Saturday, October 5, 2013

Punished by Fate

Bad luck
Winslow Homer, The Gulf Stream, 1899
“Dans ce pays-ci, il est bon de tuer de temps en temps un amiral pour encourager les autres.”

— Voltaire, Candide

Then, as his planet killed him, it occurred to Kynes that his father and all the other scientists were wrong, that the most persistent principles of the universe were accident and error.

— Frank Herbert, Dune

Marxist rationalist C.J.F. Dillow recently published a short post illustrating a central contention of his: that ordinary people systematically misallocate praise and blame to others based on their misunderstanding of the importance of chance in human outcomes. In particular, he cites an interesting experiment:
[The researchers] split subjects into a principal and agent. The agent chose between a safe option and a lottery, and the principal then split a sum of money between himself and the agent after seeing the outcome of the lottery. They found that principals’ payments depended upon the outcome of the lottery, even though this was obviously out of the agents’ control. For example, agents who chose the safe option were paid less if the lottery won than if it didn’t.

The researchers choose to explain this finding as the allocation by the principals of “unjustified blame” to the agents. Mr. Dillow finds this outcome

consistent with research... which has also found that people just can’t distinguish between luck and skill even in the elementary conditions.

I find neither of these interpretations satisfying or persuasive. In fact, I think they both fail because they severely underestimate the complexity and sophistication of ordinary people’s implicit understanding of risk, chance, and justice.

* * *

Take the experiment at hand. Consider an ordinary person’s perspective: as a principal, the experiment forces him to cede decision-making authority over whether to take a chance on a lottery or a safe option to an agent. If the agent chooses the lottery, the principal is exposed to the (presumably remote) chance of a large payout or the more likely outcome of a loss. If the agent chooses the safe option, the principal will enjoy a much smaller but far more certain positive payout. So far so good. But the experiment encourages principals to allocate “payment” to the agent with the hindsight of actual results: principals are both informed of the results of the lottery, whether they participated or not, and they have the discretion to award payment to their agent as they see fit. In other words, “payment for services” (the agent’s ex ante judgment) is rendered contingent, rather than contractual: it implicitly depends on the actual outcome of the experiment. And, from the perspective of hindsight, the agent who selected the safe option when the principal could have won much more money had he selected the lottery has made a bad decision. Ex post, the agent’s action has materially harmed the principal. Why should he not be punished for his mistake? Or, if you prefer to think of it this way, why should he not suffer some portion of the principal’s bad luck? It’s only fair.

Note, mind you, that nothing I have related above depends on the principal’s correct or incorrect understanding of the role of chance in the experiment’s outcome. Nevertheless, I find it hard to envision any group of participants the researchers were able to assemble being so dense as to fail to understand that the lottery drawing central to the experiment ties its outcome ineluctably to chance. Of course the outcome of a lottery depends on luck. That’s what lotteries are. Everybody knows this. Accordingly, if you asked them, every principal would acknowledge that the agent’s choice of lottery or safe option has absolutely no causal bearing on the eventual outcome and, therefore, that the agent can in no way be blamed for a poor result.

And yet principals in the experiment punished agents who chose wrong. What is going on here? How can we tie these seemingly contradictory understandings together?

* * *

One way to gain greater insight into this dilemma is to examine the motivations and incentives for agents in this experiment. As an agent, one bears no downside risk for making a decision on behalf of the principal. The agent will get paid one way or another. But the agent understands the contingent, ex post nature of his payment for services, too. Why would an agent not believe that, if he put the principal’s money at risk in the lottery and the principal won, the principal would not reward him with a greater payment than otherwise? Give him a big tip? Share the luck? Of course, if the principal loses the lottery, the agent should have every expectation of getting paid nothing, but this should not matter. The agent who chooses a fair lottery for his principal is also choosing to tie his payout (at least via the mechanism of normal human psychology) to chance as well.

But what of the agent who chooses the safe option on behalf of his principal? He is explicitly choosing to deny his principal the remote chance of a large payout in favor of a low risk, safe return. He is choosing a small, low risk payout for himself over a much riskier payout profile tied to the good or bad fortune of his principal. And the principal knows it. Hence, one might argue that a principal who pays an agent who chooses the safe option less if the lottery wins is punishing the agent for risk aversion, because it is the agent’s risk aversion which has, in retrospect, forced the principal to forgo a windfall of good luck.1

* * *

Unlike Mr. Dillow,2 I believe most people do have a solid if inarticulate understanding of the role of chance in their lives. Most ordinary folks can see the effects of good, bad, or indifferent luck in their own lives and the lives of their families and acquaintances very clearly, and most people understand that good and bad luck is relatively rare. They understand in a way that lucky people cannot that luck has absolutely nothing to do with talent, character, or “just deserts.” They think if only they’d had that lucky break, they could have been as successful and happy as the rich and famous people they see on television. They know that bad things can happen to good people, and good things can happen to the undeserving. They understand that they live in an indifferent or even hostile universe, and that they can fall prey at any moment to bad luck which can ruin or blight their lives or the lives of those they love. They are fatalists.

And because they are fatalists, most people have an instinctive ethos which says that because we cannot avoid the influence of luck in our lives, and good and bad luck happen for no rhyme or reason, therefore we must all take our chances. And, because we are all exposed to chance, and none of us fully deserve what happens to us because of it, we should share the burden or benefit of chance when it is suffered or enjoyed. This is why, I believe, we have so many stories like those of the executed Athenian generals Mr. Dillow recalls, so many examples through history of leaders and commoners alike punished (or rewarded) for things literally out of their control. These are sacrifices, yes, to an indifferent fate, but they are also methods of spreading the terrible burdens and benefits of ineluctable chance from the literally undeserving shoulders of the people it falls upon onto the broader shoulders of the community.

In other words, I think most of us believe, deep in our bones, that none of us blessed or cursed with fickle chance deserves to enjoy or suffer it alone. It is our common burden, because it is our common condition.

This is also known as justice.

Related reading:
C.J.F. Dillow, Misallocating blame (Stumbling and Mumbling, September 27, 2013)
Occupy Galt’s Gulch (May 8, 2012)
To Whom It May Concern (August 31, 2013)

1 I suspect there is an entire other blog post to be written following on from these remarks concerning their implications for principal–agent relations, risk aversion, and Wall Street. I will spare you it for the nonce.
2 Who is focusing on this example of “unjustifiable blame” because, methinks, he is so concerned with the phenomenon of unjustified reward; that is, the rewards and benefits accruing to (mostly socioeconomic) elites from good luck. But my intuition tells me, as I have attempted to articulate above, that most people are willing to let the lucky enjoy most of the fruits of their luck—perhaps because they hope they will have the same chance themselves one day—but not all of them. How much of the good and bad luck accruing to individuals should be redistributed to the larger community—and in what fashion: progressive taxes, social welfare systems, charity, etc.—is of course the central political question, and one which each society answers differently at different times.

© 2013 The Epicurean Dealmaker. All rights reserved.