When I was on a committee to study honor codes, the rule of thumb in the literature about this was a 20/60/20 split in the population. 20% will never cheat, 20% will always cheat, and you can only hope to affect the behavior of the 60% in the middle with policy.