Oppression and production are competing explanations for wealth inequality.

You would like to go to the beach tomorrow if it's sunny, but aren't sure whether it will rain; if it rains, you'd rather go to the movies. So you resolve to put on a swimsuit and a raincoat, and thus attired, attend the beach in the morning and the movies in the afternoon, regardless of the weather. Something is wrong with that decision process,* and it's also wrong with the decisions made by many supposedly systemic approaches to philanthropy: it does not engage with real and potentially resolvable uncertainty about decision-relevant facts.

Different popular philanthropic programs correspond to very different hypotheses about why people are doing wealth inequality, much like swim trunks and a trip to the movies represent different hypotheses about the weather. Instead of working backwards from the proposals to the hypotheses, I will lay out what I think are the two main hypotheses worth considering, and reason about what someone might want to do if that hypothesis were true. This is not because I want to tell you what to do, but to clarify that any time you think that something in particular is a good idea to do, you are acting on a hypothesis about what's going on.

The ideas of charity and philanthropy depend on the recognition of inequality; otherwise it would just be called "being helpful." The persistence of wealth inequality, in turn, depends on many people working together to recognize and enforce individual claims on private property.

If the mechanism of private property tends to allocate capital to its most productive uses, then incentives are being aligned to put many people to work for common benefit. But if wealth does not correspond to productive capacity - i.e. the people with the most are not those best able to use it - then, assuming diminishing marginal returns to wealth, coordination towards persistent wealth inequality comes from a self-sustaining misalignment of incentives, i.e. conflict.

The economic ideology taught in introductory microeconomics courses, which is assumed by many formal analyses of how to do good at scale, including much of Effective Altruist discourse, tends to make assumptions consistent with the means of production hypothesis, so if we are considering making decisions on the basis of that analysis, we want to understand which observations would falsify that hypothesis, and which beliefs are incompatible with it.

You walk into a workshop, and see someone holding a hammer. You can infer that this is because there is some hammering to do right now, and the holder is competent to do it. Someone else has a saw, and you make a similar inference. In this context, the unequal distribution of production goods is part of how things get made; wealth inequality is a part of the means of production. If a workshop did not allocate tools in a way that justified those inferences - if perhaps you observed one person with a hoard of wrenches doing nothing while others used their bare hands as best they could - then you might infer the existence of a conflict between the wrenchmaster and the other laborers, and you would expect that workshop to do a worse job if called upon to make something. On the other hand, if someone with a hoard of wrenches were freely lending out the wrenches when appropriate, seemed like an especially good judge of which wrench (if any) is appropriate for which job, and made sure people put the wrenches back instead of putting them down at random in hard-to-find places, then you might not think worse of the workshop for its wrenchmaster.

The hypothesis that wealth inequality is part of the means of production has moral and strategic implications for charity.

From a global utilitarian perspective, having much more than others is not on its own a reason to transfer wealth to them. Instead, you should expect the return you can get on reinvesting your wealth into profit-yielding enterprises to frequently be higher than the return they can get, so you might be able to make a more important gift to the future than to the present. Even when there is a large enough market failure to justify philanthropy, some amount of paternalism is warranted, because your wealth advantage corresponds to a way in which you know better than them. An exemplar for this perspective is Andrew Carnegie, who amassed a vast fortune improving the organization of steel production, and used some of that fortune to provide a public good, specifically the information good of public libraries. Readers who want his perspective in his own words might do well to read The Gospel of Wealth and his autobiography.

While the details of the return on investment calculation from the selfish perspective will be different, the basic tradeoffs are similar. Due to diminishing marginal returns, at some point it becomes so prohibitively expensive to solve your problems by buying commodity goods or even custom services that the most selfish thing to do is contribute to undersupplied public or coordination goods. For example, Elon Musk's interest in acquiring Twitter and relaxing its censorship regime - and creating Starlink - may be the selfish one of wanting to maintain access to lines of communication with sympathetic strangers (which has been important for things like his ability to find a compatible reproductive partner).

If, on the other hand, wealth inequality is mainly due to systemic oppression, i.e. coordination by an extractive class against producers, then the world looks very different. The simplest implication is that the possession of a fortune is no longer evidence that you know better than others. And before we can even generate the idea of charity under this framework, we run into a justification for a radical form of economic skepticism: what are we even doing when we try to buy a good?

Under the means of production hypothesis, the answer was straightforward: when I buy a good, I am sending a price signal which causes some combination of reallocation of resources to produce more of that good, and the reallocation of that good and its inputs away from those with the least productive use for them. On balance I should expect such price signals to enrich those alleviating scarcity by improving the efficiency with which scarce goods are produced. It follows analytically that under the oppression hypothesis, since the enrichment of producers doesn't happen, any price signals I send do not reallocate resources to produce more in-demand goods on net. There must be a loser, so either I am paying for a weapon to extract from others, or I myself am the target for extraction, i.e. I am being scammed. The pure oppression hypothesis implies that wealth has no real purchasing power for goods; at most it has an illusory or dramatic one.

I have enough money to pay a modest premium for high quality ingredients, and I really do seem to feel better after eating them, which is some evidence for the hypothesis that wealth inequality is part of the means of production. But a friend of mine lives nearby in public housing and cooks on a food stamp budget, and my millionaire housemate enjoys my friend's cooking more than mine. The friend in public housing has complained to the two of us that a much wealthier friend and potential donor to her nonprofit likes to take her out to eat at an expensive club with dismally bad food to waste her time, and won't actually financially support her programs, even the ones he's agreed are good ideas. This is not consistent with the story that money buys good things, but is consistent with the oppression hypothesis.

The pure oppression hypothesis is difficult to imagine. If wealth is nothing but a way to threaten others, and has no independent purchasing power, then it has no way to threaten anything outside of the system; it is a closed system of domination and those outside it can safely ignore it. The rule of the Roman Catholic church in Europe is not a perfect example, but provides a suggestive resemblance. The church made the most extreme metaphysical threats towards its constituents, mixed with what were in most cases mild physical threats if any. The very large sums of gold paid in indulgences or contributions to crusaders show how strongly motivated people were to get out from under this threat. People who rose in the ranks acquired more power to make or withdraw threats towards others, but were not supposed to correspondingly control more productive capital, and they were discouraged from reproducing.

From a global utilitarian view, on the oppression hypothesis, what should a rich person do? The arguments for paternalism or reinvestment do not apply here; your wealth does not imply that you are a good steward, because the allocation of resources does not conform to the function of meeting people's needs. You have no reason to think that you know better than others how to help them, and the idea of a return on investment is perverse. But needs are getting met somehow, so the coordination to do so must be happening outside the system of oppression.

One thing you might try to do in this situation is to use your position as someone validated by a system of oppression to invalidate it, e.g. by publicly setting your money on fire. (This differs from conspicuous consumption because it eliminates motive ambiguity; intentionally wasteful spending still pretends to be receiving something of value, while literally making a pile of cash and setting it on fire does not, so it sends a credible signal that you think the money is worse than useless.) Another thing you might do is try to deescalate threats towards others, in the hope that this frees up their capacity to solve problems, including the existence of the system of threats you're caught up in. In other words, cash transfers.

You might try applying some selection by concentrating your gifts on people with reputations as good actors within the system. The Bezoses seem to have done something like this, with MacKenzie Scott distributing money widely among nonprofits working on things that seem good, and Jeff Bezos making one-time $100 million grants to Van Jones and José Andrés. On the other hand, you might reasonably worry that the reputational system - or at least, the mechanism by which news gets delivered to you, a wealthy person - is part of the system of oppression. In that case, you might apply Rawlsian skepticism and simply try to help whoever is worst off, e.g. cash transfers to the global poor, programs to help prisoners, etc. But then you need to trust that you can pay for the cash transfers to actually happen, which is not clearly justified (remember, under this hypothesis money facilitates threatening people, not providing goods and services) - the best available option might be to wander around incognito looking for people who seem like they could use help but aren't seeking attention.

We live in a mixed economy, but it can't be a homogeneous mixture. Instead, there are details to investigate: who gets paid to produce, and who gets paid to destroy, under what circumstances?

This post was inspired by the state of public discourse on effective altruism, in which cash transfers to the global poor, paternalistic global health interventions, animal welfare interventions in explicit conflict with incumbent powers, and extremely high-leverage high-trust speculative AI design, are put on a single list as though the same set of assumptions could calculate an ROI for all of them, and the main thing that's left to do is pick from the list, or add items. This seems crazy to me like planning to put on swim trunks and a rain coat, and go to the beach in the morning and the movies in the afternoon. It represents a huge missed opportunity: to clarify what our hypotheses actually are about the world in which we live, and test these hypotheses in ways that prevent us from wasting huge sums of money and a corresponding number of human lifetimes on programs that do not matter.

A community without the discursive apparatus to clarify such disagreements, and the ability to invest an appropriate level of work into testing them, is operating on assumptions too low-trust to justify any of the predominant EA hypotheses, all of which require the ability to delegate a lot of work to strangers, including much of the work of evaluating the output of the work you are funding.

Addendum: If you don't already find yourself with a large surplus of wealth or power, and are considering how to make yourself helpful to yourself or others, the model laid out above implies that one thing worth paying a lot of attention to is, as you make your way in life, whether the skills and behaviors you are learning and being rewarded for seem like the sort of thing that is likely to be able to help someone solve a practical, material problem. Sometimes the connection may be real but unclear, but the less reason you have to think that your society is a just one, the more open you should be to the hypothesis that you're being rewarded for bad behavior. If so, you might want to look for another game to play. On global-utilitarian grounds, if you thought that capital accumulation is a gift to the future (or that accumulating "career capital" would improve your ability to help others), you might want to update away from that. On selfish grounds, you should become more skeptical about what money can buy you.


* The image of someone relaxing at the beach in a swimsuit and raincoat is equally ridiculous whether it's raining or not, as is the image of someone similarly attired in a movie theater. I'm pretty sure most readers have found a better solution to a similar problem, than the one in my hypothetical, but I think they would gain a lot from thinking about exactly what their solution would be, and what principles of decisionmaking they are using. I recommend doing that before reading the next paragraph, in which I explain what I'd do and why.

I expect to have more information about tomorrow's weather tomorrow than today. If, in the morning, conditions look good for the beach, I might head there first, bringing my raincoat but not wearing it. If at some point it starts raining, I would abandon my beach plans, put on my emergency raincoat, and head indoors to a movie. If conditions don't look good for the beach, I'd head straight for the movies. In either case, if the movie finishes during the daytime, then I can make another observation of the sky, and use that to decide whether the beach seems promising, or whether I should pursue my best rainy-day option.

I'm not going to give an explicitly mathematized decision-theoretic account, as I think the implied principles I'm using here are pretty obvious. On LessWrong, Lukeprog recommends Peterson's An Introduction to Decision Theory. How to Measure Anything by Douglas Hubbard has more detail about how to use Bayesian methods in practical business applications. The Lean Startup by Eric Ries gives examples, also in a business context, of how we can better achieve our goals by structuring our plans as a series of experiments testing the highest value of information hypothesis, than by committing in advance to a highly conjunctive plan.

9 thoughts on “Oppression and production are competing explanations for wealth inequality.

  1. ety

    > But if wealth does not correspond to productive capacity - i.e. the people with the most are not those best able to use it - then, assuming diminishing marginal returns to wealth, coordination towards persistent wealth inequality comes from a self-sustaining misalignment of incentives, i.e. conflict.

    Surely sometimes one person will have more wealth than another, not because of her greater productive capacity, but due to luck. One person might inherit a fortune, for instance.

    Is the claim that, in the LONG RUN, individuals and organizations should accumulate wealth in proportion to their productive capacity, because if you can reinvest resources better, then you deferentially end up with more resources?

    That seems like an argument for a directional trend or correlation between wealth and productive capacity, but it seems like trend might be swamped by luck. Every generation, some fraction of distribution of wealth is due to efficient allocation of resources, but some other (possibly larger) fraction is due to random factors. Usually when I'm thinking about conventional charity, I'm operating from a frame in which some people have much more resources, not because they can better use them, but because they were luckier. In that frame, charity, or redistribution, is attempting to redress those luck-originating inequalities.

    Isn't this a third hypothesis for what determines the allocation of resources, in addition to allocation by the market to the most productive uses, and allocation according according to privilege in a system of oppression? Am I missing the point somehow?

    > If the mechanism of private property tends to allocate capital to its most productive uses, then incentives are being aligned to put many people to work for common benefit.

    Strictly speaking, that capital is being directed to its most productive uses doesn't mean that those gains are widely allocated. It can be the case that I own an automated factory that can make better use of resources than anyone else, but I will keep the outputs of that factory to myself, or use them to extract or destroy the resources of others.

    1. Benquo Post author

      I agree that luck matters, but luck can't be the motivating case for property rights on its own. If luck inequality gets preserved as wealth inequality, it has to be hitching a ride on some system of motivations for that preservation.

  2. Simon

    Excellent teaching style.

    My solution was to go to the beach and enjoy it, regardless of the rain.

    I think that’s a valid solution, that gets at the heart of an interesting challenge to instrumental rationality that I learned from a friend here.

    In particular: decisions are constitutive of identity (classic insight from Kierkegaard, and later existentialists). I can suspend my identity and plan for both film and beach; but another solution is to throw my lot in with one or the other. I shall be a beach person, and take my lumps, or I shall be a film person, and miss the sky. In my case, I’d rather be a beach person in the rain than a “neither” person at the films.

    This is the either/or vs the both/and. Kierkegaard understands this as accepting the position of freedom, and avoiding the false pretense of being as yet unborn.

    I think it’s also quite close to reality. Stuffing a raincoat in your bag, metaphorically, rarely works out in truly personal cases. You’re neither one nor the other, at sea and uncommitted. It might work in a stock portfolio or startup, but this is a sign of the pathologies of that world.

    When I see young people set out on careers, by contrast, I do see the best ones go all in, choosing and throwing-in as part of the task of building a self.

    It is also a matter of solidarity. Sartre understood this, and his great advance was to see the social implicit in Kierkegaard. “Les Jeux Sont Fait / Rien ne va plus” — you place your bets but, importantly, with others. Sartre’s novel ends in a suicide, as the character rejects the do-over (or, perhaps better, the suicide shows that getting a do over is not worth the candle).

    I appreciated the big ask; that people get together to think together. It’s dead on. I’d add that you can’t get together at the beach if half the people who come have a backup raincoat. Because in the end, it’s always only half the people, at most, who have the escape room to run to. If it can’t fit everyone you might meet, better to blast it. (I don’t really mean to refer to the material here — that’s less important than your time.)

  3. Simon

    Let me play this out a little further.

    Say the EA crowd had gone all in on its weirdest “rainy beach”, MIRI. No backup plan. They start funding it, but commit and don’t stop until they collapse.

    In time, they accumulated hundreds of millions. Instead of squandering it on regifting to standard non-profits, and destroying the community, they now have an endowment approaching that of Carnegie Mellon — but outside of the ordinary funding mechanisms and obligations. Maybe they have a weirdos problem with CFAR, but it’s no worse than what the standard R1 has, and there are problems it definitely doesn’t have.

    Like the Santa Fe Institute, MIRI leadership is smart: they don’t sell out to a famous University, nor do they brand as a think tank. The people involved have gone all-in. This is what they have, and they won’t quit.

    What happens? One answer is that, eventually everyone at MIRI figures out that the malevolent AI is already here, it’s called the market economy. This is, for them, insane and way out of scope, but they’re all in. They can’t be like oop, that’s too weird, we’re just computer geeks, let’s get our raincoats and go to a rerun of The Matrix.

    Instead, they transform themselves. They pivot, and try to rescue the world. Hijinks ensue. Legend outcome. A century later, students at the University of Mars try to cancel statues of the MIRI founders.

    Alternative take: same as above, but MIRI collapses into drugs and cults, because this is not the right crowd.

    The spectacle of the collapse is too salient. The sight of a bunch of people writing esoteric LaTeX fan fiction about Prime Intellect means that (the wrong form of) AGI fear is permanently discredited. Everyone can move on. Still a win for the human race.

    What actually happened: everyone gradually shifted resources over to the next bullshit thing without ever having a reckoning.

  4. Noah

    Wrote this up after starting a second reading of this post:

    check the weather first and, in case of a favorable forecast, go to the beach in or with a swimsuit. bringing a raincoat is cheap, so I might do that also, though really, i often don’t think of some sorts of contingency plans, which is good to an extent due to planning / preparedness costs

    one could also control the environment instead of behavior: invent a weather-controller, or pick between beaches based on weather. or, if it’s rainy, go to a pool to get your swimming fix instead

    wearing a swimsuit and a raincoat actually doesn’t sound so bad. one reason it seems ridiculous to me is i feel like such a person potentially *looks* odd…i’m not sure why, exactly. maybe they just look odd in the same way someone wearing a jacket and shorts looks odd, or maybe specifically i’m thinking of myself or men in general, because of comments about men’s legs looking odd or ugly. somehow stylistically the shorts seem to go better together with a t-shirt, and jackets with … what? pants? I’m not sure what a rain jacket looks good with, i guess it could look OK with jeans?

    besides style, I think the main problem with the coat+suit solution to your hypothetical would occur if you don’t actually want to go to the beach in the sun. in this case, the suit+coat solution just leaves value on the table. if you still would want to go to the beach in the rain, the suit+coat solution seems fine

    a general principle i might use when thinking about this is “delay as many decisions as you can”. i seem to think about this heuristic a lot playing Kittens Game, though i’m not sure i actually use it that much there - an application in this case would be e.g. keeping catnip instead of refining it into wood, because you can always refine catnip into wood, but you can’t directly perform the other transformation. in Kittens Game, there’s also opportunity cost since many other actions than refining catnip are available to you. a purer example might be poker, where people prefer later positions, for more informed decisions

  5. Aaron

    Interesting post, which frames well two different ways of understanding basic economic transactions and some of the consequences of doing so. I think it's interesting to think about the ways one might go about using each of these understandings as heuristics that correct of inform one another (similar to the way that ACX characterizes mistake versus conflict in politics).

    One way of doing this, as I understand it, is the Marxist view according to which both of these things are going on at once. The rich are necessary for capitalist production not because they are necessarily the smartest or most deserving, but because by definition they have the capital. They are able to use that capital to be productive and to allow workers to be productive by, for example, creating factories. In doing so, they are making goods more efficiently and thus lowering their price---allowing more people to buy things for less. That's the production hypothesis. But at the same time, they do so by hiring workers on the labor market, and as production gets more efficient capitalists are also getting more efficient at using less workers and using them more productively. So the demand for workers goes down, so the price of workers goes down, ie., workers' wages goes down---meaning workers are less and less able to afford the things they are making. (Also the price of land goes up as it becomes financialized, so workers are less and less able to afford rent.) So here is where the oppression hypothesis comes into play, as workers must work to afford to live (the threat), but their working buys them less and less.

    There are surely other ways of combining these two understandings fruitfully, but at the very least I think applying the oppression hypothesis to the labor market is fairly helpful and tracks with the way many people think about their jobs.

  6. tailcalled

    Is there any reason you don't consider "people are doing wealth inequality in order to incentivize production" a third valid hypothesis to consider?

    1. Benquo Post author

      That hypothesis overlaps strongly with the capital-allocation hypothesis. We can think of “consumption” in neoclassical ideology as having two components.

      The major one is simply inputs to processes of production that operate on different timescales than the one of the input-output matrix you’re using. If you’re making planning decisions on a one-year timescale, your model doesn’t know that you need to feed your workers every day, or that you need to care for children to eventually replace aging or otherwise depreciating workers, so you model those expenses as though they were pure consumption. (In a classical framework this consideration roughly maps onto Ricardo’s iron law of wages - you don’t keep what you don’t pay for the reproduction of.)

      The second component is conspicuous consumption, a costly signal of military capacity, the sorts of things states or individual warrior-aristocrats want. It’s not inconcievable that one could purchase nonmilitary services from warrior-aristocrats, but there is little precedent, and purchasing military services in this way has a mixed track record.


Leave a Reply

Your email address will not be published. Required fields are marked *