He's ignoring a subset of crimes in order to evaluate what society would be like if that subset wasn't illegal. The point is if you're using lawbreaking as a metric for which people are bad, it may be misleading when used to inform future laws.