On Wealth-Dependent Penalties

Greg Mankiw responds to Tyler Cowen's post of a news story, and I respond again. First, the news:

A Swiss court has slapped a wealthy speeder with a chalet-sized fine — a full $290,000. Judges ... based the record-breaking fine on the speeder's estimated wealth of over $20 million.

A statement on the court's Web site says the driver — a repeat offender — drove up to 35 miles an hour (57 kilometers an hour) faster than the 50-mile-an-hour (80-kilometer-an-hour) limit.

Mankiw's take:

Is it optimal to base fines on wealth?

My first thought is no. We fine activities that have negative externalities, such as putting others at risk. If X is the size of the externality, and p is the probability of being caught, then the optimal fine is X/p. That will give people the right incentive to produce the optimal quantity of the externality. Under this policy, the rich may choose to speed more, but that is optimal. If we have an optimal carbon tax, the rich will produce more carbon too. Optimal pigovian taxes do not eliminate income effects.

He then finds a "theoretical" exception for when externalities are very large [such as possible loss of human life] but there is also a risk of mistaken or malicious ticketing. That could introduce wealth-dependent fines: "Because the mistaken ticket is a proportionately smaller fraction of their wealth, we need to worry less about the uncertainty large fines impose. The result is larger fines for richer offenders."

I responded:
Where externalities are purely monetary, I agree. The rich can afford to clean up more of their own messes, so let them. When the externality involves actively endangering other human life, I feel a moral impulse that tells me the rich should have no more right than a poor person. Money can't make it all better when a child is run over or dies in an accident.

What would you think of the following proposition? In most of economics, we consider u=u(y) with positive first and negative second derivatives. When we pose fines without regard for wealth, we are implicitly fining a poor person more utility than a rich person for committing the same immoral, unlawful, or socially-suboptimal act. If fines are imposed in constant utility terms, the rich will pay more money than the poor, but feel the same punishment. This seems to me in this case a more egalitarian (and probably more effective) solution to deterrence.

Thoughts?