yep, immediately thought of that too. PG writes VERY good stuff most of the time and is very smart about a lot of things, but when he strays into areas in which he is not well versed (the unions essay comes to mind), he ends up writing pieces with obvious holes.
I read the union essay a while ago, but your failure to sufficiently consider risk-aversion in this one was a major hole. I don't mean to really criticize that much, you write consistently great stuff.
That works, except then I guess the hole was not making it clear enough that you were talking about utility rather than expected value. The problem with your response, and the reason that I don't really see it as a good explanation, is that your basic principle is far less useful if it requires the user to calculate their expected utility.
