Independent revision built for the Bristol first- and second-year Economics & Finance syllabus. Not affiliated with or endorsed by the University of Bristol.

Efficiency in Exchange: the Edgeworth Box and the Contract Curve

Why one market is never the whole story, and what it means for an allocation to be efficient when nobody can be compared to anybody. The pure endowment economy, the two conditions for Pareto efficiency, and the geometry that makes both visible at once.

Partial Equilibrium, General Equilibrium

Core concept

This block steps back from one market and looks at the whole economy at once. Everything you did last year was partial equilibrium: draw supply and demand for one good, hold everything else fixed, read off the answer. That is a useful lie. Taxing cider changes what people spend on pasties; the pasty price then feeds back into the cider market. General equilibrium asks whether there is a set of prices that clears every market simultaneously — and over the next four weeks uses that machinery to say something genuinely surprising about efficiency, about what markets achieve, and about what redistribution costs.

This week lays the groundwork for all of it — what efficiency even means in an economy where nothing is produced and no two people can be compared.

Partial equilibrium analysis finds the price and quantity in one market while treating every other price as a fixed constant. General equilibrium analysis lets all prices move together and insists that all markets clear at once.

The distinction matters whenever markets are linked — and they are linked whenever goods are substitutes or complements in demand, or compete for the same inputs in supply. A tax on good X does not simply vanish: the money not spent on X reappears somewhere, and the price that reappears there bounces back into the market for X.

How many markets? In principle everything is connected to everything. In practice we choose the markets that carry most of the feedback and ignore the rest. Taxing rail fares, you would look at coach travel and fuel — not at garden furniture. Almost all of this unit works with two or three goods, which is enough to make every general-equilibrium point visible and few enough to draw.

What the theory does and does not claim

General equilibrium theory asks: do prices exist that clear all markets? It does not ask how a market finds them. That second question needs a theory of price adjustment out of equilibrium, and economics does not have a settled one — it would require knowing not just supply and demand but how each reacts, in real time, to every other variable moving at once.

Short run vs long run

We work in the short run, where production capacity is roughly fixed. Long-run general-equilibrium effects — capital accumulating, technology shifting — belong to growth and development models.

Where you already see it

A tariff raises import prices at home, but can lower prices in unrelated third countries as supply is diverted there. Co-movement between equity prices and commodity prices is largely a general-equilibrium phenomenon.

How the four weeks are ordered

Weeks 1–3 strip production out entirely: just endowments, and trade. Week 1 is efficiency, Week 2 equilibrium, Week 3 welfare. Week 4 puts production back and runs the same three ideas again.

A Tax, Worked Through Both Markets

The calculation

The previous section claimed that a tax on one good does not stay in that market. Here is what that looks like when you actually do the arithmetic, in the smallest economy that can show it: two goods that are substitutes, two markets, one tax.

Cider and pasties. Quantities in thousands, prices in pounds:

cider    DC = 12 − 3PC + 2PP     SC = 2 + PC
pasties   DP = 5 − PP + 3PC       SP = 8 + 2PP

Each demand curve slopes down in its own price and up in the other — that is what makes them substitutes, and it is the only reason the two markets have anything to say to each other. Set demand equal to supply in both and the economy rests at PC = £4 and PP = £3, with 6 thousand litres and 14 thousand pasties sold.

Now impose a tax of £2 a litre on cider

Specifically, a tax collected from buyers: the price a drinker pays is PC + t, while the brewer still receives PC. The whole problem turns on where that PC + t has to appear.

Obviously it appears in cider demand. The step that is easy to miss is that it appears in pasty demand too. What a pasty competes against is not the price a brewer receives; it is the price a drinker actually hands over. So both demand curves change, and only one of them is the taxed good:

DC = 12 − 3(PC + t) + 2PP     SC = 2 + PC
DP = 5 − PP + 3(PC + t)    SP = 8 + 2PP
That single substitution is the whole of general equilibrium here. If you write the tax into cider demand and leave pasty demand untouched, you have done partial equilibrium with extra steps — and the next section shows exactly how far wrong it goes. Every cross-price term in the economy takes the price the buyer faces, tax included.

Clearing both markets at once

Equate demand and supply in each, and tidy:

12 − 3(PC + t) + 2PP = 2 + PC   →   10 − 3t + 2PP = 4PC
5 − PP + 3(PC + t) = 8 + 2PP   →   3PC + 3t − 3 = 3PP

The second line gives PP = PC + t − 1. Substituting into the first, 10 − 3t + 2PC + 2t − 2 = 4PC, so 8 − t = 2PC and:

PC = 4 − t/2      PP = 3 + t/2      buyers pay PC + t = 4 + t/2

Read those three. The brewer's price falls by half the tax, the drinker's price rises by half the tax — an even split, which is what you get when the slopes happen to be symmetric like this and not a general law. And the pasty price rises, by the same half. That last one is the general-equilibrium effect, and it has a sign worth remembering: tax a good and its substitutes get dearer, because demand moves across to them.

At t = 2 everything comes out whole:

no taxtax of £2change
price brewers receive£4£3−£1
price drinkers pay£4£5+£1
pasty price£3£4+£1
cider sold65−1
pasties sold1416+2
tax revenue—£10

Revenue is t × QC = t(6 − t/2), which is a parabola: it rises, peaks at t = 6 with revenue of 18, and falls back to nothing at t = 12 when the last litre has gone. Nothing mysterious about that shape — the tax collects more per litre but there are fewer litres, and past a point the second effect wins.

Taxing sellers instead

Write SC = 2 + (PC − t) and leave demand alone. The algebra differs, the answer does not: the same £5 and £3, the same quantity, the same revenue. Who hands the money over does not decide who bears it.

Why the split is even here

Because cider supply and demand happen to be equally responsive at the margin once the feedback is counted. Write supply as SC = 2 + ePC and brewers bear 1/(e+1) of each pound: at e = 1 that is half, at e = ½ two-thirds, at e = 4 a fifth. The steeper, less responsive side bears more — whoever cannot move away from the tax pays it.

What Partial Equilibrium Gets Wrong

The comparison

Now do what last year's course would have done: look at the cider market, hold the pasty price at its old value of £3, and solve one equation.

12 − 3(PC + t) + 2(3) = 2 + PC   →   PC = 4 − 3t/4

At t = 2 that gives PC = £2.50, drinkers paying £4.50, and 4.5 thousand litres sold — against the true £3, £5 and 5. Revenue comes out at 2 × 4.5 = £9 thousand where the truth is £10. The partial answer is wrong in every figure, and wrong in the same direction each time.

The error has a formula

Compare the two expressions for the brewer's price: 4 − t/2 with the pasty market free, 4 − 3t/4 with it frozen. Subtract:

error in PC = t/4      error in QC = t/4      error in revenue = t²/4

Partial equilibrium understates all three, always, in this economy. The reason is visible on the diagram: freezing the pasty price throws away the one force holding cider demand up. Pasties really do get dearer, drinkers really do come back to cider, and a model told to ignore that sees only the fall.

The revenue error is the one to carry. It grows with the square of the tax, so the approximation is at its worst exactly where a government would care most — at t = 2 it is £1 on £10, a tenth; at t = 4 it is £4 on £16, a quarter.

When partial equilibrium is good enough. Not never — the error here is driven by the cross-price terms, and if cider and pasties barely substituted, the 2 and the 3 in those demand curves would be near zero and both answers would nearly coincide. The practical rule is the one the previous section gave: include the markets that carry most of the feedback. What this example adds is that you can now say how much the feedback was worth, rather than asserting that it matters.
A warning about signs, and a trap inside it. It is tempting to say that this all depends on the goods being substitutes, and that complements would reverse it. Only half of that is true. The feedback runs round a loop — cider price moves pasty demand, which moves the pasty price, which moves cider demand — so what governs the error is the product of the two cross-price terms, here 2 × 3. Make both negative, as complements are, and the product is positive again: partial equilibrium still understates, by exactly the same amount.

What complements genuinely reverse is the substitute’s price. Tax cider and a substitute gets dearer; tax cider and a complement gets cheaper. The error only changes sign in the mixed case, where one good substitutes and the other complements, and the product turns negative. So the honest rule is: you cannot sign the error from one cross-price term, only from both.

The Economy We Will Use All Term

Setup

Strip the economy to its bones. Two people, two goods, no production at all. Each person simply starts holding some of each good — their endowment — and the only economic activity available to them is trade. No-one makes anything; they just swap.

That sounds impossibly thin, and it is deliberately so. Every result in the next three weeks — the welfare theorems, the case for markets, the cost of redistribution — can be stated and proved in this economy. Adding production later changes the machinery but not the conclusions.

Our two traders are Morwenna, who works a boat out of Mousehole and lands pilchards, and Tegan, who keeps an orchard above the village. Good 1 is pilchards, good 2 is apples. Write ωkj for person j's endowment of good k, and xkj for what they end up consuming.

UM(x₁M, x₂M) = ln x₁M + α ln x₂M    UT(x₁T, x₂T) = β ln x₁T + ln x₂T

with α > 0 and β > 0. Nothing hangs on the logarithms. They are chosen because log utility makes each person spend a fixed share of their income on each good, which keeps the algebra short enough to do in an exam.

The numbers we will carry through

For every worked example in Weeks 1 to 3, take α = 1 and β = ½, and these endowments:

Morwenna — the boat

Endowment (12 pilchards, 16 apples). Utility ln x₁ + ln x₂: she weights the two goods equally.

Tegan — the orchard

Endowment (6 pilchards, 36 apples). Utility ½ ln x₁ + ln x₂: apples matter twice as much to her at the margin.

The economy

Totals 18 pilchards and 52 apples. Nothing else exists and nothing can be made. Every allocation we discuss is a way of dividing exactly these.

Notice the setup is lopsided on purpose. Morwenna is pilchard-rich relative to the economy and Tegan is apple-rich, and their tastes run the other way to their holdings. That is what will make trade worth doing.

An allocation is four numbers, not two. Writing it as (x₁M, x₁T, x₂M, x₂T) looks pedantic until you notice that the whole of this week is about which of those four-number lists are good ones. A great deal of the difficulty students have here comes from quietly collapsing the allocation into a single person's bundle and losing track of what the other one is holding.

Feasibility, and Why Waste Is Never Efficient

Condition one

An allocation is feasible if it does not hand out more of a good than the economy actually has:

xkM + xkT ≤ ωkM + ωkT    for each good k = 1, 2

Written with our numbers: x₁M + x₁T ≤ 18 and x₂M + x₂T ≤ 52. The inequality allows for throwing things away.

Now ask whether an allocation that wastes anything could ever be efficient. Suppose only 17 of the 18 pilchards are eaten. That spare pilchard can be handed to either person. Because both goods are normal — more is better — whoever receives it is strictly better off, and the other person is untouched. We have made someone better off at nobody's expense.

So the first condition for efficiency is simply that the feasibility constraint binds:

x₁M + x₁T = 18     x₂M + x₂T = 52

This is worth more than it first appears. It means we can stop treating the allocation as four free numbers. Once Morwenna's bundle (x₁M, x₂M) is named, Tegan's is forced: she gets (18 − x₁M, 52 − x₂M). Four numbers collapse to two, which is exactly why the whole economy can be drawn on a flat page.

Check the size of the loss. With Morwenna at (10, 20), throwing away a single apple costs her about 0.051 utils. That number is meaningless on its own — log utility has no units — but its sign is the entire argument. Efficiency rules out waste not because waste is expensive but because it is strictly, unambiguously improvable, with no trade-off to weigh and nobody to compensate.

Equalising the Marginal Rate of Substitution

Condition two

Now suppose nothing is wasted, so both constraints bind, and ask whether that is enough. It is not.

The marginal rate of substitution is the rate at which someone will swap one good for the other and feel no better or worse. For Morwenna it is the ratio of her marginal utilities:

MRSM = MU₁M / MU₂M = x₂M / (α x₁M)     MRST = β x₂T / x₁T

Read MRSM as apples per pilchard: how many apples Morwenna would hand over for one more pilchard and be left indifferent.

At the endowment, with α = 1 and β = ½, those come out as

MRSM = 16/12 = 1.33     MRST = ½ × 36/6 = 3.00

These are not equal, and the gap is an opportunity. Morwenna will give up only 1.33 apples for a pilchard. Tegan would pay 3 apples for one. Any exchange rate strictly between those two numbers — say 2 apples per pilchard — leaves both of them better off than before. Morwenna sells a pilchard for more apples than it was worth to her; Tegan buys it for fewer apples than she was willing to pay.

That trade can be repeated while the gap persists. As Morwenna gives up pilchards and gains apples her MRS rises; as Tegan does the reverse hers falls. The gains are exhausted exactly when the two rates meet:

MRSM = MRST
The direction of the argument matters. We did not assume a market, a price, or even money. All that was used is that the two people value the goods differently at the margin, and that both prefer more to less. The conclusion — unequal MRS means an unexploited gain — holds in a household, a prison camp or a competitive economy. What markets add, in Week 2, is a mechanism for finding the trade. They are not what makes the trade worth doing.

Pareto Efficiency, and What It Deliberately Ignores

The standard

The two conditions together define the standard this whole unit runs on.

An allocation is Pareto efficient if no-one's utility can be raised without lowering someone else's.

Equivalently, an allocation can be Pareto improved if there is another feasible allocation that makes at least one person better off and nobody worse off. Pareto efficient means: no such improvement exists.

Three things it does not do

It never compares people

Utility here is ordinal. "Morwenna gains 0.04 and Tegan gains 0.03" is not a statement the criterion can use — their utility numbers are not in the same units and cannot be added, subtracted or ranked against each other.

It says nothing about fairness

The allocation where Morwenna takes all 18 pilchards and all 52 apples and Tegan gets nothing is Pareto efficient. You cannot improve Tegan without taking something from Morwenna. Efficiency and justice are different questions.

It is nearly always silent

Most real policy makes someone worse off. The criterion simply declines to rank those options. It is a very weak test — which is precisely why it is useful.

That weakness is the point. Pareto efficiency is close to the only normative standard economists can agree on without importing a theory of justice. Because it asks so little, anything it does rule out is uncontroversially bad: if an allocation can be Pareto improved, there is a change available that literally no-one objects to.

Efficient is not the same as good. This is the single most common misreading of the next three weeks. When Week 2 proves that competitive markets deliver a Pareto efficient outcome, it will not have proved that markets deliver a desirable one. It will have proved that they leave no mutually agreeable trade on the table — which is a real and surprising achievement, and a much narrower claim than it sounds. Week 3 takes up what is missing.

The Edgeworth Box

Key diagram

The Edgeworth box is the single most useful picture in this unit, and it is built on a trick. Draw a rectangle whose width is the total stock of pilchards — 18 — and whose height is the total stock of apples — 52. Measure Morwenna's consumption from the bottom-left corner in the usual way. Then measure Tegan's consumption from the top-right corner, with both axes reversed.

Every single point in the box is now a complete feasible allocation that uses up exactly all of both goods — because whatever Morwenna does not consume, Tegan does. Two people's consumption bundles collapse into one point, which is the whole reason the diagram works.

This is Section 3's first condition made geometric. We showed that efficiency forces both feasibility constraints to bind, which let us write Tegan's bundle as (18 − x₁M, 52 − x₂M). The box is that substitution, drawn. Points outside the rectangle are infeasible; points inside it with something thrown away do not exist on this diagram at all, because the construction has already assumed the constraints bind.

Morwenna's indifference curves have their usual convex shape, rising away from her origin at the bottom left. Tegan's are drawn from the opposite corner, so they look convex "upside down" from where you are standing — but they are ordinary indifference curves seen from a rotated viewpoint, and hers improve towards the top right.

Tangency is efficiency. Where two indifference curves are tangent, their slopes are equal — and the slope of an indifference curve is the MRS. So tangency is exactly condition two. Since every point in the box already satisfies condition one by construction, a tangency point satisfies both, and is therefore Pareto efficient. Where the curves cross rather than touch, they carve out a region between them, and anything in that region beats both — so a crossing point is never efficient.

The Lens: Where Both of Them Can Gain

Gains from trade

So far the box has said nothing about who owns what. Put that back in. Mark the endowment point E at (12, 16) measured from Morwenna's corner — which automatically puts Tegan at (6, 36) measured from hers.

Now draw the indifference curve through E for each person. They cross at E, and between them they carve out a lens-shaped region.

Every allocation strictly inside that lens makes both people better off than the endowment does. These are the mutually beneficial trades — the ones that could actually happen between two people who are free to walk away.

The lens exists because the two indifference curves cross at E rather than touch. And they cross because, as Section 4 computed, MRSM = 1.33 while MRST = 3.00 at the endowment. Had those been equal the curves would have been tangent at E, the lens would have collapsed to a point, and there would be nothing to trade.

The lens shrinks as they trade. Suppose they agree a swap and land somewhere strictly inside the lens. Draw the two indifference curves through that new point and you get a new lens, smaller than the old one: the gains they have already banked are no longer on offer. Trade can carry on while a lens remains, and has to stop once the lens has closed to a point — which, by Section 6, is exactly where the two curves are tangent instead of crossing. So the question "where does trade end?" is the question "where are the tangencies?", and that is the next section.

The Contract Curve, Derived

The efficient set

The contract curve is the locus of every tangency point in the box — the whole Pareto efficient set. Here it is, derived rather than asserted.

Start from condition two and substitute condition one, which lets us write everything in terms of Morwenna's bundle alone:

MRSM = MRST   ⟹   x₂M / (α x₁M) = β (52 − x₂M) / (18 − x₁M)

Cross-multiply, collect the x₂M terms on one side, and solve. In general:

x₂M = αβ (ω₂M + ω₂T) x₁M / [ ω₁M + ω₁T + (αβ − 1) x₁M ]

With α = 1, β = ½, totals 18 and 52, that simplifies to a curve you can sketch from memory:

x₂M = 52 x₁M / (36 − x₁M)    for x₁M between 0 and 18

Check the ends. At x₁M = 0 it gives x₂M = 0: Morwenna has nothing, Tegan has everything, and that corner is efficient. At x₁M = 18 it gives x₂M = 52: the mirror image. The curve runs corner to corner, as it must — both corners are allocations nobody can improve on without taking from the person holding everything.

x₁Mx₂MTegan getsCommon MRS
34.73(15, 47.27)1.58
610.40(12, 41.60)1.73
917.33(9, 34.67)1.93
1020.00(8, 32.00)2.00
1226.00(6, 26.00)2.17
1537.14(3, 14.86)2.48

Every row satisfies MRSM = MRST exactly. The common MRS rises as Morwenna moves along the curve: the more pilchards she holds, the more apples a further pilchard is worth to the economy at the margin.

Two sets, easily confused

Inside the lens

Allocations that Pareto improve on E. Both people accept them voluntarily. Most of them are not efficient — there are still further gains available.

On the contract curve

Allocations that are Pareto efficient. Most of them are not improvements on E — many would leave one of the two worse off than simply refusing to trade.

In both

The segment of the contract curve lying inside the lens. Efficient and acceptable to both without coercion. This is where voluntary trade between two people ends up.

Why "mutually beneficial" is not the same as "efficient". Students routinely treat the lens and the contract curve as the same object. They are not even the same shape: the lens is a two-dimensional region, the contract curve is a one-dimensional line, and they overlap only in a short segment. An allocation can be a genuine improvement on where you started and still leave money on the table.
The curve is a set of answers, not an answer. Every point on it is Pareto efficient. The criterion cannot separate them — ranking the corner where Tegan has everything against the corner where Morwenna does would require comparing their utilities, which Section 5 ruled out. Efficiency narrows an infinite set of feasible allocations down to a line, and then stops.

Where on the Curve? The Box Cannot Say

Looking ahead

We have gone as far as efficiency alone can take us. Of all the ways to divide 18 pilchards and 52 apples, the contract curve picks out the ones that waste nothing in either sense. Which single point on it actually happens is a different question, and it is not a question about efficiency at all. It is a question about institutions.

In a household

The parents decide, or the two children negotiate. The outcome depends on bargaining, custom and authority — none of which appears anywhere in the model.

In a jungle

Strength decides. Whoever can take the pilchards takes them. Still potentially efficient; nobody would call it just.

In an economy

Property rights plus markets decide. Morwenna owns her catch, Tegan owns her crop, and they trade at a common price. That is Week 2.

The move to markets is doing two separate jobs, and it is worth keeping them apart. First, it selects one point: given the endowment and a price ratio, there is a determinate outcome. Second — and this is the surprising part — the point it selects turns out to lie on the contract curve, every time, without anyone intending it. That second claim is the First Welfare Theorem, and it is the reason this whole apparatus was built.

Ownership has been silently doing work. Up to the lens we never used the endowment at all; we just asked which allocations were good. The moment we drew E and asked which trades both parties would accept, we assumed property rights — that Morwenna's 12 pilchards are hers to withhold. Without that assumption the lens means nothing, because nobody has the standing to refuse. Markets are a mechanism for allocating resources given property rights, and they take those rights as an input rather than justifying them.

A Worked Allocation, Start to Finish

Putting it together

Take the allocation where Morwenna consumes (10 pilchards, 20 apples) and run every test from this week against it.

TestWorkingVerdict
Feasible?Tegan gets (18 − 10, 52 − 20) = (8, 32). Both non-negative.Yes
Nothing wasted?10 + 8 = 18 pilchards; 20 + 32 = 52 apples. Both bind.Yes
MRS equal?MRSM = 20/10 = 2. MRST = ½ × 32/8 = 2.Yes
On the contract curve?52 × 10 / (36 − 10) = 520/26 = 20. ✓Yes
Pareto efficient?Both conditions hold.Yes
Better than the endowment for Morwenna?ln 10 + ln 20 = 5.298 against ln 12 + ln 16 = 5.258.Yes
Better than the endowment for Tegan?½ln 8 + ln 32 = 4.506 against ½ln 6 + ln 36 = 4.479.Yes
Inside the lens?Both strictly better off than at E.Yes

So this allocation is both efficient and a genuine improvement on where the two of them started — it sits in that short segment where the contract curve passes through the lens. Morwenna has given up 2 pilchards and received 4 apples; the implied exchange rate is 2 apples per pilchard, which sits neatly between her valuation of 1.33 and Tegan's of 3.00, exactly as Section 4 predicted it would have to.

That exchange rate is not a coincidence. Nothing in this week mentioned prices, yet a price has appeared: two apples per pilchard. Week 2 shows that if you hand this economy a competitive market and let it find its own price ratio, the number it finds is p₁/p₂ = 2 — and the allocation it lands on is precisely (10, 20) and (8, 32). The market does not merely reach an efficient point. In this economy it reaches this one.
Flashcards

Tap card to reveal answer

Quick-reference note sheet
Your score
0 / 13
Self-check — fill in the figures
Type each value, then press Check answers. Cells turn green when correct; any you miss show the right figure beside them.