AndrewH's observation and opportunity costs
Scott argues that opportunity costs should be calculated based on what you'll realistically do with resources, not on theoretical optimal alternatives, using examples from cryonics spending, NASA budgets, and charitable giving.
In his discussion of "cryocrastination", AndrewH makes a pretty good point. There may be some better things you can do with the money you'd spend on cryonics insurance. The sort of people who are into cryonics would probably accept that donating it to the Singularity Institute is probably, all in all, a higher utility use of however many dollars. Andrew's conclusion is that you should figure out what maximizes utility and do it, regardless of how small a contribution is involved. He's right, but I want to use the same example to push a point that is very slightly different, or maybe a little more general, or maybe the exact same one but phrased differently.
Consider an argument frequently made when politicians are discussing the budget. I frequently hear people say it would cost between ten and twenty billion dollars a year to feed all the hungry people in the world. I don't know if that's true or not, and considering the recent skepticism about aid it probably isn't, but let's say the politicians believe it. So when they look at (for example) NASA's budget of fifteen billion dollars, they say something like "It's criminal to be spending all this money on space probes and radio telescopes when it could eliminate world hunger, so let's cut NASA's budget."
You see the problem? When we cut NASA's budget, it doesn't immediately go into the "solve world hunger" fund. It goes into the rest of the budget, and probably gets divided among the Congressman Johnson Memorial Fisheries Museum and purchasing twelve-thousand-dollar staplers.
The same is true of cryocrastination. Unless you actually take that money you would have spent on cryonics and donate it to the Singularity Institute, it's going into the rest of your budget, and you'll probably spend it on coffee and plasma TVs and famous statistician trading cards and whatever else.
I find myself frequently making this error in the following way: a beggar asks me for money, and I want to give it to them on the grounds that they have activated my urge to help people. Then think to myself "I can't justify giving the money to this beggar when it would help many more people if I gave it to a responsible charity." So I say no, and forget all about it, and never give the money to anyone. Even though (from a charity point of view) I know of a superior alternative to giving the money to the beggar, I would still be better off just giving the beggar the money!
All this means that for any entity that does not use its resources with maximum efficiency, the opportunity cost of spending a certain amount of resources should not be calculated as what you'd get earn from the best possible use of those resources, but what you'll earn from the use of those resources which you expect to actually occur.