This is something like Milton Friedman's 'pool player analogy'. It doesn't matter if a pool player isn't calculating all the angles and frictions when he pots a

This is something like Milton Friedman's 'pool player analogy'. It doesn't matter if a pool player isn't calculating all the angles and frictions when he pots a ball, because it's as if he was. So as long as your model has predictive power, unrealistic assumptions don't matter. The unfortunate part is that most models in economics have lousy predictive power.