— economics · policy · moral philosophy
fair warning before we begin: honest moral philosophy sometimes leads to conclusions that feel wrong before you understand why they're right. that discomfort isn't a sign you've gone off track. it's often a sign you're getting somewhere.
most of us operate on intuitive moral codes. we encounter a situation, it triggers a feeling—fair, unfair, cruel, kind—and we act on that feeling. this works well enough for everyday life. but intuition has a serious weakness: it's inconsistent. and inconsistency is a pretty serious flaw in a moral framework you'd want to defend publicly.
consider the trolley problem. most people say they would pull a lever to divert a runaway trolley from five people to one. fewer say they would push a large man off a bridge to stop the same trolley, even though the arithmetic is identical. the emotional framing changes the answer even when the underlying situation hasn't changed. this isn't a quirk—it's the default mode of human moral reasoning.
now consider a more mundane example.
ask people whether they support surge pricing—a rideshare company charging more during a rainstorm or on new year's eve—and you'll get a lot of negative reactions. people find it exploitative. opportunistic. some find it almost unethical.
then ask them whether they support happy hour.
nearly everyone loves happy hour. cheaper drinks during slow periods? obviously fine. good for business, good for customers, good for the neighborhood bar.
but here's the thing: happy hour is surge pricing, viewed from the other direction. if you set the baseline price at the peak price, the discount becomes the off-peak adjustment. if you set the baseline at the off-peak price, the premium becomes surge pricing. the underlying economics are identical. the only difference is the story we tell about which price is "normal."
this means the moral intuition isn't tracking anything real about the economic situation. it's tracking a framing. and a moral intuition that reverses itself based on framing isn't doing much moral work.
here's another case of the same phenomenon, and this one is more uncomfortable.
if you walked past a child drowning in a shallow pond and did nothing—when you could easily have waded in at trivial cost to yourself—most people would consider you morally monstrous. the philosopher peter singer made this intuition famous, and it's hard to argue with. you could save a life. you didn't. that seems clearly wrong.
now: you are reading this right now, probably on an expensive device, possibly drinking expensive coffee. and you know—with complete certainty, even without knowing their names or locations—that children are dying today from malnutrition and preventable disease in other parts of the world. organizations like givewell have done the rigorous work of identifying charities that can save a child's life for a few hundred dollars, sometimes less. if you've spent money on luxuries this month that you could have donated instead, you have, by singer's logic, walked past the drowning child.
are you morally monstrous?
most of us resist that conclusion. and that resistance is itself the point. our moral intuitions are not reliably tracking human welfare in any systematic way. they're doing something more local, more emotional, more tied to proximity and visibility. we respond to the child in front of us and not to the child on the other side of the world, even when the stakes are identical or higher.
i'm not raising this to condemn anyone—i include myself. i'm raising it because it establishes something important before we go any further: all of us are already operating with moral intuitions that are dramatically inconsistent when examined rigorously. acknowledging that is the precondition for thinking more carefully.
before getting to the question i actually want to explore, i want to introduce a distinction that i think is underused in policy debates: the difference between efficiency effects and equity effects.
an efficiency effect is about the size of the pie—whether available resources are being allocated to the people and uses that value them most. an equity effect is about the distribution of that pie—who has how much.
here's a simple illustration. suppose alice has an orange but prefers apples. bob has an apple but prefers oranges. this is an inefficiency—a suboptimal allocation. neither person has "too much" or "too little." the problem is purely that resources aren't going to the people who value them most. the solution is a voluntary trade, which makes both parties better off without taking anything from anyone. economists call this a pareto improvement.
now suppose bob has a mountain of fruit and alice has almost none. that's an equity problem. the disproportion is the issue. the solution here isn't a clever trade. it's redistribution.
confusing these two categories leads to bad policy. consider carbon emissions. people emit more CO₂ than is socially optimal because they don't pay the full cost—the cost of climate damage gets spread across society while the benefits of cheap energy accrue to the individual emitter. this is a classic market failure: a negative externality. the efficient solution is a carbon tax that makes emitters pay the true social cost of their emissions.
a common objection to carbon taxes is that they're regressive—they hit lower-income households harder because those households spend a higher share of their income on energy. this is a real concern. but it's an equity concern, not an efficiency concern. the solution is not to abandon the carbon tax—that would leave the efficiency problem unsolved. the solution is to address the equity problem directly: a dividend or universal basic income that offsets the burden on lower-income households. alaska's permanent fund does something like this. so does the idaho grocery credit, which despite its name functions as unrestricted cash.
when we reach for equity tools to solve efficiency problems, or efficiency tools to solve equity problems, we end up with policies that do neither job well—and often create new problems in the process.
with that framework in place, i want to return to a question a friend recently asked me: do i support the civil rights act of 1964—specifically its provisions banning discriminatory hiring based on race, sex, and other protected characteristics?
my instinct was to give a quick answer. but i think the question deserves better than that.
let me be unambiguous about the historical context first. the conditions that motivated the act were genuinely horrific. people in small southern towns were systematically denied employment, housing, and basic economic participation on the basis of race, with no practical alternatives and often with violent enforcement of those conditions. anyone who engages with this question without acknowledging that is not engaging in good faith. the moral urgency of intervention was real. the suffering was immense.
with that foundation laid, let me apply the consistency check we established earlier.
consider ladies' night at a bar. most people find this entirely unobjectionable—a charming discount for women. but apply the framing lesson from earlier: ladies' night is equally describable as a surcharge on men. it's sex-based price discrimination in which men pay more than women for the same service. if you find the "discount for women" framing perfectly acceptable but would find "surcharge on men" troubling, you're doing exactly what we did with surge pricing and happy hour: letting the story change your moral judgment about an identical economic fact.
similarly, google for many years ran diversity hiring programs that systematically favored applicants from underrepresented racial groups—meaning they systematically disfavored others. many people who strongly support anti-discrimination law also strongly supported google's program. but if it's legally and morally acceptable for an employer to prefer candidates of certain racial backgrounds, it becomes harder to argue that it's categorically impermissible for an employer to disfavor them. the principle cuts both ways.
consider one more extension of the same logic. should there be laws compelling consumers to shop at minority-owned businesses—prohibiting them from choosing a competitor on the basis of the owner's race? most people would find that idea somewhere between absurd and authoritarian. and yet the economic structure is nearly identical to employment discrimination law. when you hire someone, you are purchasing their labor. when you shop somewhere, you are purchasing their goods or services. in both cases, a private economic transaction is being made. if we're uncomfortable compelling a consumer's purchasing decisions on anti-discrimination grounds, it's worth asking why we're more comfortable compelling an employer's hiring decisions on the same grounds—and whether the distinction we're drawing is principled or merely habitual.
these aren't gotchas designed to produce a particular conclusion. they're invitations to figure out what principle is actually doing the work in our intuitions—and whether we're applying it consistently. the point isn't that discrimination is fine. the point is that the question of what government should compel in private economic transactions is genuinely more complex than a snap judgment suggests.
let's ask the question directly: is discrimination in hiring an efficiency effect or an equity effect?
suppose bias causes you to end up with your second-best option—a job that pays $10,000 a year less than you would have earned otherwise, whether that gap shows up in flat compensation, job satisfaction, skill acquisition, or some combination. that's effectively like being born poorer. it diminishes your total resources through no fault of your own. that is clearly an equity effect, not an efficiency effect.
now consider two interventions:
option 1: a law that reverses this—compelling the employer to hire you at the higher wage.
option 2: $11,000 a year in cash—more than the $10,000 differential, making you unambiguously better off.
which is better for you? the cash, clearly. it gives you the freedom to allocate resources according to your own preferences. maybe you don't want to work for an employer who had to be legally compelled to hire you. maybe you'd rather start a business, retrain for a different field, or move somewhere with better opportunities. all the compounding career benefits of the better job—trajectory, network, experience—can in principle be captured as a net present value and folded into the cash amount. whatever that job was worth over a lifetime, you can just give the person that.
this isn't subjective or debatable. it's revealed preference. no intervention can be called more moral than the alternative that the person subjected to the harm would themselves prefer. the cash wins by that standard, definitionally.
now consider an analogy. suppose instead of facing racial bias, a person is born with a genetic predisposition to depression, or a learning disability, or any other condition that reduces their productivity or narrows their options. this too acts as a tax on their lifetime earnings—less income, less opportunity, less economic security, through no fault of their own. nobody suggests we solve this by compelling employers to hire people regardless of whether they can perform the job. nobody argues the nba should be forced to sign players who can't dribble. we intuitively understand that the right response is a social safety net—redistribution that compensates people for the bad hands they were dealt.
racial bias works the same way economically, even though it feels different. it feels different because we sense that discrimination—unlike a genetic condition—could in principle just be eliminated. ban it, and the problem goes away. but this is illusory. you might think: the discriminated-against person is just as capable of doing the job as anyone else, so there's no real cost to forcing the hire. but that's not the only cost. there is also a cost to the employer in terms of their dissatisfaction with the arrangement. you may find that cost morally repugnant—and that's understandable—but it is still a real cost, and it is absolutely part of the total size of the pie. banning discriminatory hiring doesn't eliminate that cost. the cost is simply whatever the employer would have required to make the transaction voluntary—that's definitional, not speculative. compulsion doesn't make it disappear; it just makes it invisible. and invisible costs are still costs.
the forced hire also creates costs on the other side. suppose you asked the biased employer: "what would we have to pay you to willingly hire this person?" they might say $15,000 a year. that means the compulsion imposes a $15,000 annual burden that doesn't benefit anyone—it simply disappears as deadweight loss. this is the characteristic problem with price controls generally, whether rent control, minimum wage floors, or anti-discrimination mandates. the harm of the underlying bias is real. but the policy instrument generates its own friction on top of it.
contrast this with well-designed tax policy. pigouvian taxes—like a carbon tax—are specifically designed to correct market failures with zero or negative deadweight loss, because they address the actual distortion rather than imposing a price control. land value taxes are even more efficient: because land is fixed in supply, taxing it doesn't reduce the amount of it, meaning essentially no deadweight loss at all. anti-discrimination mandates in hiring are not in this category.
there's also a practical dimension worth naming. proving discriminatory intent in any specific hiring decision is extraordinarily difficult. a friend of mine—an economist and former data scientist at a large tech company you've heard of—described to me the enormous resources his employer spent on defensive monitoring: documentation systems, review processes, legal infrastructure, all specifically designed to reduce litigation exposure. this overhead is pure waste from a welfare perspective. it doesn't help the people who were discriminated against. it helps lawyers. a broad redistributive policy that quietly and invisibly lifts everyone's floor has none of this overhead. it just does its work.
i want to acknowledge something directly. part of what makes this conversation so difficult is that the underlying motive—racial bias, gender bias—strikes most people as so fundamentally abhorrent that an outright legal prohibition feels like we're actually solving the problem. most of us don't walk around with economics degrees in our heads. we don't systematically ask whether a given policy instrument is the right tool for the job, or whether we're confusing an equity effect with an efficiency effect. we see something wrong, we want it to stop, and a law that says "stop it" feels like it's doing exactly that.
there's also an association that's worth naming and then carefully distinguishing. discrimination that utilizes the full coercive power of the state—jim crow laws, apartheid, legally mandated segregation—is categorically different from discrimination exercised by private actors in a market. jim crow didn't merely permit discrimination; it commanded it, backed by the force of law and often by violence. government has unique coercive power, and government-mandated discrimination represents a different and more profound kind of wrong than a biased private employer making a bad hiring decision.
this distinction matters because some of the moral urgency we feel around the 1964 act comes from conflating these two things. when we hear "discrimination in hiring," our minds can conjure the full horror of state-enforced racial exclusion. and that horror is real and warranted. but a private employer's biased decision, however harmful and morally wrong, is not the same thing—and the appropriate remedy may not be the same either.
there's a further complication worth acknowledging honestly, because a thoughtful critic pushed me on this after reading an earlier draft.
if anti-discrimination law is largely unenforceable in practice—which it largely is, for the reasons described above—then it isn't really functioning as a price control at all. it's functioning as a norm-setting signal. and norm-setting signals can do real work that cash transfers cannot.
there's also a deeper mechanism here. discriminatory preferences aren't fixed. they're often the product of conditioning, limited exposure, and social environment—not immutable features of human psychology. gordon allport's contact theory, developed in the 1950s, showed that intergroup contact under the right conditions reliably reduces prejudice. the empirical literature since has largely borne this out. legal prohibition creates conditions for contact and norm updating that gradually erode the underlying preferences themselves.
this matters for the economic argument. a rent control law doesn't change landlords' preferences about rent—it just overrides them, generating deadweight loss for as long as it remains in force. an anti-discrimination law that successfully shifts social norms may actually reduce discriminatory preferences over time, which means the behavioral distortion shrinks as the law does its cultural work. if that's true, the deadweight loss argument weakens considerably—not because the economics is wrong, but because the law is operating through a different mechanism than a standard price control.
i find this argument genuinely persuasive, and it's a meaningful update to the piece. the case for anti-discrimination law as a norm-setting and preference-shaping instrument is stronger than the case for it as a direct behavioral mandate. whether it's the most efficient instrument for that purpose is still a fair question—a well-designed pigouvian tax on discriminatory behavior might do the same norm-shifting work more precisely. but the historical record does suggest the law has moved things in ways that persuasion alone did not.
none of this is an argument against government action. it's an argument for thinking carefully about what mechanism the action is actually operating through—and designing policy around that mechanism rather than around a story we tell ourselves about what we're accomplishing.
we started with peter singer's drowning child, and i want to return there briefly.
singer's argument reveals that most of us already hold moral beliefs we systematically fail to act on—not because we're evil, but because human moral psychology is wired for proximity and visibility rather than abstract welfare maximization. we know this about ourselves. the question is whether we're willing to use that knowledge to design better policy rather than just better-feeling policy.
the 1964 act was a response to a genuine and urgent moral catastrophe. the question of whether it was the optimal policy instrument for that problem is separate—and asking it is not the same as denying the catastrophe or tolerating what caused it. rigorous moral reasoning demands that we hold both things at once: the full weight of the historical wrong, and a clear-eyed assessment of whether the tools we've chosen to address it are actually doing the job.
that's harder than a snap judgment. but it's the only way to actually help.
some people would call you a monster for questioning laws that prohibit racially discriminatory hiring practices. but an analytically rigorous person might turn that around: you're a monster if you don't support a universal basic income—the policy instrument that would actually address the underlying inequity directly, efficiently, and without deadweight loss. and yet a comfortable majority of the population readily dismisses the idea of a universal basic income without a second thought.
funny how that works.