In this world, people are knowingly committing offenses all the time. Often they have a rather clear indication of the penalty involved in case they would be caught in action. For them, the decision to commit the offense is based on a two-sided economic assessment. On the one hand people consider the expected gains that are obtained by committing the offense; on the other hand they assess cost of punishment. Naturally, the cost of punishment is for the most part determined by the probability of getting caught, and the penalty in case of getting caught. The penalty of getting caught does not just refer to a financial penalty, but also to factors such as loss of face.
##Prospect theory In 1979, Daniel Kahneman and Amos Tversky launched a model called prospect theory in their paper Prospect Theory: an Analysis of Decision Under Risk. In this paper, the showed the inconsistency of real world behaviour in the face of risky prospects and basic utility theory. Concretely, Kahneman and Tversky indicated that "they [people] prefer a small loss, which can be viewed as an insurance premium, over a small probability of a large loss." This means that when people are presented the options A) to pay $5 with 100% certainty, or B) to pay $5000 with 0.1% certainty, most people will prefer option A. For me, this makes intuitive sense. While the expected costs are equal, option A would not affect my lifestyle in any significant way, whereas option B would have a rather large impact on my financial situation.
##Discouraging scheme In my view, penalties should not serve their purpose in any sort of ex-post retaliation, but rather in the ex-ante discouragement of committing offenses. Puting these two together, I believe we could come up with a better discouragement scheme than a one-to-one mapping of offenses and their corresponding penalties. I propose drawing the penalties from an heavy tailed distribution, such as the Pareto distribution. Basing penalties on such a distribtuion, offenders have a great probability of received only a small fine, and a small probability of receiving a very large fine. I will not discuss the height of the penalties, but I would suggest to initially set the parameters for the distribution so that the mean of the distribution corresponds to the current height of the penalty for a given offense.