第2章和第3章表明，竞争市场会产生使总剩余最大化的均衡。正如我们在第1章所论述的，价格体系以惊人的效率协调分散的决策。但这一结果依赖于有时并不成立的条件。当这些条件不成立时，市场就会低效地配置资源——某些东西生产过多，另一些则生产不足。
市场效率的条件包括：(1) 没有成本或收益落在交易之外的第三方身上，(2) 物品是竞争性的和排他性的，(3) 买卖双方拥有充分的信息，(4) 存在许多买家和卖家（没有市场势力——将在第7章另行讨论）。当任何一个条件不满足时，就会出现市场失灵——市场均衡不是帕累托有效的。
本章确定了四类市场失灵：外部性、公共物品、公共资源和信息不对称。这些不是需要死记硬背的例外，而是具有共同结构的系统性模式。对于每一类，我们提出相同的问题：市场为什么出错？偏离多远？可以做什么——代价是什么？
外部性无处不在。当工厂污染河流时，它对下游渔民施加了成本，而这些成本并未出现在工厂的成本计算中。当房主维护美丽的花园时，它提高了邻居的房产价值——园丁无法获取的收益。当司机驶入拥堵的高速公路时，她减慢了所有其他司机的速度——她不需要支付的成本。在每种情况下，私人决策者只考虑自己的成本和收益，而不考虑对他人的影响。
当交易对第三方施加成本时，存在负外部性。生产者或消费者基于私人成本做出决策，忽略了对他人施加的成本。结果是：该活动过多。
市场均衡出现在需求（边际收益）等于供给（MPC）的位置。但社会最优数量是需求等于MSC的位置——它考虑了所有成本，包括第三方承担的成本。由于 $MSC > MPC$，社会最优数量低于市场数量。市场过度生产了产生外部性的商品。
外部性造成的无谓损失等于从 $Q^*$（社会最优）到 $Q_M$（市场数量）之间MSC和需求曲线围成的面积。这个三角形代表过度生产给社会带来的净成本——在这些单位上，完全的社会成本超过了消费者的收益。
图 4.1. 负外部性。拖动MEC滑块查看边际外部成本如何在私人成本和社会成本之间产生楔子。MSC曲线与MPC分离，社会最优数量下降，无谓损失三角扩大。最优庇古税等于MEC。悬停查看数值。
负外部性的现实案例：
当交易对第三方产生收益时，存在正外部性。市场对这些商品的生产过少，因为私人收益低估了社会收益。
其中MSB是边际社会收益，MPB是边际私人收益（反映在需求曲线中），MEB是边际外部收益。
正外部性的现实案例：
我们如何纠正外部性？一种方法是：改变价格以反映真正的社会成本。
What this says: The optimal Pigouvian tax equals exactly the damage each extra unit of production imposes on third parties. Set the tax equal to the marginal external cost at the socially optimal quantity, and the polluter's private cost becomes the true social cost.
Why it matters: The tax makes the polluter "internalize" the externality — they now face the full cost their production imposes on society, not just their own costs. The market equilibrium shifts to the social optimum without anyone needing to ban or mandate anything. Prices do the work.
What changes: If the external cost rises (pollution becomes more damaging), the optimal tax rises and the socially optimal quantity falls. If the external cost is zero, no tax is needed — the market already gets it right.
In Full Mode, Eq. 4.3 states this formally.征税后，生产者的有效成本变为 $MPC + t^* = MSC$，市场均衡与社会最优一致。外部性造成的无谓损失被消除。
对于正外部性，庇古补贴等于社会最优数量处的MEB。补贴降低了消费者面临的有效价格，鼓励他们购买更多——将数量推至社会最优水平。
钢铁需求：$P = 100 - Q$。MPC（供给）：$P = 20 + Q$。恒定 $MEC = 10$ 每单位。
市场均衡： \$100 - Q = 20 + Q \Rightarrow Q_M = 40$，$P_M = 60$。
社会最优： $MSC = 30 + Q$。令 \$100 - Q = 30 + Q \Rightarrow Q^* = 35$，$P^* = 65$。
无谓损失： $\frac{1}{2}(10)(5) = 25$。
最优庇古税： $t^* = MEC = \\$10$ 每单位。征税后，生产者面临的成本为 \$10 + Q = MSC$。新均衡：$Q = 35$，$P_B = 65$，$P_S = 55$。无谓损失消除。
税收收入：\$10 \times 35 = \\$150$。庇古税产生"双重红利"——它既纠正了外部性，又增加了财政收入。
图 4.2. 庇古税纠正。在无管制市场和最优税收之间切换。征税后，有效供给曲线上移至MSC，无谓损失被消除。悬停查看数值。
庇古税在理论上运作完美，但面临实际挑战：
Greta Thunberg has called carbon offsets and carbon trading "a scam," arguing that market-based climate solutions let polluters buy their way out of real change. Meanwhile, over 3,500 economists — including 28 Nobel laureates — signed a 2019 statement calling carbon pricing the most cost-effective lever against climate change. One side says price the externality and let markets work. The other says the house is on fire and you're haggling over the water bill. Who's right depends on a question the Pigouvian model can't answer by itself: how fast is fast enough?
入门政府干预的替代方案：让受影响的各方相互谈判。
Proposition (Coase). Let $TC = 0$ and property rights be fully assigned. Then for any initial allocation of rights, the bargaining outcome is Pareto-efficient. The final allocation of resources is invariant to the initial assignment of rights; only the distribution of surplus differs.
What this says: When bargaining is free and property rights are clear, the people involved will always negotiate their way to the efficient outcome — regardless of who starts with the rights. If a factory's pollution costs a farmer more than the factory earns, they'll strike a deal to stop the pollution, no matter who "owns" the right to clean air.
Why it matters: It reframes the externality problem. The issue isn't that externalities exist — it's that transaction costs prevent bargaining. When those costs are low (two neighbors, a barking dog), private deals work. When they're high (millions of people, air pollution), markets fail and we need other tools.
What changes: As transaction costs rise, bargaining becomes harder and eventually fails. As the number of affected parties grows, coordination costs explode — this is why Coase works for neighbor disputes but not for climate change.
In Full Mode, the formal proposition above states the conditions precisely.一家工厂的污染对邻近农民造成每单位\$10的损害。工厂每单位利润为\$10。有效结果：不生产（成本 \$10 > 收益 \$10）。
情形1——农民拥有权利：工厂需要获得污染许可。必须向农民支付 ≥ \$10，但只赚 \$10。无力支付。结果：无污染。有效率。
情形2——工厂拥有权利：农民向工厂支付 \$10 至 \$10 之间的金额以停止污染。双方获益。结果：无污染。有效率。
无论哪种情形结果相同。唯一的区别是财富分配不同。
图 4.3. 科斯谈判。切换产权分配并滑动交易成本。当TC = 0时，无论产权如何分配，都会产生有效结果（不生产）。随着TC上升，谈判剩余缩小，最终谈判失败。悬停查看详情。
科斯定理要求三个在实践中经常不满足的条件：
1. 明确界定的产权。谁拥有清洁空气的权利？拥有稳定气候的权利？在许多外部性情形中——尤其是环境问题——产权是模糊的、有争议的或不可执行的。
2. 低交易成本。谈判必须廉价。科斯定理适用于两个邻居就狗叫问题的协商。但对于空气污染，数百万受影响方需要与数千家污染企业谈判，科斯定理则完全失效。
3. 无策略行为或信息不对称。各方必须诚实谈判。实际上，每一方都有动机虚报自己的成本或收益。钉子户问题即使在存在互利交易的情况下也能阻止协议达成。
科斯定理最有用的地方不是作为实际解决方案，而是作为诊断工具。它揭示了市场无法处理外部性的原因：交易成本。
这两个属性——非竞争性和非排他性——产生了不同的问题。非竞争性意味着有效价格为零（额外用户的边际成本为零）。非排他性意味着私人企业无法收取任何价格。两者结合意味着私人市场无法有效提供公共物品。
| 排他性 | 非排他性 | |
|---|---|---|
| 竞争性 | 私人物品：食品、服装 | 公共资源：海洋鱼类、清洁空气 |
| 非竞争性 | 俱乐部物品：有线电视、收费公路 | 公共物品：国防、灯塔 |
公共物品的有效水平是多少？对于私人物品，效率要求每个消费者满足 $MB_i = MC$。对于公共物品，所有消费者同时消费相同数量。效率要求边际收益之和等于边际成本：
What this says: To decide how much of a public good to provide, add up how much every person values one more unit. If that total exceeds the cost, provide more. The efficient amount is where the combined willingness to pay exactly equals the cost of production.
Why it matters: Unlike private goods, where each person decides for themselves, public goods are shared by everyone simultaneously. So the question is not "does any one person value it enough?" but "does society collectively value it enough?" This is why markets underprovide public goods — no single buyer captures the full social value.
What changes: If more people benefit from the public good, the sum of marginal benefits rises, so the efficient quantity increases. If the cost of provision falls (better technology), the efficient quantity also rises. If some people value it less (free-rider incentives reduce revealed willingness to pay), the measured sum falls and the good is underprovided.
In Full Mode, Eq. 4.4 states the Samuelson condition formally.这就是萨缪尔森条件（Samuelson, 1954）。在图形上，我们纵向加总各个MB曲线（因为每个人消费相同数量），找到总MB等于MC的位置。
3个家庭：$MB_1 = 10 - Q$，$MB_2 = 8 - Q$，$MB_3 = 6 - Q$。边际成本：$MC = 6$。
$\sum MB = 24 - 3Q$。萨缪尔森条件：\$14 - 3Q = 6 \Rightarrow Q^* = 6$ 小时。
私人供给：家庭1在 $MB_1 = MC$ 处提供：\$10 - Q = 6 \Rightarrow Q = 4$ 小时。其他家庭搭便车。供给不足：4小时而非6小时。
图 4.4. 公共物品：纵向加总。调整每个家庭的支付意愿。粗绿色曲线是三条MB曲线的纵向加总。萨缪尔森最优数量是 ΣMB = MC 的位置。私人供给（最高个人 MB = MC 处）总是不足。悬停查看数值。
You now have externalities and public goods. Here's a surprising implication: some redistribution can be justified on pure efficiency grounds — not because inequality is unfair, but because it creates externalities.
Poverty creates negative externalities: crime, disease transmission, reduced human capital formation. Education and health have positive externalities — the social return exceeds the private return. If these externalities are large, redistribution that funds education, healthcare, and poverty reduction is efficiency-enhancing, not just equitable. Public goods — rule of law, infrastructure, basic research — benefit everyone but are funded through taxation that is inherently redistributive. The externality and public goods frameworks from this chapter provide a purely efficiency-based argument for some redistribution, bypassing the equity debate entirely.
The externality argument for redistribution is a means-tested claim, not a blanket justification. Not all redistribution targets externality correction — much of it is pure transfer (Social Security, unemployment insurance). These may be justified on equity grounds but not on efficiency grounds. And the externality magnitudes are disputed: how large are the crime-reduction or human-capital externalities of poverty programs? The efficiency case for redistribution sets a floor for how much redistribution is warranted, not a ceiling — and that floor may be lower than redistributionists assume.
The distinction between "efficiency-based redistribution" and "equity-based redistribution" is analytically useful but politically irrelevant. Real programs do both simultaneously — public education reduces inequality and corrects a positive externality. The political coalition for redistribution doesn't distinguish between efficiency and equity justifications, and the economic analysis often can't cleanly separate them either.
Some redistribution is justifiable purely on efficiency grounds — human capital investment, poverty reduction that addresses negative externalities, public goods provision. This is a stronger claim than "redistribution is fair" because it doesn't require any normative judgment about equity. But it sets a floor, not a ceiling. Whether to go beyond the efficiency-justified level is a question this chapter's tools can't answer. The efficiency framework tells you the cost of redistribution (deadweight loss from taxation) and some of the benefits (externality correction) — but not whether the benefits beyond externality correction are worth the costs.
How do you design redistribution to minimize efficiency costs? That requires tools you don't have yet. Come back in Chapter 12 (§12.1), where mechanism design formalizes the redistribution problem under incentive constraints — the government can't observe ability, only income, so every tax schedule faces an incentive-compatibility constraint. Then in Chapter 16 (§16.7), Ramsey optimal taxation gives quantitative answers: tax inelastic goods more, and optimal top marginal rates are probably 50-70% (Diamond & Saez 2011) — higher than most countries implement but lower than "tax everything" implies.
If poverty creates externalities, the simplest fix is a cash floor. But does UBI reduce the labor supply enough to offset the externality gains?
入门例子比比皆是：海洋鱼类资源、地下水含水层、作为碳汇的大气、公共牧场、高峰时段的公路以及野生动物。在每种情况下，资源都是可耗竭的（竞争性的）但对所有人开放（非排他性的）。
其逻辑与负外部性完全相同。每个渔民多捕一条鱼获得该鱼的全部市场价值，但通过减少剩余鱼群对所有其他渔民施加了成本。私人边际成本低于社会边际成本，因此资源被过度开采。
With $N$ users, each user $i$ maximizes private profit: $\pi_i = B(E) \cdot e_i - c \cdot e_i$, where $B(E) = a - E$ is the diminishing benefit, $E = \sum e_i$ is total extraction, and $c$ is the unit cost. The Nash equilibrium total extraction is $E_N = \frac{N}{N+1}(a - c)$, while the social optimum is $E^* = \frac{a - c}{2}$. As $N \to \infty$, $E_N \to (a - c)$ — the resource is driven to exhaustion.
What this says: Each user grabs more than their fair share because they enjoy the full benefit of extraction but bear only a fraction of the depletion cost. With many users, the resource gets hammered far past the efficient level.
Why it matters: A single owner would extract efficiently (they bear the full cost of depletion). But open access splits the cost across everyone while concentrating the benefit — so each person overextracts. More users means worse overextraction. This is why open-access fisheries collapse.
What changes: Adding more users pushes extraction further past the optimum. Raising the cost of extraction (a tax) or reducing the number of users (quotas, property rights) moves the outcome back toward efficiency.
In Full Mode, the Nash equilibrium derivation above shows this precisely.图 4.5. 公地悲剧。拖动滑块增加用户数量。每个用户的使用量超过其社会最优份额，因为他们忽略了自己对他人施加的资源耗竭外部性。在单一所有者下，开采是有效的；在多个用户下，资源被严重过度开采。悬停查看数值。
1. 产权（私有化）。将所有权分配给个人或企业。所有者内化全部资源耗竭成本。冰岛的个体可转让配额（ITQ）渔业制度是一个成功的例子。
2. 管制。政府对开采施加限制：渔业配额、狩猎季节、水资源使用许可证、排放标准。
3. 庇古税。对每单位开采按等于边际外部成本的税率征税。道路拥堵收费就是一个例子。
4. 社区治理（奥斯特罗姆）。埃莉诺·奥斯特罗姆（2009年诺贝尔奖）研究了在没有私有化或政府管制的情况下成功管理公地的社区。成功需要：明确界定的边界、适应当地条件的规则、用户参与规则制定、有效监控、阶梯式制裁以及可及的冲突解决机制。
You now have the full catalog of market failures — externalities, public goods, common resources. Time for a reality check on the efficiency claim from Chapter 3.
Each market failure has a specific mechanism and, in theory, a specific fix. Externalities: $MSC \neq MPC$, so the market quantity is wrong — fix with a Pigouvian tax or property rights (Coase). Public goods: markets underprovide because of free-riding — fix with government provision funded by taxes. Commons: markets overuse because no one bears the depletion cost — fix with property rights, regulation, or Ostrom-style community management. Information asymmetry: adverse selection collapses markets (Akerlof's lemons); moral hazard changes behavior after contracts — fix with signaling, screening, or regulation. The framework is clean: identify the failure, apply the corrective instrument, restore efficiency.
Government failure. Every market failure diagnosis implies a government intervention, but governments face their own failures — public choice problems (rent-seeking, regulatory capture), information problems (the planner knows less than the market), and political constraints. The cure may be worse than the disease. Coase's deeper point: With low transaction costs, private bargaining resolves externalities regardless of initial property rights. The question isn't "is there a market failure?" but "are transaction costs low enough for private solutions?" The existence of a market failure is necessary but not sufficient for government intervention — you must also show that the government can do better than the flawed market.
The mainstream acknowledges both market failure and government failure. The Coase theorem shifted the debate from "should the government intervene?" to "what are the transaction costs?" Ostrom (Nobel 2009) demonstrated a third path: communities can manage commons without either markets or governments. The framework became comparative institutional analysis — compare the realistic market outcome to the realistic government outcome, not to an idealized planner.
Market failures are real and pervasive — they are not rare exceptions to an otherwise perfect system. Externalities are everywhere (pollution, congestion, network effects), public goods are critical (defense, research, rule of law), and information is always asymmetric. But the existence of a market failure doesn't automatically justify intervention. The right framework is: identify the failure, assess the transaction costs, compare the market outcome to the feasible intervention — not the ideal one. Markets work remarkably well in most settings precisely because the conditions for efficiency hold approximately, even if they never hold perfectly.
The informal treatment here lacks rigor. You know that markets fail under these conditions, but you can't prove exactly when they succeed. Come back in Chapter 11 (§11.6–11.7), where the First and Second Welfare Theorems tell you precisely which conditions must hold for market efficiency — and Greenwald-Stiglitz (1986) proves that when markets are incomplete, competitive equilibria are generically inefficient. Then in Chapter 12 (§12.1–12.5), mechanism design asks: if markets aren't up to the job, can we engineer better institutions?
Greta Thunberg calls market-based climate solutions a scam. 3,500 economists call carbon pricing indispensable. One side says price the externality; the other says the house is on fire. Who's right depends on how fast is fast enough.
入门Sanders' viral rallying cry meets Arrow's 1963 paper. The moral force is real — but declaring a right doesn't solve the allocation problem.
入门市场假设买卖双方拥有足够的信息来做出良好的决策。当一方比另一方知道更多重要信息时——信息不对称——市场可能以可预测的方式失灵。
卖家知道自己的车是可靠的（"好车"，价值 \$10,000）还是有缺陷的（"柠檬"，价值 \$1,000）。买家无法分辨。在50/50的概率下，买家出价 \$1,500。但好车车主拒绝——他们的车值 \$10,000。只有柠檬车被出售。买家意识到这一点后只出价 \$1,000。
结果：优质二手车市场消失了。高质量卖家退出，只剩下低质量卖家。
Let quality $q \in \{H, L\}$ with values $v_H > v_L$. Sellers observe $q$; buyers observe only the prior $\Pr(q = H) = \lambda$. A pooling price $p = \lambda v_H + (1 - \lambda)v_L$ makes type-$H$ sellers exit whenever $p < v_H$ (i.e., $\lambda < 1$). With type-$H$ gone, buyers revise to $\lambda' = 0$, and only lemons trade at $p = v_L$. The market unravels.
What this says: When buyers cannot tell good products from bad, they offer an average price. But that average price is too low for sellers of good products, who walk away. Once good sellers leave, only bad products remain — and buyers adjust their offers downward. The market spirals: quality drops, prices drop, more good sellers exit.
Why it matters: This explains why markets can collapse even when gains from trade exist. Health insurance without mandates, used car markets without warranties, and labor markets with unobservable skill all face this unraveling pressure. The information gap — not bad intentions — destroys the market.
What changes: If buyers gain information (inspections, warranties, reputation), the unraveling slows or stops. If the share of high-quality sellers rises, the pooling price rises and fewer exit. Mandatory participation (insurance mandates) prevents the spiral by keeping good types in the pool.
In Full Mode, the formal setup above shows the unraveling mechanism precisely.逆向选择的现实解决方案：
有了火灾保险，房主可能对防火变得不那么小心。有了医疗保险，患者可能更频繁地就医。道德风险本质上是隐藏行动的问题。解决方案包括：
逆向选择和道德风险在这里仅作直觉介绍。第11章通过揭示原理和机制设计正式化逆向选择。第10章提供了思考信息和激励的正式框架。
Bernie Sanders made this line the centerpiece of his 2016 and 2020 presidential campaigns — viral clips with tens of millions of views, crowd roaring. The moral force is undeniable: Americans spend \$4.5 trillion a year on healthcare and get worse outcomes than countries that spend half as much. But declaring something a "right" doesn't answer the question economics actually asks: who allocates the scarce MRI machines, surgeon hours, and hospital beds — and by what mechanism?
入门玛雅的柠檬水摊产生了正外部性。邻居报告说，玛雅顾客带来的人流增加了附近商店的客流量。估计的边际外部收益为每杯 \$1.30。
城市应该补贴玛雅吗？
$MSB = MB + MEB = (5 - Q/20) + 0.30 = 5.30 - Q/20$。令 $MSB = MPC$：
\$1.30 - Q/20 = 0.50 + Q/20 \Rightarrow Q^{**} = 48$ 杯（相比市场 $Q = 45$）。
每杯 \$1.30 的庇古补贴可以实现这一目标。但城市对玛雅征收了每杯 \$1.50 的税（第3章），将产量压至40杯——方向完全错误。该税收的动机是财政需要，而非效率。理解外部性框架有助于明确我们在权衡什么。
| 标签 | 公式 | 描述 |
|---|---|---|
| 公式 4.1 | $MSC = MPC + MEC$ | 负外部性下的边际社会成本 |
| 公式 4.2 | $MSB = MPB + MEB$ | 正外部性下的边际社会收益 |
| 公式 4.3 | $t^* = MEC$ 在 $Q^*$ | 最优庇古税 |
| 公式 4.4 | $\sum_{i=1}^{N} MB_i = MC$ | 公共物品的萨缪尔森条件 |
Coming in Part II: calculus makes everything precise. The intuitions you built are correct — the math lets you say exactly how much.
