I remember once reading about a theory that option premiums were empirically higher than the value implied from models like black-scholes as it allowed the holder of the option to capture (rather than incur) the bid/offer spread. Does anyone know what papers/books talk about such theory?