I have long been convinced that teams punt way, way too much. I'd like to see some computer simulations done in which they played through several hundred games with average teams and changed around the punting logic, just to see what maximizes your chances of winning the game. I suspect that it's almost always a mistake to punt any time on the opponent's side of the field, certainly any time you've got less than 10 yards to go, and that you should go for it almost every time it's less than 4th and 5 any time you're beyond your own 35.
Obviously that would have to be modified depending on who you're playing, how good your defense is, etc. But I think that computer simulations might show that the default approach that most coaches have towards punting is far too conservative.