第10章提问：给定偏好和禀赋，竞争性市场能否产生有效结果？答案——是的，在福利定理条件下——将市场机制视为既定。本章反转这个问题：给定期望结果，我们能否设计一个机制来实现它？
机制设计常被称为"逆向博弈论"。不是预测博弈的结果，而是设计博弈以产生期望结果。市场设计将这些思想应用于现实制度——拍卖、匹配市场、频谱分配、肾脏交换。
前置知识：第7章（博弈论基础、纳什均衡）和第10章（福利定理、一般均衡）。
重要文献：Myerson (1981); Vickrey (1961); Clarke (1971); Groves (1973); Gale & Shapley (1962); Roth (2002); Milgrom (2004)。
挑战在于：代理人的类型是私人信息。我们如何让他们如实披露其类型？
图 12.1.机制设计时间线。
机制设计者选择规则（消息空间和结果函数）以实现期望的社会选择函数。
直接机制要求每个代理人简单地报告其类型（私人信息）。如果如实报告是均衡策略——没有代理人能从撒谎中获益——则该机制是激励相容的（IC）。
这是机制设计中最强大的简化——可以说是整个经济理论中最强大的简化。原则上，可能的机制空间是无限大的。拍卖可以有任意数量的轮次、任意竞标规则、任意支付公式。匹配算法可以以任何可想象的方式运行。在所有可能的机制中搜索最优者似乎毫无希望。
显示原理指出：你不必搜索。无论任何机制能实现什么结果，一个直接机制（只需要求每个人如实报告）可以实现相同的结果。因此，机制设计问题简化为：找到最优的分配规则和支付规则作为报告类型的函数，受制于如实报告是最优的约束。这将一个无限广泛的搜索转化为一个明确定义的优化问题。
DSIC更强但更难实现。BIC更弱但允许更多机制。
You now have mechanism design tools — the revelation principle, incentive compatibility, and the distinction between DSIC and BIC. These tools formalize what a government can and cannot achieve when it can't observe people's types directly.
Mechanism design formalizes the redistribution problem with startling clarity. The government wants to transfer from high-ability to low-ability agents but can't observe ability directly — only income, which is a choice variable. The revelation principle says any redistribution scheme can be analyzed as a direct mechanism where agents report their type. The binding constraint is incentive compatibility: high-ability agents must not find it profitable to mimic low-ability agents by working less. A tax-and-transfer system is literally a mechanism — it maps reported incomes to after-tax incomes — and the revelation principle tells you that if any scheme can achieve a redistributive goal, a truthful direct mechanism can too. This is the conceptual foundation of optimal income taxation (Mirrlees 1971): the tax schedule is a mechanism designed to maximize social welfare subject to incentive compatibility.
Incentive compatibility creates an irreducible tradeoff between redistribution and efficiency — and it's worse than the intuitive version. The Myerson-Satterthwaite theorem (§12.4) shows that in bilateral trade with private information, no mechanism simultaneously achieves efficiency, incentive compatibility, individual rationality, and budget balance. Apply this logic to redistribution: the government faces a version of the same impossibility. It cannot design a tax system that fully redistributes, respects incentives, and avoids deadweight loss. Furthermore, the mechanism design framework assumes a benevolent, well-informed planner who knows the distribution of types even if not individual types. In practice, redistributive policy is shaped by political economy — median voters, interest groups, lobbying. The design problem is well-understood; the implementation problem is not.
The mechanism design framework connects directly to optimal income tax theory. Mirrlees (1971) showed that the optimal tax schedule depends on the distribution of abilities and the elasticity of labor supply — both empirical quantities. The mechanism design approach gives the conceptual architecture; the quantitative answers require data. Myerson's optimal auction is structurally identical to optimal taxation: both maximize an objective subject to incentive compatibility and individual rationality. The same math that designs revenue-maximizing auctions designs welfare-maximizing tax schedules.
The efficiency-equity tradeoff is real, but mechanism design makes it precise rather than vague. The tradeoff isn't "redistribution is costly" — it's "redistribution is costly by exactly the amount that incentive compatibility constraints bind." The magnitude depends on specific parameters: how elastic is labor supply? How fat is the tail of the ability distribution? These are empirical questions with empirical answers, not ideological ones. Mechanism design transforms the inequality debate from philosophy into engineering — but the engineering is constrained by informational limits that no cleverness can circumvent.
Mechanism design gives you the framework; optimal tax theory gives the numbers. Come back in Chapter 16 (§16.7) for the Ramsey optimal tax result — tax inelastic goods more — and the quantitative estimates: optimal top marginal rates are probably 50–70% (Diamond & Saez 2011), higher than most countries implement but lower than "tax everything" implies. Then in Chapter 20 (§20.5, §20.8), the problem goes global: within-country inequality is dwarfed by between-country inequality, and the tools for addressing it — institutions, human capital, development interventions — are entirely different from domestic tax design.
Elizabeth Warren's proposal meets mechanism design: the binding constraint on redistribution is incentive compatibility — agents can hide their type. Wealth is harder to hide than income. Does that make wealth taxes better mechanisms?
高级这是机制设计中对应阿罗不可能定理的结果。它表明，在一般社会选择设定下，没有非独裁机制能在占优策略中引出真实偏好。
突破口：限制定义域。在准线性偏好（$U_i = v_i(a) + t_i$，其中 $t_i$ 是货币转移）下，吉巴德-萨特斯韦特障碍被突破。VCG机制通过转移支付实现效率和DSIC。
维克里-克拉克-格罗夫斯（VCG）机制通过货币转移，以如实报告为占优策略实现有效分配。
有效分配：$a^*(\theta) = \arg\max_a \sum_i v_i(a, \theta_i)$ — 最大化总价值。
代理人 $i$ 支付她对他人施加的外部性——有她和没有她时其他人福利的差额。
为什么如实报告是占优策略？在如实报告下，代理人 $i$ 的收益为：
$$v_i(a^*(\theta)) + t_i = v_i(a^*(\theta)) + \sum_{j \neq i} v_j(a^*(\theta_{-i})) - \sum_{j \neq i} v_j(a^*(\theta))$$
这简化为 $\sum_j v_j(a^*(\theta)) - \sum_{j \neq i} v_j(a^*(\theta_{-i}))$。第二项不依赖于 $i$ 的报告。因此 $i$ 通过选择报告来最大化 $\sum_j v_j(a^*(\theta))$ 以最大化其收益——这恰好在她如实报告时发生，因为 $a^*$ 已经最大化了总价值。
输入代理人对单一不可分割物品的估值。计算器计算VCG支付（对单物品等同于第二价格拍卖）。
图 12.2.代理人估值与VCG支付。每个代理人支付其对他人施加的外部性。获胜者支付第二高价值（在单物品拍卖中，VCG简化为维克里拍卖）。
三位市民对一座桥的估值分别为 $v_1 = 30$、$v_2 = 25$、$v_3 = 15$。成本为 $C = 60$。
若 $\sum v_i > C$ 则建造：\$10 > 60$ → 是。
克拉克税支付：
总收取：\$10 + 15 + 5 = 40 < 60$。存在20的预算赤字——VCG通常不能实现预算平衡。每个代理人支付其"枢纽"贡献。
| 形式 | 规则 | 获胜者支付 |
|---|---|---|
| 英式（升序） | 竞标者提高出价；最后竞标者获胜 | 第二高价值（近似） |
| 荷兰式（降序） | 价格下降直到有人认领 | 其出价 |
| 第一价格密封投标 | 最高出价获胜 | 其出价 |
| 第二价格密封投标（维克里） | 最高出价获胜 | 第二高出价 |
维克里拍卖（第二价格密封投标）是DSIC的：每个竞标者的占优策略是按其真实价值 $v_i$ 出价。高于 $v_i$ 出价有以高于价值的价格中标的风险；低于 $v_i$ 出价有在第二高出价低于 $v_i$ 时错失的风险。
这是一个惊人的结果。它表明拍卖形式之间看似巨大的差异——公开与密封、升序与降序、第一价格与第二价格——在这些条件下对期望收入是无关的。
收入等价何时失效：
设置竞标者数量及其价值分布。运行单次拍卖查看个别结果，或运行100轮观察收入等价（各种形式的平均收入趋于一致）。调整风险厌恶滑块以打破等价。
图 12.3.拍卖结果。在单次运行中，由于随机性，各种形式的收入不同。经过100次运行，平均收入趋于一致——展示了收入等价。增加风险厌恶（$\rho > 0$）可以打破等价：第一价格收入高于第二价格。
当卖方想要最大化收入（而非效率）时，迈尔森证明了最优机制使用虚拟价值：
其中 $F$ 是竞标者价值分布的CDF，$f$ 是PDF。
最优拍卖将物品分配给虚拟价值最高的竞标者，前提是其为正值。如果所有虚拟价值均为负，卖方保留物品。这意味着一个保留价——卖方设置等于 $\psi^{-1}(0)$ 的最低出价。
价值在 $[0, 1]$ 上均匀分布：$F(\theta) = \theta$，$f(\theta) = 1$。
$\psi(\theta) = \theta - (1-\theta)/1 = 2\theta - 1$
$\psi(\theta) = 0 \implies \theta = 1/2$。最优保留价 = \$1/2$。
带保留价 \$1/2$ 的第二价格拍卖是最优的：只有当至少一个竞标者的估值超过 \$1/2$ 时，物品才会售出。
对于从Uniform$[0, V_{\max}]$中抽取的价值，虚拟价值为 $\psi(\theta) = 2\theta - V_{\max}$。拖动保留价滑块。收入曲线显示期望收入作为保留价的函数。最优保留价（最大化期望收入）被突出显示。
图 12.4a。虚拟价值函数 $\psi(\theta) = 2\theta - 1$（对于 $U[0,1]$）。保留价设在 $\psi(r) = 0$ 处。估值 $\theta < r$ 的竞标者被排除（红色阴影区域）。
图 12.4b。期望收入作为保留价的函数。绿色圆点标记最大化期望收入的最优保留价。您选择的保留价显示为蓝色圆点。
政府向两家公司之一分配许可证。公司 $i$ 的私人价值 $\theta_i \in \{L, H\} = \{10, 50\}$，各以等概率出现。
提议机制：将许可证分配给报告更高价值的公司；平局时分配给公司1。支付：获胜者支付30。
检验高价值公司（$\theta = 50$）的IC：
如实报告更优。IC对类型 $H$ 成立。
检验低价值公司（$\theta = 10$）的IC：
如实报告更优。IC对类型 $L$ 成立。该机制是激励相容的。
两个竞标者的价值独立地从 $U[0, 100]$ 中抽取。
第二价格拍卖：期望收入 = $E[\text{2nd highest value}] = 100/3 \approx 33.33$。
第一价格拍卖：2个竞标者的最优出价：$b(\theta) = \theta/2$。期望收入 = $E[\max(b_1, b_2)] = E[\max(\theta_1/2, \theta_2/2)] = E[\max(\theta_1, \theta_2)]/2 = (200/3)/2 = 100/3 \approx 33.33$。
两种形式都产生 \$100/3$ 的期望收入，验证了收入等价。第一价格拍卖产生较低的收入波动（每个获胜者恰好支付其价值的一半），而第二价格拍卖的波动较高（支付取决于第二高价值，可能变化很大）。
直觉：卖方想要夸大其成本（以获取更高价格）。买方想要低报其价值（以少付款）。激励相容要求向双方留下"信息租金"。这些租金成本高昂，在预算平衡下，没有足够的剩余来支付双方的租金并确保所有有效交易发生。
私人信息下的现实谈判——工资谈判、二手车购买、并购交易——总是涉及某些低效率。发布价格、声誉系统和标准化合同等制度缓解了这一问题，但无法完全消除。
某些物品不能通过价格分配——我们不应该（或不该）出售学校入学名额、器官移植或住院医师职位。匹配市场使用算法替代。
定理（Gale & Shapley, 1962）。该算法在最多 $n^2$ 轮内终止，并产生稳定匹配——没有未匹配的配对双方都偏好对方而非其当前匹配。
性质：
输入学生和学校的偏好列表。算法动画展示每一轮：提议、暂时接受和拒绝。以逗号分隔的名称输入偏好（例如"W,X,Y,Z"）。
四名学生（A、B、C、D）和四所学校（W、X、Y、Z）。学生提议。
| 学生 | 偏好 | 学校 | 偏好 |
|---|---|---|---|
| A | W > X > Y > Z | W | B > A > D > C |
| B | X > W > Y > Z | X | A > B > C > D |
| C | W > Y > X > Z | Y | C > D > A > B |
| D | Y > W > X > Z | Z | D > C > B > A |
最终匹配：A-W、B-X、C-Y、D-Z。这是稳定的：没有配对想要偏离。使用上面的互动工具逐步验证。
分别运行学生提议和学校提议的Gale-Shapley。比较两个稳定匹配。提议方总是获得其最优稳定匹配；回应方获得其最差稳定匹配。
阿尔文·罗斯（2012年诺贝尔奖，与劳埃德·沙普利共享）将此描述为"经济学家即工程师"——运用经济理论不仅解释世界，还设计改善人们生活的现实制度。
更广泛的启示：市场不是自发产生的自然物体。它们是被设计的制度——决定谁获得什么、以什么价格、通过什么过程的规则、算法和执行机制。设计至关重要。
该市决定拍卖在市中心黄金地段经营柠檬水摊的专营权。三位潜在供应商：Maya（$v_M = 50$/天）、Nate（$v_N = 35$/天）、Olivia（$v_O = 20$/天）。价值从 $U[0, 60]$ 中抽取。
第二价格拍卖（维克里）：占优策略是如实竞标。Maya出价50，Nate出价35，Olivia出价20。Maya获胜，支付35。
最优拍卖（迈尔森）：虚拟价值，其中 $F(\theta) = \theta/60$，$f(\theta) = 1/60$：
$\psi(\theta) = \theta - (60 - \theta) = 2\theta - 60$
保留价：$\psi(\theta) = 0 \implies \theta = 30$。
Maya的虚拟价值：\$1(50) - 60 = 40$。Nate的：\$10$。Olivia的：$-20$（被最优拍卖排除）。
在保留价为30的第二价格拍卖中：Maya获胜，支付 $\max(35, 30) = 35$。
Roth的"经济学家即工程师"。阿尔文·罗斯（2012年诺贝尔奖）将机制设计从纯理论转化为重新设计真实市场的实用学科。他的工作表明，市场是被设计的制度，而非自然现象。
全国住院医师匹配项目（NRMP）：Roth诊断了原始住院医师匹配失败的原因（不稳定性、策略操纵），并使用延迟接受算法重新设计。新系统每年匹配约40,000名住院医师。
肾脏交换：Roth、Sonmez和Unver设计了交换协议，允许不兼容的供体-患者配对通过移植链交换供体，挽救了数千人的生命。这是纯粹的市场设计——在没有价格的情况下创建一个市场。
择校：Roth及其同事用策略防护系统替代了波士顿可操纵的学校分配机制。在旧系统下，如实报告偏好的家长会受到惩罚；在新系统下，诚实总是最优的。
频谱拍卖：Milgrom和Wilson（2020年诺贝尔奖）为FCC设计了组合拍卖，在有效分配频谱许可证的同时筹集了数十亿美元。2017年的激励拍卖单独筹集了198亿美元。
共同线索：经济理论提供蓝图，但实施需要理解具体的制度背景——纯理论所抽象掉的那些"细节"。
You now have the complete toolkit: the welfare theorems told you when markets work (Chapter 11); mechanism design and market design show you what to do when they don't. This is the final stop.
When traditional markets fail — when the welfare theorem conditions don't hold — you can engineer better institutions. The revelation principle says the design space is tractable: focus on direct truthful mechanisms. VCG implements efficient outcomes with dominant-strategy incentives when preferences are quasi-linear. And where prices can't work at all — kidneys can't be bought, school seats can't be auctioned — Gale-Shapley's deferred acceptance produces stable matchings without any monetary transfers. These aren't hypotheticals. Kidney exchange has saved thousands of lives by creating a market where none could exist. School choice redesigns replaced manipulable systems with strategy-proof ones, making honesty the optimal strategy for every parent. Spectrum auctions (Milgrom, Wilson — Nobel 2020) raised billions while allocating licenses efficiently. Roth's "economist as engineer" program demonstrates that economic theory can design real institutions that outperform both unregulated markets and blunt government intervention.
The Myerson-Satterthwaite impossibility is a cold shower for mechanism design optimism: in bilateral trade with private information, no mechanism can simultaneously achieve efficiency, incentive compatibility, individual rationality, and budget balance. This isn't a technical limitation — it's a fundamental impossibility. The success stories of market design (matching, auctions, kidney exchange) share a crucial feature: they operate in structured, well-defined environments where the "rules of the game" are clear and the designer has substantial control. In messier environments — healthcare systems, financial markets, labor markets, macroeconomic policy — the institutional design problem is orders of magnitude harder. The mechanism designer needs to know the distribution of types, the set of feasible allocations, and agents' utility functions. In complex real-world settings, this knowledge is precisely what the designer lacks. The mechanism design revolution may have succeeded in the easy cases while leaving the hard ones untouched.
Market design matured into a pragmatic discipline that takes the limitations seriously. Roth's methodology is explicitly "design, implement, observe, redesign" — not "prove optimality and deploy." The NRMP matching algorithm has been revised multiple times as new problems emerged (couples matching, rural hospital shortages). Spectrum auction formats evolved from simple simultaneous ascending auctions to complex combinatorial designs as the FCC learned from earlier rounds. The profession moved from proving impossibility results to asking: given the impossibilities, what's the best achievable mechanism? Computational mechanism design — integrating algorithmic constraints with incentive constraints — is the active frontier, particularly relevant as digital platforms become the dominant market institutions.
Markets allocate resources efficiently when the welfare theorem conditions hold — and they hold approximately enough to make markets the default for most goods. When they fail, mechanism design offers a genuine alternative: not "let the government decide" but "design an institution whose incentives produce the outcome you want." The success stories are real and important. But mechanism design is not a universal solvent. It works best in structured, well-defined settings. The frontier — digital markets, algorithmic pricing, AI-mediated transactions, platform monopolies — raises questions that existing theory doesn't fully address. The answer to "do markets allocate resources efficiently?" is: yes, when conditions hold; and when they don't, we can sometimes engineer something better — but "sometimes" is doing heavy lifting in that sentence, and the engineering is harder than the theory suggests.
This is the final stop on BQ #7. The arc ran from surplus as benchmark (Ch 3) through market failures (Ch 4), the formal welfare theorems (Ch 11), and now mechanism design. The question "do markets allocate resources efficiently?" turns out to be the wrong question — the right one is "under what conditions, and what can we build when conditions fail?" The answer involves welfare theorems and mechanism design and the practical wisdom that design is constrained by politics, information, and computation. The next frontier is where mechanism design meets behavioral economics (Chapter 19) — agents who aren't fully rational may not respond to incentive-compatible mechanisms the way theory predicts. Bounded rationality may be the binding constraint that mechanism design hasn't yet solved.
Bernie Sanders' rallying cry meets mechanism design: healthcare fails every welfare theorem condition. Can mechanism design do better? Kidney exchange says yes for organs. For the rest of healthcare, the design problem remains unsolved.
中级Khan's antitrust paradox: platform markets are designed institutions — but designed by the platforms, for the platforms. The consumer welfare standard is blind to it.
高级| 标签 | 公式 | 描述 |
|---|---|---|
| 式 12.1 | $U_i(\theta_i, \theta_i) \geq U_i(\hat{\theta}_i, \theta_i)$ for all $\hat{\theta}_i, \theta_{-i}$ | DSIC |
| 式 12.2 | $E[U_i(\theta_i, \theta_i)] \geq E[U_i(\hat{\theta}_i, \theta_i)]$ | BIC |
| 式 12.3 | $t_i = \sum_{j \neq i} v_j(a^*(\theta_{-i})) - \sum_{j \neq i} v_j(a^*(\theta))$ | VCG支付 |
| 式 12.4 | $\psi(\theta) = \theta - (1-F(\theta))/f(\theta)$ | 迈尔森虚拟价值 |
Coming in Part V: graduate macro. The models get serious — and so do the policy debates.