with sign-ups; not by pretending you don't have a registration page until the user is deeper into the process. We'll just leave all the same.