Chapter 5 introduced consumer theory through utility maximization and the Lagrangian. This chapter strips away the crutch of specific functional forms and builds the theory from axiomatic foundations. We ask: when can preferences be represented by a utility function? What properties must demand functions satisfy? And under what conditions does a system of competitive markets allocate resources efficiently?
The shift in method is from computation to proof. Part II solved optimization problems. Part III proves theorems — establishing which results are robust and which depend on special assumptions.
By the end of this chapter, you will be able to:
State the preference axioms and the conditions for utility representation
Define WARP and SARP and test revealed preference consistency
Derive the expenditure function and Hicksian demand using duality
State and verify properties of the Slutsky matrix
Define Walrasian equilibrium and prove the First Welfare Theorem
State the Second Welfare Theorem and explain its policy implications
Prerequisites: Chapters 6–7. Mathematical prerequisites: real analysis basics (open/closed sets, continuity, fixed-point theorems), convex analysis, matrix algebra. See Appendix A.
11.1 Choice Theory: Axioms and Utility Representation
Preference Axioms
Preference relation.A binary relation $\succsim$ on a consumption set $X \subseteq \mathbb{R}^n_+$. We define: $x \succ y$ (strict preference) if $x \succsim y$ and not $y \succsim x$; and $x \sim y$ (indifference) if $x \succsim y$ and $y \succsim x$.
The standard axioms:
Axiom 1 (Completeness).For all $x, y \in X$: $x \succsim y$ or $y \succsim x$ (or both).
Axiom 2 (Transitivity).For all $x, y, z \in X$: if $x \succsim y$ and $y \succsim z$, then $x \succsim z$.
Axiom 3 (Continuity).For all $y \in X$, the sets $\{x : x \succsim y\}$ and $\{x : y \succsim x\}$ are closed. Equivalently, the strict preference sets $\{x : x \succ y\}$ and $\{x : y \succ x\}$ are open.
Theorem (Debreu).If $\succsim$ is complete, transitive, and continuous, then there exists a continuous utility function $u: X \to \mathbb{R}$ such that $x \succsim y \iff u(x) \geq u(y)$.
Proof sketch. Fix a ray $\{te : t \geq 0\}$ where $e = (1,1,\ldots,1)$. For each $x$, by completeness and continuity, there exists a unique $t(x) \geq 0$ such that $x \sim t(x)e$. Set $u(x) = t(x)$. Transitivity ensures the representation is consistent; continuity ensures $u$ is continuous.
Intuition
What this says: If your preferences never contradict themselves and don't jump around, they can be summarized by a single number attached to each bundle — that number is utility. Debreu's theorem is the guarantee that the summary exists: assign every bundle the point on a fixed reference line you'd be equally happy with, and that gives you a consistent score.
Why it matters: It means economists don't have to assume a utility function — they earn it from three plain conditions on choice. Everything downstream (demand, the welfare theorems, mechanism design) leans on this guarantee. No representation, no objective to maximize.
What changes: Drop continuity and the guarantee breaks: lexicographic preferences (always prefer more of good 1, breaking ties with good 2) are complete and transitive but jump, so no continuous utility number can track them. Continuity is the condition that rules out those jumps.
In Full Mode, the theorem and its proof sketch build the utility number explicitly from a reference ray.
The utility function is ordinal — any monotonic transformation $v = g(u)$ with $g' > 0$ represents the same preferences. Cardinal properties (magnitudes of utility differences) are meaningless.
Additional Properties
Monotonicity (more is better).If $x \geq y$ (component-wise) and $x \neq y$, then $x \succ y$.
Convexity.If $x \succsim y$, then $\lambda x + (1-\lambda)y \succsim y$ for all $\lambda \in [0,1]$. Convexity means indifference curves are convex to the origin — the consumer prefers mixtures.
Strict convexity.If $x \succsim y$, $x \neq y$, and $\lambda \in (0,1)$, then $\lambda x + (1-\lambda)y \succ y$. Strict convexity guarantees unique optimal bundles.
Example 11.1a — Checking Preference Axioms
Consider lexicographic preferences on $\mathbb{R}^2_+$: $x \succ y$ if $x_1 > y_1$, or $x_1 = y_1$ and $x_2 > y_2$.
Completeness: Satisfied — for any $x, y$, either $x_1 > y_1$, $y_1 > x_1$, or $x_1 = y_1$ and we compare $x_2, y_2$.
Transitivity: Satisfied — if $x \succ y$ and $y \succ z$, then $x \succ z$ (follows from transitivity of $>$ on $\mathbb{R}$).
Continuity:Fails. Consider $y = (1, 1)$. The set $\{x : x \succ y\}$ includes $(1, 1.5)$ but not $(0.999, 100)$. The "at least as good" set is not closed — there is a jump at $x_1 = 1$.
Consequence: No continuous utility function represents lexicographic preferences. This shows that continuity is essential for Debreu's utility representation theorem.
Instead of assuming preferences, we can infer them from observed choices.
Weak Axiom of Revealed Preference (WARP).If bundle $x$ is chosen when bundle $y$ is affordable (i.e., $p \cdot x \geq p \cdot y$ where $p$ is the price vector), then $y$ is never chosen when $x$ is affordable.
Formally: if $x$ is revealed preferred to $y$ ($xRy$: $x$ chosen at prices where $y$ was affordable), then $y$ is not revealed preferred to $x$.
Strong Axiom of Revealed Preference (SARP).The revealed preference relation has no cycles: there is no sequence $x^1 R x^2 R \cdots R x^k R x^1$.
SARP is necessary and sufficient for observed choices to be consistent with utility maximization (Afriat's theorem). WARP is necessary but not sufficient in general (though it is sufficient with two goods).
Example 11.1 — WARP Check
A consumer's choices at two price-income situations:
Situation
Prices $(p_1, p_2)$
Chosen bundle $(x_1, x_2)$
Expenditure
A
(1, 2)
(4, 2)
8
B
(2, 1)
(2, 4)
8
Check WARP: At prices A, could the consumer afford bundle B? \$1(2) + 2(4) = 10 > 8$. No. At prices B, could the consumer afford bundle A? \$1(4) + 1(2) = 10 > 8$. No. WARP is satisfied — the data are consistent with utility maximization.
Interactive: Revealed Preference Checker
Enter price vectors and chosen bundles for up to 6 observations. The checker will test WARP and SARP automatically.
Obs.
$p_1$
$p_2$
$x_1$
$x_2$
Expenditure
1
8.0
2
8.0
3
6.0
4
—
5
—
6
—
Click "Check WARP & SARP" to analyze the data.
Interactive 10.1. Enter price-bundle observations and test for revealed preference consistency. WARP checks direct pairwise reversals; SARP checks for cycles of any length. Violations are highlighted with explanations.
11.3 Duality: Expenditure Function and Hicksian Demand
Chapter 5 solved the primal problem: maximize utility subject to a budget. The dual problem minimizes expenditure to achieve a target utility level.
The Expenditure Minimization Problem
Expenditure function.$e(p, \bar{u}) = \min_{x \geq 0} p \cdot x$ subject to $u(x) \geq \bar{u}$. It gives the minimum cost of achieving utility level $\bar{u}$ at prices $p$. The expenditure function is homogeneous of degree 1 in prices and concave in prices.
$$e(p, \bar{u}) = \min_{x \geq 0} \; p \cdot x \quad \text{subject to} \quad u(x) \geq \bar{u}$$(Eq. 11.1)
Hicksian (compensated) demand.The demand function $h(p, \bar{u})$ that solves the expenditure minimization problem. It shows how consumption responds to price changes holding utility constant (compensating the consumer for the price change). Unlike Marshallian demand, Hicksian demand isolates the pure substitution effect.
The solution is the Hicksian (compensated) demand $h(p, \bar{u})$:
Shephard's lemma.Hicksian demand can be recovered directly from the expenditure function by differentiation: $h_i(p, \bar{u}) = \partial e(p, \bar{u}) / \partial p_i$. This is the dual analogue of Roy's identity.
Homogeneous of degree 1 in $p$: $e(tp, \bar{u}) = te(p, \bar{u})$
Non-decreasing in $p$: Higher prices mean more expenditure to reach $\bar{u}$
Concave in $p$: $e(\lambda p + (1-\lambda)p', \bar{u}) \geq \lambda e(p, \bar{u}) + (1-\lambda)e(p', \bar{u})$
Non-decreasing in $\bar{u}$: Higher target utility means more expenditure
Connecting Primal and Dual
The indirect utility function $V(p, m)$ gives the maximum utility achievable at prices $p$ with income $m$:
$$V(p, m) = \max_{x} \; u(x) \quad \text{s.t.} \quad p \cdot x \leq m$$
The key duality relationships:
$$e(p, V(p, m)) = m$$(Eq. 11.3)
$$V(p, e(p, \bar{u})) = \bar{u}$$(Eq. 11.4)
$$h(p, \bar{u}) = x(p, e(p, \bar{u}))$$(Eq. 11.5)
Roy's identity.Marshallian demand can be recovered from the indirect utility function: $x_i(p, m) = -(\partial V / \partial p_i) / (\partial V / \partial m)$. A price increase reduces welfare in proportion to the quantity consumed, scaled by the marginal utility of income.
Roy's identity provides a shortcut for deriving Marshallian demand from the indirect utility function:
$$x_i(p, m) = -\frac{\partial V / \partial p_i}{\partial V / \partial m}$$(Eq. 11.6)
Intuition
What this says: Two shortcuts fall out of asking the same question from opposite sides. Shephard's lemma: the cheapest way to hit a fixed happiness target shifts, as a price rises, in exact proportion to how much of that good you were buying — so the demand curve is just the slope of the cost-to-stay-happy function. Roy's identity: your demand for a good is how fast your best-affordable happiness falls when its price ticks up, divided by how much one more dollar would have helped.
Why it matters: Demand can be recovered by differentiating a single function instead of re-solving an optimization from scratch. The minimize-cost and maximize-utility problems are two views of one choice, so the answer to either hands you the other for free. This duality is the engine behind welfare measurement and the Slutsky decomposition that follows.
What changes: Raise a price and both shortcuts move the same way: Shephard says you substitute away from the now-expensive good along the constant-utility path; Roy says your achievable welfare drops by roughly the quantity you were buying. The bigger your purchases of a good, the harder its price increase bites.
In Full Mode, Eqs. 11.2 and 11.6 state Shephard's lemma and Roy's identity as derivatives of the expenditure and indirect-utility functions.
Intuition for Roy's identity: A small increase in $p_i$ has two effects on welfare (measured by $V$): (1) it directly reduces utility by making good $i$ more expensive (the numerator $\partial V/\partial p_i < 0$), and (2) the magnitude of this effect is proportional to how much of good $i$ the consumer buys ($x_i$) times the marginal utility of income ($\partial V/\partial m$). Dividing (1) by the marginal utility of income gives the quantity of good $i$.
As $\rho \to 0$ (elasticity of substitution $\sigma = 1/(1-\rho) \to 1$), this converges to the Cobb-Douglas case.
Interactive: Duality Explorer
Cobb-Douglas utility $u = x_1^{0.5} x_2^{0.5}$ with income $m = 10$. Slide $p_1$ to see how all three representations — budget-line tangency, Marshallian demand, and expenditure function — encode the same information.
Interactive 11.2. Three views of the same consumer. Left: indifference curve tangent to budget line (primal). Center: Marshallian demand for good 1 as a function of $p_1$. Right: expenditure function $e(p_1, p_2, \bar{u})$ needed to achieve the current utility level. All three encode the same preferences.
11.4 The Slutsky Matrix
Slutsky matrix.The $n \times n$ matrix $S$ with entries $S_{ij} = \partial h_i / \partial p_j$, measuring substitution effects between goods. If demand is generated by utility maximization, $S$ must be symmetric and negative semidefinite. These are testable restrictions on observed demand.
The Slutsky equation from Chapter 5 (Eq. 6.7) generalizes to a matrix. Define the Slutsky (substitution) matrix with entries:
If demand is generated by utility maximization, the Slutsky matrix must be:
Symmetric: $S_{ij} = S_{ji}$ (cross-substitution effects are equal)
Negative semidefinite: $v'Sv \leq 0$ for all vectors $v$ (own-substitution effects are non-positive: $S_{ii} \leq 0$)
$S \cdot p = 0$: Compensated demand is homogeneous of degree zero in prices
Intuition
What this says: The Slutsky matrix is bookkeeping for two facts about substitution. First, cross-effects are symmetric: how good A's compensated demand responds to good B's price exactly mirrors how B responds to A's price. Second, own-substitution effects never go the wrong way — raise a good's price, holding happiness fixed, and you buy weakly less of it, never more. That is all "symmetric and negative semidefinite" means.
Why it matters: These two facts are testable. A demand system that an analyst estimates from real data either obeys them or it doesn't — and if it doesn't, no rational, utility-maximizing consumer could have produced it. The matrix turns the abstract claim "people maximize utility" into a checkable pattern in observed purchases.
What changes: Run it the other way (the integrability theorem): if a demand system does satisfy symmetry, negative semidefiniteness, homogeneity, and budget exhaustion, then some utility function must have generated it. The restrictions aren't just necessary — they're enough to reconstruct preferences from choices.
In Full Mode, Eq. 11.7 defines the matrix and the listed properties state symmetry, negative semidefiniteness, and homogeneity formally.
These are testable restrictions — if observed demand violates them, it cannot have been generated by a rational consumer maximizing a well-behaved utility function.
Integrability.Conversely, if a demand system satisfies: (a) Walras' law ($p \cdot x(p,m) = m$), (b) homogeneity of degree zero, (c) Slutsky symmetry and negative semidefiniteness — then there exists a utility function that generates it. This is the integrability theorem.
Adjust the price of good 1 to see how Marshallian demand, Hicksian (compensated) demand, and the income effect respond. Uses Cobb-Douglas utility $u(x_1,x_2)=x_1^a x_2^{1-a}$ with $a=0.6$, $p_2=3$, $m=120$.
1 (cheaper)4 (more expensive)
Figure 11.2. Left: Slutsky decomposition in commodity space. The original bundle (blue), compensated bundle (orange, on original indifference curve at new prices), and new bundle (green). The substitution effect moves from blue to orange; the income effect moves from orange to green. Right: Slutsky matrix entries $S_{11}$ and $S_{12}$ as $p_1$ varies, confirming negative semidefiniteness ($S_{11} \leq 0$) and symmetry.
Exchange economy.An economy with $I$ consumers and $L$ goods but no production. Each consumer has an initial endowment $\omega_i$ and preferences $\succsim_i$. Trade occurs at market prices; the question is whether a set of prices exists that clears all markets simultaneously.
Consider an economy with $I$ consumers and $L$ goods. Consumer $i$ has endowment $\omega_i \in \mathbb{R}^L_+$ and preferences $\succsim_i$.
At prices $p$, consumer $i$'s wealth is $m_i = p \cdot \omega_i$. She demands $x_i(p, m_i)$.
Walras' law.For any price vector $p$: $p \cdot z(p) = 0$. The total value of excess demand is always zero. This follows from budget exhaustion: $p \cdot x_i = p \cdot \omega_i$ for each consumer.
Implications: (1) If $L - 1$ markets clear, the $L$th clears automatically. (2) Only relative prices matter — we can normalize one price to 1 (the numeraire).
Existence
Theorem (Arrow-Debreu, 1954).Under standard conditions (continuous, strictly convex, locally nonsatiated preferences; positive aggregate endowment of each good), a Walrasian equilibrium exists.
Proof strategy (sketch). Normalize prices to the unit simplex $\Delta$. Define a price-adjustment map $f: \Delta \to \Delta$ that raises the price of goods in excess demand. By Brouwer's fixed-point theorem, $f$ has a fixed point $p^*$. At the fixed point, $z(p^*) = 0$ — all markets clear.
General equilibrium is the formal culmination of the marginalist program: value, fully formalized, becomes the equilibrium price vector. The intellectual lineage — Walras's 1874 vision through to the Arrow-Debreu existence proof — is traced in History of Economic Thought, Ch. 5 (The Marginalist Revolution and Formalization).
History of economic thought — value lineage
The Edgeworth Box
Edgeworth box.A diagram for a 2-consumer, 2-good exchange economy. The box dimensions equal total endowments. Consumer 1's origin is at the bottom-left, consumer 2's at the top-right. Every point in the box is a feasible allocation; the contract curve connects all Pareto-efficient points (tangencies of indifference curves).
For a 2-consumer, 2-good economy, the Edgeworth box provides a complete visualization. The box dimensions equal total endowments. Consumer 1's origin is at bottom-left, consumer 2's at top-right. Every point in the box is a feasible allocation.
Interactive: Edgeworth Box
Two consumers with Cobb-Douglas preferences. Drag the endowment point to explore how the Walrasian equilibrium, contract curve, and core change.
Figure 11.1 (Interactive). The Edgeworth box. The orange dot is the endowment. The green dot is the Walrasian equilibrium. The red curve is the contract curve (all Pareto-efficient allocations). The shaded core region shows allocations both consumers prefer to the endowment. The budget line passes through the endowment with slope $-p_x/p_y$.
First Welfare Theorem.If preferences are locally nonsatiated, then every Walrasian equilibrium allocation is Pareto optimal.
Pareto optimal (efficient).An allocation $x^*$ is Pareto optimal if there is no other feasible allocation $x'$ such that $u_i(x'_i) \geq u_i(x_i^*)$ for all $i$ and $u_j(x'_j) > u_j(x_j^*)$ for some $j$.
Proof. We proceed by contradiction. Suppose the Walrasian equilibrium allocation $x^*$ at prices $p^*$ is not Pareto optimal. Then there exists a feasible allocation $x'$ with everyone at least as well off and someone strictly better off.
Step 1. For consumer $j$ who is strictly better off: since $x_j^*$ was utility-maximizing and $x_j'$ is strictly preferred, $x_j'$ must have been unaffordable: $p^* \cdot x_j' > p^* \cdot \omega_j$.
Step 2. For every consumer $i$: by local nonsatiation, $p^* \cdot x_i' \geq p^* \cdot \omega_i$.
What this says: Suppose the market outcome could be rearranged to help someone with no one else losing. That lucky person would now be holding a bundle they prefer to what they bought at equilibrium — but if they preferred it and could afford it, they would have bought it already. They didn't, so it must have been out of reach. Add up everyone's spending and the "improved" allocation costs more than the economy actually owns. Impossible. So no such improvement exists: the equilibrium is already efficient.
Why it matters: This is Adam Smith's invisible hand stated exactly. Self-interested traders, each minding only their own budget, land on an allocation no central planner could improve without making someone worse off — and the proof needs no benevolence, no coordination, just prices everyone faces in common.
What changes: The argument leans on only two things: people don't leave money on the table (local nonsatiation) and they spend their whole budget. Break the shared-price assumption — market power, externalities, missing markets, private information — and the "they could have bought it" step fails. That is precisely where real markets stop being efficient.
In Full Mode, the four-step proof derives the contradiction from budget exhaustion and feasibility.
The proof uses only local nonsatiation and budget exhaustion. It does not require convexity, differentiability, or any specific functional form. This generality is what makes the theorem powerful.
Interpretation. The First Welfare Theorem is the formal statement of Adam Smith's "invisible hand." Competitive markets produce an allocation that no rearrangement can improve upon without making someone worse off. But the assumptions (complete markets, price-taking, no externalities, no public goods, full information) define exactly when the invisible hand fails.
History of economic thought — value lineage
Example 11.6 — First Welfare Theorem in a 2-Consumer Economy
From Example 11.4, the equilibrium is $x_1^* = y_1^* = x_2^* = y_2^* = 2$ at $p_x = p_y$.
Check Pareto optimality: At the equilibrium, $MRS_1 = y_1/x_1 = 1$ and $MRS_2 = y_2/x_2 = 1$. Since $MRS_1 = MRS_2 = p_x/p_y$, the indifference curves are tangent — the allocation is on the contract curve.
Verify no Pareto improvement: Any reallocation giving Consumer 1 more of good $x$ (say $x_1 = 3$) requires $x_2 = 1$. Then $u_1 = \sqrt{3 \cdot y_1}$ and $u_2 = \sqrt{1 \cdot y_2}$ with $y_1 + y_2 = 4$. For Consumer 1 to gain ($u_1 > \sqrt{4} = 2$), we need $y_1 > 4$, so $y_1 > 4/3$, leaving $y_2 < 8/3$, giving $u_2 = \sqrt{8/3} < 2 = u_2^*$. Consumer 2 is worse off. No Pareto improvement exists.
Interactive: First Welfare Theorem Visualization
The Walrasian equilibrium lies on the contract curve (Pareto efficient). Toggle "Pareto improvements?" to verify: at the equilibrium, the lens-shaped region where both consumers can gain is empty. At the endowment, it is not.
At the Walrasian equilibrium: No Pareto improvements exist — the lens-shaped region is empty. This IS the First Welfare Theorem.
Interactive 10.3. Toggle between viewing the equilibrium (where no Pareto improvements exist) and the endowment (where the shaded lens shows mutually beneficial trades). The equilibrium's position on the contract curve proves efficiency visually.
Take
"Every billionaire is a policy failure" — viral slogan, popularized by Dan Riffle / AOC's office
Dan Riffle, AOC's former policy aide, turned this line into a social media mantra — shared millions of times, printed on T-shirts, chanted at rallies. The claim is stark: billionaires don't exist because they created extraordinary value. They exist because the system is broken — tax loopholes, monopoly power, rigged rules. The First Welfare Theorem you just proved gives you the tools to test this precisely: does extreme wealth concentration represent the market working correctly (and we just dislike the endowment), or the market failing (and efficiency is not achieved)?
Advanced
The popular version
"Every billionaire is a policy failure" collapses a genuinely complex question into a bumper sticker. The slogan treats all billionaire wealth as rent extraction, as if Jeff Bezos, a Russian oligarch who bought state oil assets at gunpoint, and a pharmaceutical CEO exploiting patent monopolies all got rich through the same mechanism.
It conflates wealth with income and ignores that most billionaire wealth is equity: claims on future profits, not cash extracted from workers' paychecks. It treats the economy as zero-sum when the welfare theorems are precisely about how exchange creates surplus.
The opposing person-level failure is equally bad: "Billionaires are the economy's heroes — they earned every penny." This ignores that many fortunes depend on market power, network effects, IP rents, and political influence, mechanisms that have nothing to do with the competitive market the First Welfare Theorem describes. Neither side engages with what the theorem actually says or what its conditions actually require.
Riffle's claim at its strongest
Here's the version of "every billionaire is a policy failure" that would survive an economics seminar. The First Welfare Theorem requires competitive markets: price-taking behavior, no externalities, no market power, no barriers to entry. Many billionaire fortunes violate these conditions systematically.
Platform monopolies (Google, Meta, Amazon) benefit from network effects that create winner-take-all dynamics: these are natural monopolies with massive barriers to entry, not the competitive markets the theorem describes. Pharmaceutical fortunes depend on patent rents, which are government-granted monopolies. Finance fortunes benefit from too-big-to-fail subsidies, information asymmetries, and regulatory capture. Even "legitimate" tech fortunes rely on non-rival goods (software, algorithms) where marginal cost is zero, so standard competitive pricing ($P = MC$) would generate zero revenue; profits require market power.
The deeper structural argument: Piketty's $r > g$ result shows that when the return on capital exceeds the growth rate, wealth concentrates mechanically over time regardless of individual merit. The Second Welfare Theorem says any efficient allocation can be achieved with lump-sum transfers, but those transfers don't exist. So Riffle's instinct has formal backing: the policies that would prevent extreme concentration (lump-sum redistribution, competition enforcement, rent taxation) are either politically impossible or deliberately weakened.
The strongest case that billionaires are NOT a policy failure
Bezos didn't extract wealth from a fixed pie; he created Amazon, which expanded the pie. Consumer surplus from Amazon (lower prices, faster delivery, wider selection) is estimated at hundreds of billions per year. The fact that Bezos captured some of that surplus as profit is not a policy failure. It's the market's reward mechanism working.
The formal argument: the First Welfare Theorem says competitive equilibrium is Pareto efficient: no reallocation makes someone better off without making someone else worse off. Confiscating Bezos's wealth and redistributing it would make many people better off (diminishing marginal utility), but it would also make Bezos worse off. That's a Pareto comparison, not a failure. The theorem doesn't promise equality; it promises efficiency. If you want equality, use the Second Welfare Theorem: redistribute endowments, then let markets clear. That's an equity intervention, not a failure correction.
Furthermore, not all billionaires are created equal. Bezos built a logistics empire delivering real consumer value. Russian oligarchs acquired state assets in rigged privatization auctions. Riffle's slogan, "every billionaire," lumps value creators with rent extractors, innovators with oligarchs. That's rhetoric.
The judgment
Was Riffle right that every billionaire is a policy failure? No. The word "every" is where the claim breaks.
Some billionaire fortunes are genuine market failures: fortunes built on network effects, patent rents, regulatory capture, or political connections involve market power that violates the conditions of the First Welfare Theorem. These are departures from competitive equilibrium, and the resulting wealth concentration is not Pareto optimal. Other fortunes are closer to the welfare theorem ideal: they reflect surplus creation in reasonably competitive markets, and calling them "policy failures" requires a definition of failure that includes any outcome you dislike.
But even the "legitimate" billionaires expose a limitation the welfare theorems themselves acknowledge: Pareto optimality says nothing about equity. A world with one trillionaire and seven billion in poverty is Pareto optimal if redistribution makes the trillionaire worse off. Riffle's slogan captures a real truth: that policy choices (weak antitrust, capital gains preferences, lobbying access) amplify wealth concentration beyond what competitive markets alone would produce.
But "every billionaire" replaces analysis with vibes. The empirical question, how much billionaire wealth is value creation versus rent extraction, is genuinely contested, the mix differs across individuals, and blanket claims in either direction are wrong.
11.7 The Second Welfare Theorem
Second Welfare Theorem.Under convexity assumptions (convex preferences, convex production sets), any Pareto optimal allocation can be achieved as a Walrasian equilibrium — after appropriate redistribution of endowments (lump-sum transfers of wealth).
Intuition
What this says: Pick any efficient outcome you'd want — any fair, equal, or otherwise-desirable split that wastes nothing. The Second Welfare Theorem says a market can reach it. You don't override prices; you hand out the right starting endowments first, then let competitive trade run. Equilibrium does the rest.
Why it matters: It seems to separate two arguments people usually tangle: how big the pie is (efficiency) and how it's sliced (equity). In principle you can have both — choose the distribution you want, then keep all the efficiency of markets. The political left and right can both, on paper, get what they want from the same mechanism.
What changes: The catch the convexity assumption hides is the word "lump-sum." Reaching the chosen split requires transfers that don't distort anyone's choices — which means the government must already know each person's type without watching their behavior. Real taxes (income, wealth, capital gains) react to behavior, so they distort. The clean separation is true in the model and unreachable in practice.
In Full Mode, the theorem statement names the convexity conditions and the lump-sum-transfer requirement precisely.
Interpretation. The Second Welfare Theorem says efficiency and equity are separable problems. Society can choose any Pareto-efficient distribution through two steps:
Redistribute endowments using lump-sum transfers
Let markets operate from the new endowments
The markets will then produce a competitive equilibrium that is both efficient (by the First Welfare Theorem) and achieves the desired distribution.
Why it matters for policy. Don't distort markets to achieve equity (that sacrifices efficiency). Instead, use lump-sum transfers to redistribute, then let markets work. The right-wing implication: let markets operate freely. The left-wing implication: redistribute as much as you want. Both can be achieved simultaneously — in theory.
Why it fails in practice. Lump-sum transfers require information about individuals' types that the government does not have. Real-world redistribution uses distortionary taxes (income, capital gains, wealth) that change incentives and create deadweight loss. This information problem is the subject of mechanism design (Chapter 12) and optimal taxation (Chapter 16).
Core Equivalence
In large economies, the set of core allocations (allocations that no coalition can improve upon) shrinks to the set of Walrasian equilibrium allocations. This is the core equivalence theorem — competitive equilibrium is the unique outcome that survives competition among all possible coalitions.
Maya's Enterprise
We model Maya's lemonade market as a 2-consumer, 2-good Edgeworth box exchange economy.
Setup: Maya and Alex. Two goods: lemonade ($L$) and cookies ($C$). Maya starts with 45 lemonade and 0 cookies. Alex starts with 0 lemonade and 40 cookies.
By the First Welfare Theorem, this allocation is Pareto optimal.
The Historical Lens
Arrow-Debreu (1954): The Existence Proof. Kenneth Arrow and Gerard Debreu proved that a competitive equilibrium exists under weak assumptions (convex preferences, no externalities). Using Kakutani's fixed-point theorem, they showed that a set of prices exists clearing all markets simultaneously — formalizing Adam Smith's "invisible hand" two centuries after The Wealth of Nations.
The mathematical achievement was remarkable: reducing the problem to showing that a certain correspondence (excess demand as a function of prices) satisfies the conditions for a fixed point. The result required only local nonsatiation and convexity — not differentiability or specific functional forms.
Debreu's Theory of Value (1959) distilled this framework into a rigorous axiomatic system, earning him the 1983 Nobel Prize. Arrow had already received the Nobel in 1972 for his broader contributions to general equilibrium and social choice. Their existence proof remains the mathematical foundation for welfare economics and the two welfare theorems proved in this chapter.
Where this proof sits in the history of ideas — the marginalist program that started with Walras and reached its formal peak in Arrow-Debreu — is the subject of History of Economic Thought, Ch. 5 (The Marginalist Revolution and Formalization).
Summary
Preference axioms (completeness, transitivity, continuity) guarantee the existence of a continuous utility function (Debreu's theorem). Additional properties (monotonicity, convexity) ensure well-behaved demand.
Revealed preference (WARP, SARP) allows testing rationality from observed choices without assuming utility. SARP is necessary and sufficient for consistency with utility maximization.
Duality connects the primal (utility maximization) and dual (expenditure minimization) problems. The expenditure function $e(p, \bar{u})$ and Hicksian demand $h(p, \bar{u})$ are linked by Shephard's lemma. Roy's identity recovers Marshallian demand from the indirect utility function.
The Slutsky matrix must be symmetric and negative semidefinite if demand is utility-generated. These are testable restrictions; the integrability theorem says they are also sufficient.
Walrasian equilibrium requires all markets to clear simultaneously. Walras' law ($p \cdot z(p) = 0$) means one market clears automatically. Existence follows from fixed-point arguments (Arrow-Debreu).
The First Welfare Theorem: every competitive equilibrium is Pareto optimal. The Second Welfare Theorem: every Pareto optimum can be achieved as a competitive equilibrium after lump-sum redistribution.
Preferences are defined by $x \succsim y \iff x_1 + x_2 \geq y_1 + y_2$. Verify completeness, transitivity, and continuity. Write down a utility function that represents these preferences.
A consumer makes the following choices: at prices (2, 1) she buys (3, 4); at prices (1, 3) she buys (5, 1). Check WARP.
For Cobb-Douglas utility $u = x_1^{1/3}x_2^{2/3}$: (a) derive the expenditure function, (b) verify Shephard's lemma, (c) verify Roy's identity.
In a 2-consumer, 2-good exchange economy: $u_1 = x_1 y_1$, $\omega_1 = (6, 2)$; $u_2 = x_2 y_2$, $\omega_2 = (2, 6)$. Find the Walrasian equilibrium prices and allocation.
Apply
A consumer's observed demand function is $x_1 = m/(p_1 + p_2)$ and $x_2 = m/(p_1 + p_2)$. (a) Check Walras' law. (b) Check homogeneity of degree zero. (c) Compute the Slutsky matrix and check symmetry and negative semidefiniteness. (d) Can this demand be generated by utility maximization?
Explain why the First Welfare Theorem does not apply to an economy with externalities (connect to Chapter 4). Identify the specific assumption that fails.
The Second Welfare Theorem says any efficient allocation can be achieved via competitive markets with lump-sum transfers. Explain why, in practice, governments use distortionary taxes instead. What information problem makes lump-sum transfers infeasible?
Using the Edgeworth box, illustrate: (a) an allocation in the core but not a competitive equilibrium, (b) a Pareto improvement from the endowment point, (c) why the endowment point itself is generally not Pareto efficient.
Challenge
Prove that if the expenditure function $e(p, \bar{u})$ is concave in $p$, then the Slutsky matrix is negative semidefinite. (Hint: the Hessian of a concave function is negative semidefinite, and $\partial^2 e/\partial p_i \partial p_j = \partial h_i/\partial p_j = S_{ij}$.)
Prove the First Welfare Theorem for the 2-consumer, 2-good case with locally nonsatiated preferences. Then identify where the proof breaks down if one consumer has satiated preferences (a bliss point).
In an Edgeworth box economy with Leontief preferences ($u = \min(x, 2y)$) for both consumers, does a Walrasian equilibrium exist? If so, find it. If not, explain which existence condition fails.
State Afriat's theorem precisely. Using a dataset of 4 observations (price vectors and chosen bundles), construct an example where WARP is satisfied but SARP is violated.
Sources
Mas-Colell, Whinston & Green (MWG); Debreu Theory of Value; Arrow & Debreu (1954); Varian Microeconomic Analysis.
