Even in a world where people would be universally willing to do this, it may fail in the case where a unified plot of land is required for a building or project of whatever kind (as in the recent Kelo vs. New London eminent domain case).
Also, “routing around” a particular house may be difficult; it’s made impossible if a group of people in the way band together and collectively hold out to take the builder’s surplus.
It’s my understanding that a good amount of social science does the same—I was recently in a class on organization theory that made few if any testable predictions and ended up being a bunch of just-so stories about how people in organizations think.
One of the side effects of prediciton markets of the Hansonian variety is that people would rapidly see who is doing good science and who isn’t: wealth would be rapidly transferred from those who don’t know how to make good evidence-centric predictions to those who do.