Two drivers leave a bar at 1:00 AM. Both have had four drinks. Both get behind the wheel. Both run a red light at the same intersection, at the same speed, with the same level of impairment. One of them passes through the intersection and arrives home safely. The other one hits a child who had stepped off the curb.
Driver A gets a DUI, maybe. Loses their license for a while. Pays a fine. Goes to an alcohol education course. Their life continues, more or less, as it was.
Driver B gets charged with vehicular manslaughter. Goes to prison. Loses their family, their career, and the next several years of their life. When they get out, they carry a felony conviction that will follow them to every job application and housing form for the rest of their existence.
Here is the question that the philosopher Thomas Nagel posed in 1979, and that remains genuinely unresolved in the philosophical literature: what is the moral difference between these two people?
Both made the same decision. Both exercised the same degree of negligence. Both ran the same red light with the same blood alcohol level and the same disregard for other people's safety. The most significant variable separating their fates is whether a child happened to be in the crosswalk at that particular moment, and neither driver had any way of knowing or controlling that fact.
And yet. Ask anyone on the street whether Driver B deserves a harsher punishment than Driver A, and most people will say yes without hesitation. Of course they deserve worse. They killed someone. The fact that the other driver didn't kill someone is, in the popular moral imagination, not evidence that they were equally negligent. It is evidence that they were luckier. And luck, we believe, should not determine moral blame.
Except it does. Constantly. In nearly every system of moral judgment humans have ever constructed.
Nagel called this phenomenon "moral luck," and the term was deliberately chosen to sound like an oxymoron. Morality, in the way we typically think about it, is supposed to be the domain of human life where luck does not apply. You are supposed to be judged for your choices, your character, your intentions. The things within your control. Luck governs outcomes, fortune, circumstance. Morality governs the will. Kant made this argument more or less explicitly: the only thing that is unconditionally good is a good will, and a good will remains good regardless of what it achieves or fails to achieve.
This sounds right. It feels right. And the moment you start pulling on the thread, the entire fabric comes apart.
The Four Layers of Luck
Nagel identified four categories of moral luck, and each one peels back another layer of the assumption that we are the authors of our own moral character.
The first is resultant luck, which is the driver example above. Two people perform the same action, and outcomes diverge based on factors beyond their control. We treat them differently anyway. The legal system formalizes this treatment: attempted murder carries a lighter sentence than completed murder in virtually every jurisdiction, even though the moral culpability of the attempt is, by the logic of the Control Principle, identical. The only difference is whether the bullet hit.
The second is circumstantial luck. This is the luck of what situations you happen to face. Nagel's most striking example involves Nazi Germany. An ordinary German citizen in 1933 who was transferred to Argentina by their employer before the war never faced the moral test of whether they would collaborate, resist, or look away. A nearly identical person who remained in Germany did face that test, and many of them failed it. The person in Argentina is not morally praised for avoiding collaboration. They simply never encountered the circumstances that would have tested their character. And we have no way of knowing what they would have done if they had stayed.
This one cuts deeper than resultant luck, because it implicates not just outcomes but the entire terrain on which moral judgments are made. We judge people for what they do. But what they do is profoundly shaped by what they encounter. A person who grows up in a stable household, in a safe neighborhood, with adequate resources, simply never faces the same set of pressures that someone born into poverty, violence, and deprivation faces daily. The first person may be, in some abstract sense, no more virtuous than the second. They just never had their virtue tested by the same exam.
The third category is constitutive luck, which is the luck of who you are. Your temperament, your genetic predispositions, your neurological wiring, the traits that make up your character. These are not things you chose. Nobody selects their baseline level of impulsivity, their capacity for empathy, their threshold for anger. These are determined by a combination of genetic inheritance and early developmental environment, neither of which is under the control of the person who ends up possessing them. And yet, we treat character traits as though they are achievements. We praise the patient person for their patience and blame the impulsive person for their impulsivity, as though both could have chosen otherwise from the same starting conditions.
The fourth and most radical category is causal luck, which is really just the problem of free will wearing a different hat. If your actions are the product of prior causes (your brain chemistry, your upbringing, your experiences, your genes, the entire causal chain stretching back to conditions that existed before you were born), then in what sense are you the ultimate author of any of them? Nagel does not resolve this question. He simply points out that if you take the Control Principle seriously and follow it all the way down, you end up in a place where no one is morally responsible for anything, because there is nothing that is ultimately within anyone's control.
This is the part where most people stop and say: well, that's absurd, so the argument must be wrong somewhere. And that reaction is understandable. But it is worth sitting with the discomfort rather than dismissing it, because the people who have tried to identify exactly where the argument goes wrong have not had an easy time of it. The premises are individually plausible. The logic connecting them is valid. The question is which premise you are willing to reject, and every available rejection has consequences that are uncomfortable in their own right.
The Compatibilist Reply
The strongest philosophical response to the moral luck problem comes from compatibilists, and it deserves serious engagement because it identifies a genuine weakness in the argument I am making.
John Martin Fischer and Mark Ravizza, in their 1998 book Responsibility and Control, argue that moral responsibility does not require ultimate authorship of the kind Nagel's argument seems to demand. What it requires, they claim, is "guidance control": the behavior in question must issue from a mechanism that is (a) the agent's own, and (b) appropriately responsive to reasons. On their account, you are morally responsible for an action if, at the moment of acting, you were capable of recognizing and responding to moral reasons, even if the deep causal history behind your character is something you did not choose.
This is a genuinely powerful response, because it redraws the boundary of what "control" means. Instead of asking whether you are the ultimate cause of your character (which nobody is), it asks whether your character, as it actually exists, is the kind of thing that responds to reasons. A person who drives drunk had the capacity to recognize that driving drunk is dangerous and to act on that recognition. The fact that their impulsivity has a causal history does not change the fact that, at the moment of decision, reasons were available to them and they failed to respond to those reasons. Fischer and Ravizza would say this is enough for moral responsibility.
I find this response partially convincing. It does a good job of rescuing everyday moral responsibility from the most radical implications of causal luck. Most of us, most of the time, are reasons-responsive agents, and treating us as such is both practically necessary and, I think, broadly justified.
But the compatibilist response has a problem of its own, and it is a problem at the margins, which is where the concept of desert does the most damage. The capacity for reasons-responsiveness is itself unevenly distributed. It varies with mental illness, with addiction, with neurological development, with childhood trauma, with the sheer accumulated weight of circumstances that make rational deliberation harder for some people than for others. Fischer and Ravizza acknowledge this by arguing that responsibility comes in degrees, which is a significant concession. Once you accept that responsibility admits of degrees, and that the degree of someone's responsibility is influenced by factors outside their control, you have let luck back into the system. You have not eliminated the moral luck problem. You have softened it.
And the legal system does not operate in degrees. A 16-month sentence versus a 5-month sentence for the same crime is not a graduated response calibrated to the agent's degree of reasons-responsiveness. It is a blunt instrument applied uniformly to people whose capacity for rational deliberation varies enormously, and justified by a concept of desert that the compatibilist response itself concedes is more complicated than our practices assume.
The Meritocracy Problem
If the concept of moral desert is complicated by luck at every level, then the concept of economic desert is in even worse shape.
The meritocratic ideal, the idea that people deserve their economic outcomes based on their talent and effort, requires at minimum that talent and effort be things that belong to the individual rather than to the circumstances that produced them. But they are not. Your talent is a function of your genetics and your developmental environment. Your effort is a function of your temperament, which is itself a function of your genetics and your developmental environment. Your opportunity to develop and deploy both talent and effort is a function of your socioeconomic starting position, the quality of your early education, the stability of your family, the safety of your neighborhood, the presence or absence of lead in your drinking water, and a thousand other variables that were established before you could form memories, let alone make choices.
This does not mean that individual variation in effort does not exist. It clearly does. Some people work harder than others, given similar circumstances. But the capacity for hard work is itself unevenly distributed by factors outside anyone's control, and the circumstances in which hard work is rewarded versus futile are also unevenly distributed. The person who works 70-hour weeks building a startup in Silicon Valley and the person who works 70-hour weeks at two minimum-wage jobs to keep the lights on are both working hard. One of them is going to be celebrated as a genius of enterprise. The other one is going to be described as evidence that hard work alone is not enough. And the variable that separates their outcomes is not effort. It is circumstance.
John Rawls formalized a version of this argument in 1971 with A Theory of Justice. His framework is more precise than the popular summary often suggests, and getting the precision right matters. Rawls does not simply argue that meritocracy is illegitimate because talents are unearned. His claim is more structural: the basic institutions of society should be arranged so that inequalities arising from the natural and social lotteries (the unearned distribution of talents, temperament, and social position) are regulated by principles that free and equal persons would choose under fair conditions. The result is his difference principle, which holds that social and economic inequalities are permissible only to the extent that they benefit the least advantaged members of society. Inequalities are allowed, on this view, but only when they work to improve the position of those at the bottom. A doctor earning more than a janitor is acceptable if the economic structure that produces that gap also makes the janitor better off than they would be under strict equality.
This is subtler than "nobody deserves their income." Rawls is not proposing an enforced equality of outcomes. He is proposing that since the distribution of talents is, from a moral perspective, arbitrary (a product of the natural lottery rather than of choice), a just society should treat the total pool of natural talents as a shared asset and structure its institutions so that everyone benefits from their unequal distribution. The difference principle is not a claim about individual desert at all. It is a claim about institutional design.
The meritocratic ideal only looks appealing from the position of someone who already has the tools to succeed within it. From the position of someone who does not, it looks like a justification for their suffering dressed up in the language of personal responsibility. That Rawls' argument has faced serious objections (most notably from Robert Nozick, who argued that individuals have rights to their talents and their products that cannot be overridden by distributive schemes) does not make it wrong. It means the debate is real. But notice that even Nozick did not argue that people deserve their talents. He argued that people have rights to them, which is a different claim with different implications. The moral arbitrariness of the talent distribution is not something the major parties to this debate seriously contest.
The Punishment Question
If desert is philosophically unstable, what happens to punishment?
The United States incarcerates roughly 1.8 million people in adult facilities, at a rate of 541 per 100,000 residents. That is the fifth highest incarceration rate in the world, behind El Salvador, Cuba, Rwanda, and Turkmenistan. It is roughly five times the rate of the United Kingdom, eight times the rate of Germany, and nine times the rate of Japan. The average burglary sentence in the United States is 16 months. In Canada, it is 5 months. In England, it is 7 months. A first-time federal drug offense in the US carries a mandatory minimum of 5 to 10 years. In most of Europe, the same offense would result in at most 6 months.
The standard justification for this system is desert. People who commit crimes deserve to be punished, and the severity of the punishment should reflect the severity of the crime. The most philosophically serious version of this argument comes from Michael Moore, whose 1997 book Placing Blame argues that retributive desert is not merely one justification for punishment among others but the only genuine justification. Moore claims that punishment of the guilty is an intrinsic good, and that our strong emotional response to wrongdoing (the guilt we feel when we do wrong, the indignation we feel when others do) constitutes evidence of a moral fact about what wrongdoers deserve.
This is a more formidable position than it might initially sound. Moore is not merely saying punishment feels satisfying. He is making a claim about moral epistemology: that our retributive emotions are a reliable guide to moral reality, in the same way that perceptual experience is a reliable guide to physical reality. If you witness a terrible crime and feel that the perpetrator deserves to suffer, Moore would say that feeling is tracking something real about the moral world.
I have two responses. The first is that the reliability of retributive emotions as moral evidence is precisely what the moral luck problem calls into question. We feel stronger retribution toward Driver B than Driver A, even though their moral culpability is identical by any measure that respects the Control Principle. We feel stronger retribution toward a murderer than toward a person whose attempted murder failed only because the gun jammed. If retributive emotions are evidence of moral facts, they are evidence that tracks outcomes rather than culpability, which means they are evidence contaminated by luck. Moore's appeal to moral emotions proves too much, because those emotions do not discriminate between factors within and beyond the agent's control.
The second response is practical. The retributive justification requires that the person who committed the crime could have done otherwise, that their decision was a free and autonomous exercise of will rather than a predictable product of circumstances. Even compatibilists who rescue moral responsibility from the most radical implications of determinism concede that responsibility comes in degrees and is shaped by factors outside the agent's control. A system calibrated to actual degrees of culpability would look very different from the system we have.
I am not arguing that prisons should not exist. I am arguing that the philosophical foundation on which we have built the American carceral system is significantly less stable than our practices assume, and that this matters for how we think about what prisons are for.
If people do not "deserve" punishment in the deep metaphysical sense that pure retributive justice requires, then punishment must be justified on other grounds. The two most common alternatives are deterrence (punishment discourages future crime) and incapacitation (imprisonment prevents dangerous people from harming others while they are locked up). Both of these are consequentialist justifications. They do not require desert. They require only that the punishment produces better outcomes than the absence of punishment.
And here is where something worth noting happens. If you design a criminal justice system around consequences rather than desert, you end up with a different system. You end up with shorter sentences, because the deterrent effect of imprisonment is primarily a function of the certainty of being caught rather than the length of the sentence. You end up with more investment in rehabilitation, because reducing recidivism produces better outcomes than warehousing people for decades.
Norway has an incarceration rate of about 54 per 100,000. The United States has a rate of 541. Norway's recidivism rate is around 20%. The United States' recidivism rate, depending on the study and the time horizon, is somewhere between 44% and 83%. I want to be careful with this comparison, because it is easy to overstate what it proves. Norway is a small, wealthy, ethnically more homogeneous country with a robust social safety net. Its lower crime and recidivism rates are not solely a product of its criminal justice system. They reflect broader social conditions including lower inequality, universal healthcare, and stronger public education. You cannot transplant Norway's prison system into the United States and expect Norwegian outcomes without also transplanting the social context.
But the comparison is still instructive, because it demonstrates that a system designed around "what works" rather than "what is deserved" can produce outcomes that are better on every metric we claim to care about: lower crime, lower recidivism, lower cost, and less human suffering. The fact that Norway's outcomes are partially a product of its broader social structure actually reinforces the point: the factors that produce crime are largely social and circumstantial, which is exactly what you would expect if the moral luck problem is real.
The Objection
The most common objection to this line of reasoning goes something like: if nobody deserves anything, then nobody deserves praise either, and the entire moral framework collapses. If the person who donates a kidney does not deserve moral credit because their generosity is a product of constitutive luck, and the person who commits assault does not deserve moral blame for the same reason, then we have no basis for distinguishing between them at all. Morality becomes meaningless.
I take this objection seriously because it identifies the real cost of my argument. If you follow the logic of moral luck to its endpoint, you do lose something meaningful. The feeling that good people deserve good outcomes and bad people deserve bad outcomes is not just a cognitive bias. It is, for most people, a foundational moral intuition. Abandoning it entirely would require a radical reorganization of how we relate to one another, and I am not confident that such a reorganization would be desirable even if it were possible.
But I think the objection confuses two different things: the metaphysical question of whether desert is coherent, and the practical question of whether moral evaluation is useful.
You can reject desert and still maintain a functional moral framework. You can evaluate actions as harmful or beneficial without claiming that the person who performed them "deserves" to suffer or flourish as a result. You can respond to harmful behavior with interventions designed to prevent future harm without those interventions being grounded in the idea that the person has earned their suffering. You can praise prosocial behavior and criticize antisocial behavior as a social mechanism for reinforcing norms without needing a metaphysical backstop of ultimate authorship to justify it.
In fact, this is largely what we already do in other domains. When a dog bites someone, we do not agonize over whether the dog "deserves" to be punished. We ask what intervention will prevent future bites. When a bridge collapses, we do not ask whether the bridge "deserved" to fail. We investigate the cause and implement changes to prevent it from happening again. The introduction of desert into moral reasoning about humans is a distinctive feature of our psychology, and when you examine it closely, it contributes less to our actual moral practices than we assume. Most of what we do in the name of desert could be equally well justified (and often better justified) on consequentialist grounds.
What desert actually contributes, I would argue, is emotional satisfaction. It feels right that bad people suffer. It feels right that good people prosper. The just-world hypothesis, the cognitive bias that leads people to believe that outcomes are deserved, is one of the most robustly documented phenomena in social psychology. We want the world to be fair. We want suffering to be earned and success to be merited. And when the world fails to cooperate with this desire, we adjust our perception of the people involved rather than our model of how the world works. The poor person must have done something wrong. The rich person must have done something right. The prisoner must deserve to be there.
This is comforting. It is also, given what we know about the role of luck in human outcomes, almost certainly wrong.
Where This Leaves Us
I am not writing this to convince anyone that we should dismantle every institution built on the concept of desert. The practical case for acting as if people have a meaningful degree of moral responsibility is strong enough that I accept it myself, even while recognizing its philosophical limitations. Fischer and Ravizza's compatibilism is a workable approximation for most of everyday life. We are reasons-responsive agents. Treating each other as such is not a delusion. It is a necessary and largely beneficial social practice.
But there is a difference between using desert as a convenient approximation and treating it as a bedrock moral fact. The problems arise when we move from the everyday ("you should have known better") to the institutional ("therefore you deserve 16 months in a cage for stealing a television"). At the institutional level, desert is being asked to do work it cannot do. It is being asked to justify enormous differences in human suffering as though those differences are the natural and correct consequences of choices made by fully autonomous agents who could have done otherwise. And that is a claim that the philosophical literature on moral luck makes very difficult to sustain.
When desert is invoked to justify a 16-month burglary sentence in the United States versus a 5-month sentence in Canada for the same crime, something other than principled moral reasoning is happening. When desert is invoked to justify a system that incarcerates Black Americans at roughly five times the rate of white Americans, despite comparable rates of drug use, something other than principled moral reasoning is happening. When desert is invoked to justify the idea that a hedge fund manager earning $400 million a year is being rewarded for their contribution to society while a home health aide earning $28,000 is being rewarded for theirs, something other than principled moral reasoning is happening.
In each of these cases, the language of desert is being used to naturalize an arrangement that is contingent, constructed, and mutable. The arrangement could be otherwise. Other countries have demonstrated that it can be otherwise. The reason it is not otherwise here is not that the current arrangement is the one people deserve. It is that the current arrangement is the one that exists, and the concept of desert is the story we tell to make the existing arrangement feel inevitable.
Nagel knew this, I think. His 1979 essay does not conclude with a solution. It concludes with the observation that the problem of moral luck is genuine, that it resists resolution, and that the tension between the Control Principle (which says luck should not affect moral judgment) and our actual moral practices (which are saturated with luck at every level) is simply something we have to live with.
I respect the honesty of that position. But I think we can do slightly better than just living with it. We can, at minimum, let the discomfort of the problem inform the design of our institutions. We can build systems that assume less about desert and ask more about outcomes. We can fund the interventions that demonstrably reduce harm instead of the ones that satisfy our desire for retribution. We can notice when the language of "deserving" is being used to justify suffering and ask whether the suffering is actually doing anything useful, or whether it is simply making us feel like the world is more orderly than it is.
Nobody chose to be born. Nobody chose the brain they think with, the temperament they manage, the neighborhood they grew up in, or the century they inhabit. The range of possible lives available to a person is constrained before they draw their first breath by forces that have nothing to do with merit, virtue, or will. A moral framework that ignores this is not principled. It is convenient. And the people for whom it is most convenient are, predictably, the people for whom the dice came up favorably.
I think we can build something better than convenience. I do not know exactly what it looks like. But I am fairly certain it does not look like what we have now.
The core philosophical arguments here draw from Thomas Nagel's "Moral Luck" (1979), Bernard Williams' essay of the same name (1976), John Rawls' "A Theory of Justice" (1971), and John Martin Fischer and Mark Ravizza's "Responsibility and Control" (1998). The retributivist position is best represented by Michael Moore's "Placing Blame" (1997). Robert Nozick's critique of Rawls appears in "Anarchy, State, and Utopia" (1974). The incarceration statistics are from the Bureau of Justice Statistics, the World Prison Brief, and the Sentencing Project. The Stanford Encyclopedia of Philosophy entries on moral luck, retributive justice, and justice and bad luck are all thorough and freely available for anyone who wants to go deeper.