The Margin

Adam Smith set the puzzle and could not solve it. Nothing is more useful than water, he observed, yet it will buy almost nothing; a diamond is of scarcely any use, yet a great quantity of other goods may be had in exchange for it. Usefulness and price seemed to run in opposite directions, mocking the idea that one governs the other. The classicals split the difference with a distinction — value in use against value in exchange — but the distinction named the paradox without dissolving it. The dissolution waited a century, and when it came it was the single most important idea in economics: that what governs the value of a thing is neither its total usefulness nor its average usefulness, but its marginal usefulness — the worth of one more unit, at the margin where a choice is actually made. We are awash in water, so one more glass is worth almost nothing, and its price is almost nothing; diamonds are scarce, so one more is worth a great deal, and commands a great deal. Price follows the margin, and the paradox is gone.

The margin is the increment — the next unit, the little more or less around which every real decision turns. The figure makes the resolution visible. Draw the marginal value of each successive unit, falling as quantity rises. The total value of all the water consumed is the whole area beneath the curve, and it is vast — water is, in total, of immense worth. But the price tracks only the height of the curve at the last unit consumed, the marginal value, and far out along the curve, where an abundant good is consumed, that height has fallen nearly to the axis. A scarce good is consumed only a little way along the curve, where the height — and so the price — is still great. Total value is the area; marginal value is the height of the last unit; and price is governed by the height, never by the area. The whole of the paradox, and the whole of the principle, is in that distinction.

Once value is seen to live at the margin, every question of how much opens to the same key. The question is never "is this thing good or bad?" but "is one more of it worth it?" — and the answer is to go on until the marginal benefit of a little more just equals its marginal cost, and no further. How much water to drink, how much to produce, how much to clean, how much to spend: each is a question not about totals but about increments, answered at the edge. This is why a single idea can carry an entire science. The plates that follow build the whole of demand, of supply, of the firm, of distribution, and of efficiency from this one move — the move from the total to the margin, from the whole to the one more.

For a century the classical economists answered with one voice: value comes from cost, and ultimately from labor. The worth of a thing, Smith and Ricardo held, and Marx after them most insistently, is governed by the labor embodied in its making — the hours of toil congealed within it. Value, on this account, looks backward, to what production required; price is the shadow that the labor of the past casts upon the present. The theory had real power — it explained why goods costly to make tend to be costly to buy — but it could not resolve Smith's own paradox of water and diamonds, and it rested the entire edifice of worth on a single objective substance, labor, poured into things and recoverable from their prices.

Then, in the eighteen-seventies, the edifice was overturned — and overturned three times at once, by three economists who had never read one another. William Stanley Jevons in England, Carl Menger in Austria, and Léon Walras in Switzerland, within a span of a few years, each arrived at the same reversal: value looks not backward to cost but forward to use, and not to total use but to marginal use. A thing is worth what one more of it would add to the wants of the person who must choose — and that worth lives in the mind of the wanter, not in the history of the making. Value became subjective, forward-looking, and marginal, where it had been objective, backward-looking, and total. The source of worth moved from the sweat of production to the satisfaction of the last unit consumed.

The first marginalists, the Austrians above all, pressed the new principle hard enough to seem to banish cost from the theory of value entirely — if worth is marginal utility, what has labor to do with it? Alfred Marshall restored the balance, and in doing so built the apparatus that organizes the subject still. Value, he showed, is determined by both forces together: utility pulling from the side of demand, cost pushing from the side of supply, the two meeting to settle price. To ask whether utility or cost determines value, he wrote, is as reasonable as asking whether it is the upper or the lower blade of a pair of scissors that does the cutting. The marginal revolution did not abolish cost; it dethroned cost as the sole wellspring of value and seated want beside it. The plates that follow develop both blades — the marginal utility behind demand, the marginal cost behind supply — and watch them meet.

Behind the marginal idea stands a regularity about wants so reliable it is called a law: each additional unit of a good yields less additional satisfaction than the one before. To a parched traveler the first glass of water is worth almost anything; the second, merely a great deal; the fifth, little; the tenth, nothing; the twentieth, an active nuisance. This is the law of diminishing marginal utility, glimpsed by Hermann Gossen in eighteen fifty-four — in a book so neglected he died believing it lost — and placed at the center of the subject by the revolution that followed. It is not a logical necessity but an empirical truth about human want, and it holds across so vast a range of goods and circumstances that the exceptions only sharpen the rule. Wants are satiable, and the margin is where satiation shows.

From this single regularity the whole of demand descends. If the marginal utility of a good falls as one consumes more of it, then one's willingness to pay for one more falls as well — and so the demand curve slopes downward. The figure shows the connection laid bare: the declining marginal-utility curve simply is the demand curve, for a buyer will take one more unit so long as its marginal utility is at least its price. Set the price high, and only the first few units — the ones whose marginal utility is high — are worth buying, so little is demanded. Set the price low, and units further along the curve, with their lower marginal utility, become worth having too, so more is demanded. The consumer buys up to the quantity where marginal utility has fallen to meet the price, and no further. Demand is nothing other than the marginal utility of successive units, read off as a schedule of prices.

This grounds the most basic fact in all of economics — that people buy more of a thing when its price falls — not as an assumption but as a consequence. The downward slope of demand follows from the diminishing marginal utility of the thing demanded, and needs no other premise. And it resolves the diamond-water paradox a second time, now from the buyer's side: water is consumed until its marginal utility, and so its price, has fallen nearly to nothing, precisely because we have so much of it; the consumption of diamonds halts far earlier, where marginal utility, and so price, remains high. The law of diminishing marginal utility is the engine beneath the demand curve, and the demand curve is the engine beneath every price, on the side of the one who buys.

A consumer faces not one good but many, a fixed budget to divide among them, and for each good a marginal utility that diminishes as more is bought. How should the budget be allotted to draw the most satisfaction from it? The question is the consumer's whole problem, and the marginal principle answers it with a rule of great elegance, the second of Gossen's laws, known as the equimarginal principle. The answer is not to spend everything on the most useful good — diminishing marginal utility would soon make further units of it worth less than the first units of something else — but to spread the budget by a precise criterion that balances all the goods at once.

Allot the budget, the rule says, so that the marginal utility per dollar is equal across every good. The last dollar spent on bread must yield the same additional satisfaction as the last dollar spent on wine, or on books, or on anything else in the basket. The reason is a simple marginal argument: suppose the last dollar on bread bought more satisfaction than the last dollar on wine. Then you could do better by moving a dollar from wine to bread — gaining more than you lose — and you would keep moving dollars until the advantage was used up. Diminishing marginal utility guarantees that it will be used up, for as you buy more bread its marginal utility falls and as you buy less wine its marginal utility rises, until the two marginal utilities per dollar meet. At the optimum no such move remains: every dollar is doing equally well wherever it is spent, and the allotment cannot be bettered.

The equimarginal principle is the consumer's instance of the universal marginal rule, and its true depth lies in how widely the same form recurs. It says that a scarce resource — here, money — is best allocated when its marginal return is equalized across all its uses. Exactly the same condition governs a firm dividing a budget among inputs (equalize the marginal product per dollar across them), an economy allocating capital among investments (equalize the marginal return across them), a person dividing time among activities (equalize the marginal value of an hour across them). Equalize the marginal yield of the last unit of the scarce thing over every use to which it might be put: it is among the most powerful and most repeated propositions in all of economics, and it is nothing but the marginal principle applied to the problem of allocation.

Cross now to the side of supply, where the marginal principle has a perfect mirror. The relevant quantity for a producer is not the total cost of all it makes, nor the average cost per unit, but the marginal cost — what one more unit adds to total cost. This is the supply-side counterpart of marginal utility, and like marginal utility it is the magnitude that decisions actually turn on: the producer contemplating one more unit asks what that one more will cost, not what the average has cost. And just as marginal utility falls as one consumes more, marginal cost characteristically rises as one produces more — the climb that gives this plate its name.

It rises because of diminishing returns. In the short run a firm has some factors it cannot change — the size of its plant, the extent of its land — and to produce more it must apply ever larger doses of the factors it can change, the labor and the materials, to those it cannot. Crowded onto a fixed plant, each additional worker has less to work with than the last, and so adds less to output than the last: the marginal product of the variable factor diminishes. And diminishing marginal product is, seen from the other side, rising marginal cost — for if each additional worker yields less output, then each additional unit of output requires more workers, and so costs more. The figure shows the consequence: the marginal cost curve climbing as quantity grows. Diminishing returns on the input side and rising marginal cost on the output side are two faces of a single fact.

And the marginal cost curve is the supply curve. A competitive firm, as the next plate establishes, produces where price equals marginal cost; so for any price it might face, the firm supplies the quantity at which marginal cost has climbed to meet that price — and the marginal cost curve, read as the quantity offered at each price, is precisely the firm's supply curve. The upward slope of supply — the homely fact that higher prices call forth more output — is thus, exactly like the downward slope of demand, not an assumption laid down but a consequence drawn out: it reflects rising marginal cost, which reflects diminishing returns. Supply is marginal cost made visible, as demand is marginal utility made visible, and the two curves that meet to set a price are both, at bottom, margins.

A firm seeks the greatest profit — revenue less cost — and the marginal principle hands it the answer in a single line: produce where marginal revenue equals marginal cost. Marginal revenue is what one more unit sold adds to revenue; marginal cost is what that unit adds to cost. So long as marginal revenue exceeds marginal cost, the next unit adds more to revenue than to cost, and making it raises profit; once marginal cost exceeds marginal revenue, the next unit costs more than it earns, and making it lowers profit. The greatest profit is therefore had at the meeting point, where the two margins are equal, neither above nor below. The figure shows it plainly: the rising marginal cost curve crossing the marginal revenue curve, and the optimal quantity standing directly beneath the crossing.

From this one condition the entire theory of the firm unfolds. For a competitive firm — a price-taker, too small to move the market — marginal revenue simply equals the price, since one more unit sells at the going rate; the rule becomes "produce where price equals marginal cost," which is exactly why, as the last plate held, the marginal cost curve is the supply curve. For a firm with market power — a price-setter, the subject of another volume — marginal revenue lies below the price, because selling one more unit means cutting the price on all the units, and so the firm stops sooner, where its lower marginal revenue meets marginal cost, producing less and charging more. The same rule, marginal revenue equal to marginal cost, generates competitive supply and monopoly restriction alike; all that differs between them is the shape of the marginal revenue curve.

One could, in principle, find the same quantity the long way — by computing total profit at every possible output and choosing the highest — but the marginal method is at once easier and more illuminating. It reveals not merely where the optimum lies but why: at any quantity short of it, there is profit left unclaimed in producing more, since marginal revenue still exceeds marginal cost; at any quantity beyond it, there is profit being destroyed, since marginal cost now exceeds marginal revenue. The optimum is the unique quantity at which the marginal incentive to expand has exactly vanished — where the pull toward one more and the push against it precisely cancel. This is the marginal principle in its purest decision-making form, and the same logic, with benefit standing for revenue and cost for cost, locates every optimum in the book.

The figure above is fixed at one setting; the contest it draws is not. Below, the same meeting is made operable. Take the quantity dial and hunt for the optimum by hand: total profit climbs while marginal revenue still beats marginal cost, reaches its summit exactly at the crossing, and turns down the moment one unit too many is made — the surplus already built now eaten back in red. Raise the cost of production and watch the whole optimum slide left. The rule is not asserted here but felt: stop at the meeting, and nowhere else.

Quantity produced

Cost of production

Revenue, next unit—

Cost, next unit—

Profit on it—

Total profit—

The marginal principle governs not only how much a firm produces but how much of each input it employs. The marginal product of a factor is the extra output that one more unit of it yields — one more worker, one more acre, one more machine — and the value of that marginal product, the marginal product multiplied by the price of the output, is what one more unit of the factor is worth to the firm. The familiar logic then fixes how much to hire: a firm takes on a factor up to the point where the value of its marginal product equals its price. It hires labor until the value of the marginal product of labor equals the wage; land until the value of the marginal product of land equals the rent; capital until the value of the marginal product of capital equals the rate. The demand for a factor is its value-of-marginal-product curve, just as the demand for a good is its marginal-utility curve.

From this John Bates Clark drew a sweeping consequence. In competitive markets, each factor tends to be paid the value of its marginal product: labor earns the value of the marginal product of labor, capital the value of the marginal product of capital, land the value of the marginal product of land. Distribution — the vast question of who gets what, to which an entire group of this atlas is given — becomes, on this view, simply the marginal principle applied to the factors of production. The wage, the rent, the rate are not arbitrary, nor set by custom or power alone, but tend toward the marginal contributions of the things they pay for. The same logic that prices a loaf of bread prices an hour of labor and an acre of land; the theory of factor prices is the theory of goods prices, turned upon the inputs.

Here the analytical power carries a provocation that has echoed through economics ever since. If each factor is paid the value of its marginal product, is each factor's reward therefore deserved? Clark himself read his theory as a kind of natural justice — to each according to what he creates. But this is a leap from is to ought that the theory does not license. Marginal productivity explains what a factor is paid in a competitive market; it says nothing whatever about whether the distribution of the factors themselves — who owns the capital, who inherited the land, who was given the chance to acquire the skill — is just, and nothing about what happens when markets are not competitive. The explanation of factor prices is one thing; the justification of the incomes they produce is quite another, and the volume on the share is built around keeping the two apart.

The marginal principle, applied without flinching, yields a conclusion at once startling and liberating: a cost already incurred and impossible to recover — a sunk cost — should have no bearing whatever on any decision. Only marginal costs and benefits, which lie in the future and depend on the choice now to be made, can rightly enter into it. What is past is past; the sole question is what one more step gains and costs from here. Whatever has already been spent is spent no matter what one now chooses, and so it cannot tip the balance between the choices that remain. "In commerce," Jevons wrote, "bygones are forever bygones"; the maxim is the marginal principle pointed at the past.

Yet people violate it constantly, and the violation has a name — the sunk-cost fallacy, the impulse to throw good money after bad, to persist in a failing course because of what has already been poured into it. The half-finished project that ought to be abandoned but is carried on for the sake of the years already sunk in it; the costly ticket to an event one no longer wishes to attend, used only because it was paid for; the war prolonged because so many lives have already been lost, as though further loss could honor the loss already suffered. In every case a sunk cost — irrecoverable, in the past, identical across all the options now open — is wrongly permitted to govern a forward-looking choice that should turn only on the marginal benefit and marginal cost of going on from this moment.

To think at the margin is, in part, to acquire the discipline of ignoring sunk costs — to ask always "what is the best thing to do from here?" rather than "how can I justify what I have already done?" This is genuinely hard, for the past exerts a psychological pull that the marginal calculus coldly overrides: the aversion to loss, the wish not to have wasted, the sense that to abandon a thing is to confess defeat. The principle does not say that the past is without lessons — one should learn from what went wrong, and carry the lesson forward. It says that the past is irrelevant to this choice, except insofar as it has changed the marginal options now available. The accountant's sunk cost and the gambler's stubbornness are the same error in different dress, and the cure for both is to look only forward, weighing one more against one more.

Marginal and average are forever confused, and the confusion is costly, for decisions hinge on the marginal while the eye is drawn to the average. The average cost is total cost divided by quantity — the cost per unit taken over all units made. The marginal cost is the cost of one more. They are different numbers; they move differently; and it is the marginal, never the average, that answers the decision-maker's question of whether to make one more. To carry both quantities clearly, and never to mistake one for the other, is among the first marks of disciplined economic thought.

Between the two there is a precise and illuminating relation: the marginal pulls the average toward itself. When the marginal lies below the average, the average must fall — the next unit, cheaper than the running mean, drags the mean down. When the marginal lies above the average, the average must rise — the next unit, dearer than the mean, pulls it up. And therefore the average is at its lowest exactly where the marginal crosses it from below. The figure shows this signature: the U-shaped average cost curve, with the marginal cost curve passing precisely through its lowest point. It is a perfectly general truth of marginals and averages, and it answers the plate's question at once — if the average is rising, the next one must lie above it. The marginal student's grade pulls the class average; the marginal hire's output pulls the team's; and the marginal cost curve always threads the very bottom of the average cost curve.

This is why the decision must ride on the margin. A firm weighing whether to produce one more unit compares the marginal revenue with the marginal cost — never with the average cost, which concerns units already made and is beside the point for the increment. A unit may cost more than the average and still be worth making, if it earns more than its own marginal cost; a unit may cost less than the average and not be worth making, if it earns less than its marginal cost. The average describes the past — what has been produced and at what cost per unit; the margin governs the future — whether to produce more. To decide at the average rather than the margin is among the most common and most expensive errors in all of practical economics, the error of letting the cost of what is done dictate the doing of what is next.

The marginal principle answers not only the private question — how much should a firm produce for profit? — but the social one: how much should be produced for the good of all? The efficient quantity is the one at which marginal benefit equals marginal cost, where the value to society of one more unit just equals the cost to society of making it. Below this quantity, marginal benefit exceeds marginal cost: another unit would be worth more than it costs, so too little is being made and the shortfall is a loss. Above it, marginal cost exceeds marginal benefit: the last units cost more than they are worth, so too much is being made and the excess is a waste. The figure sets the optimum at the crossing of the marginal benefit and marginal cost curves — the single quantity that is neither too little nor too much.

This unlocks an entire class of questions that have no answer at the extremes and a clean answer at the margin. How much pollution should a society allow? How much safety should it buy? How much of a public good should it provide, how much should it spend on a hospital or a road? Zero pollution and infinite safety are both wrong; the right amount of each lies where the marginal benefit of a little more meets its marginal cost. The right amount of pollution is not zero but the amount at which the marginal cost of further cleanup equals the marginal benefit of cleaner air; the right amount of caution is where one more increment of protection costs exactly what it saves. The marginal principle converts impossible questions of how much — which invite shouting matches between none and all — into a single, answerable condition at the edge.

And the efficient quantity has a second meaning that binds this volume to the next. The quantity where marginal benefit meets marginal cost is precisely the quantity that maximizes total net benefit — the gap between total benefit and total cost, which is the area between the two curves up to that point. It is the same quantity at which the sum of consumer and producer surplus, the subject of the very next volume, stands at its greatest. And in a competitive market, where buyers press to where marginal benefit equals price and sellers to where marginal cost equals price, the market is led to this efficient quantity of its own accord — the deepest sense in which a competitive market can be said to work. Where it fails to reach the efficient quantity — through the power, the externalities, the missing information of the third group of this atlas — the gap between the actual and the efficient quantity is the exact measure of the failure.

The early marginalists spoke of utility as though it could be measured — so many units of satisfaction, diminishing with each additional good, summed and compared like weights upon a scale. But can it? Is the satisfaction of a glass of water a cardinal quantity, like its temperature, that might in principle be read off in some unit and set beside the satisfaction of another person's glass? The claim grew steadily more embarrassing, for no one could exhibit the unit or the instrument, and the entire edifice of marginal value seemed to rest upon a quantity no one could find. This was the deep vulnerability of marginalism in its first form — a theory of measurement with nothing measured.

Vilfredo Pareto, and after him John Hicks with R.G.D. Allen, showed that the edifice could be saved, and saved by demanding far less. Marginalism, they proved, needs only ordinal utility — the power to rank bundles, to say that this is preferred to that — and not cardinal utility, the measurement of how much satisfaction each yields. The instrument of the rebuilding is the indifference curve: the set of all bundles of goods among which a consumer is indifferent, all standing equally in their estimation. A consumer need only be able to say which bundles lie on higher indifference curves, the more preferred, and which on lower; no numbers, no units, no measurement of feeling is required. From ranking alone — from the bare fact that one prefers, without any claim about how much — the whole of consumer theory can be built again.

And the marginal principle reappears, wholly undiminished, in its new ordinal dress. The slope of an indifference curve is the marginal rate of substitution — the rate at which the consumer will trade one good for another with no change in well-being, the marginal trade-off between them. Diminishing marginal utility becomes the diminishing marginal rate of substitution: the indifference curves bow toward the origin, because the more of a good one already has, the less of another one will surrender to get one more of it. And the consumer's optimum, once described as the allotment that equalizes marginal utility per dollar, is now the tangency of the budget line with the highest reachable indifference curve — the point, shown in the figure, where the marginal rate of substitution equals the price ratio. The margin survives the loss of measurable utility entirely intact, for it was never truly about the cardinal numbers; it was always about the rates of trade-off at the very edge of choice.

The margin is the one more, and the marginal principle is the discovery that value, choice, and optimum all live there — never in the total, never in the average, but in the increment. Across eleven plates the single idea has been carried through its instances: the increment and the revolution that placed it at the center; the diminishing of marginal utility and the allotment it dictates; the climb of marginal cost and the meeting where it stops the firm; the product that prices the factors; the bygone that the margin discards; the average that the margin must not be confused with; the efficient quantity the margin locates; and the substitution into which the margin survives the loss of measured utility. It remains only to gather them into the one rule they all express.

That rule has a single universal form: do anything up to the point where the marginal benefit of one more just equals its marginal cost, and no further. Its instances are the whole of the volume. The consumer allots a budget until the marginal utility per dollar is equal across all goods. The firm produces until marginal revenue equals marginal cost. A factor is hired until the value of its marginal product equals its price. A society makes a thing until its marginal benefit equals its marginal cost. Four problems, four optima, and one condition beneath them all — for every economic optimum, without exception, is a marginal condition, a balancing of one-more-benefit against one-more-cost.

The principle joins the two sides of every market. Demand is marginal utility made visible, for diminishing marginal utility is the downward slope of the demand curve; supply is marginal cost made visible, for rising marginal cost is the upward slope of the supply curve; and the price emerges where the two margins meet — the two blades of Marshall's scissors, the marginal value to the buyer and the marginal cost to the seller, closing upon a single point. The whole of the competitive market, reduced to its essence, is the meeting of a marginal benefit and a marginal cost.

The same idea reaches into the division of the product among those who make it. Each factor — labor, land, capital — tends in competition to be paid the value of its marginal product, so that the great questions of who gets what are, in their competitive idealization, marginal questions too. But here the principle marks its own limit: the step from is paid its margin to deserves its margin is a leap from fact to justice that the principle does not make and cannot make, and the confusion of the two is the oldest error the marginal theory of distribution invites.

To think at the margin is a discipline as much as a doctrine. It is to ignore the sunk cost and the average that mislead; to weigh the one-more against the one-more; to ask always, of any course already begun or about to begin, not "how do I justify what I have done?" but "what is the best thing to do from here?" More than any other single habit of mind, this is what it means to think like an economist — to see, behind every question of whether and how much, a comparison of increments at the edge.

The principle has proved deep enough to survive the loss of its own first foundation. When the measurable utility on which it was built turned out to be a fiction, the margin did not fall with it but reappeared, undiminished, as the marginal rate of substitution, needing only the ranking of bundles and not the measuring of feeling. Yet it is not a universal solvent, and its strains are as instructive as its reach. It strains where the world is lumpy — indivisibilities, all-or-nothing choices, the half a bridge one cannot buy; where returns increase rather than diminish — economies of scale, network effects, the falling marginal cost of the second age; where choosers do not optimize — the bounded, biased, habitual agents of the behaviors group; and where it is mistaken for a verdict on desert — the is that is read as an ought. To wield the marginal principle well is to know exactly where it bites and where it slips.

The Margin is the first of The Quantities, and the keystone beneath the rest of the group. It underlies The Surplus, which measures the gains from trade as the very areas between the marginal curves this volume has drawn. It underlies The Utility, whose theory of preference the margin presupposes and whose cardinal-and-ordinal question the substitution plate has already broached. It underlies The Stock, where the marginal logic extends across time to capital, inventory, and the assets held against the future. And it reaches far beyond its group — to The Wage, The Rent, and The Rate, where the factors meet their marginal products; to The Externality and The Power, where the gap between the private and the social margin is the precise measure of failure; to The Firm, built on the meeting of marginal revenue and marginal cost. The margin is less a single element than the grammar in which much of the rest of the atlas is written.

What, in the end, is a thing worth? Not the whole of what it does for us, nor the average, but the one more — the marginal unit, at the edge where we actually choose. And this small idea, once seen, reorganizes everything: value and demand, cost and supply, the firm and the factor, the efficient quantity and the optimum of every kind. The marginal revolution was rightly named a revolution, for it was the quiet replacement of a subject built on cost and totals by one built on want and increments, and the discovery that almost every economic question, rightly posed, is a question about one more. The quantity, like every quantity this atlas weighs, is a verdict on a hidden contest: between the total and the margin, between cost and want as the wellspring of value, between the whole that impresses the eye and the one more that moves the hand. The argument remains open.