Chapter 1: Foundations of Mechanism Design#

Imagine you have to design the rules of a game. But the players know things you don’t, and they will act in their own self‑interest. How do you set things up so that, even with hidden information and selfish motives, the outcome is as good as possible? That is the central puzzle of mechanism design. This chapter lays the groundwork: the assumptions that make the puzzle solvable and a stunning shortcut—the revelation principle—that lets us focus on the simplest kind of rules.

The Big Picture#

Mechanism design is the engineering side of economics. Instead of taking a market or voting system as given and predicting what will happen, we start with a desired outcome and ask: what rules, what “game,” will lead self‑interested people to that outcome, even when they hold private information? To answer that, we need a clear language for describing what people want, what they know, and what the rule‑maker can commit to. This chapter builds that language. We will meet the two classic types of hidden information, the simple utility form that makes money a convenient measuring stick, the assumption that people care only about expected values, the power of a designer who can tie her own hands, and finally the revelation principle—a result that lets us study only direct, truth‑telling mechanisms without losing anything.

Two Kinds of Hidden Information: Adverse Selection and Moral Hazard#

When a designer cannot see everything the participants know or do, the situation falls into one of two broad categories. The distinction matters because each calls for a different kind of solution.

Adverse selection is about hidden characteristics—things the participant knows about herself before the interaction begins, but the designer does not. Think of a used‑car market. The seller knows whether the car is a peach or a lemon; the buyer only sees an average car. If the buyer offers a price that reflects average quality, sellers of good cars may walk away, leaving mostly lemons. The market “selects” the wrong types. In mechanism design, a firm that wants to sell a product to consumers with different willingness‑to‑pay faces adverse selection: each consumer knows her own valuation, but the firm does not.

Adverse selection: A situation where one party has private information about her own type (e.g., skill, taste, health) before any contract is signed, and this hidden information can lead to an undesirable mix of participants.

Moral hazard is about hidden actions—things a participant does after the rules are set, which the designer cannot fully observe. Imagine you have full insurance on your phone. You might be less careful about dropping it, because the cost of a cracked screen falls on the insurer, not on you. The action (carefulness) is hidden, and the insurance contract changes it. In mechanism design, a manager who cannot observe an employee’s effort faces moral hazard: the employee may shirk once the wage is fixed.

Moral hazard: A situation where one party can take an unobservable action after an agreement is in place, and the action affects the other party’s payoff.

Many real problems mix both: a job applicant knows her ability (adverse selection) and will later decide how hard to work (moral hazard). For clarity, we usually study them separately. In this chapter and much of the classical theory, we focus on adverse selection—hidden information about types—because it is the natural setting for the revelation principle and for the design of allocation and payment rules. But the same modeling tools can be adapted to moral hazard, and we will note the connections.

📝 Section Recap: Hidden information comes in two flavors: adverse selection (private knowledge of one’s own type before the interaction) and moral hazard (hidden actions after the rules are set). Mechanism design often starts with adverse selection, where the challenge is to elicit truthful type information.

Quasi‑linear Preferences: Money Talks#

To make sharp predictions, we need a simple way to describe what people want. In mechanism design, the workhorse is quasi‑linear utility. The name sounds technical, but the idea is beautifully simple: a person’s overall satisfaction splits into two separate pieces—one that comes from the “stuff” being allocated (a good, a service, a public project) and another that comes from money.

Formally, if an agent gets an outcome (say, whether she receives an item, or the quality of a service) and pays (or receives) a transfer , her utility is

Here is her valuation—how much she subjectively likes outcome , measured in money terms—and is the amount of money she gives up. If the agent receives money, is negative and her utility increases. The key property is that the two parts don’t interact. The total utility is simply the value of the outcome minus the payment. Money enters linearly, with a coefficient of . This means an extra dollar always adds the same amount of happiness, regardless of how much money the agent already has or what outcome she receives.

Why is this so useful? Because it lets us measure everything in a common unit—money—and makes trade‑offs transparent. If an agent values an item at 10$7$3$12$, she would rather not buy. The designer can use transfers (payments, subsidies) to align incentives without worrying about complicated income effects.

Quasi‑linear utility: A utility function of the form , where the benefit from the outcome and the disutility of the money transfer are added together, with money entering linearly. This assumes no wealth effects—an extra dollar is always worth the same to the agent.

Of course, real people do not always think this way. A billionaire might value an extra dollar less than a student does. But quasi‑linearity is a powerful approximation when the stakes of the mechanism are small relative to the participants’ total wealth, or when we are designing rules for firms that can borrow and lend freely. It also makes the mathematics easier to handle, allowing us to focus on the strategic use of private information rather than on the complications of risk and wealth.

📝 Section Recap: Quasi‑linear preferences split utility into a valuation for the outcome and a linear money term. This makes money a universal yardstick for incentives and lets the designer use transfers to guide behavior without worrying about how changes in wealth affect choices.

Risk Neutrality: Agents Care Only About Expected Payoffs#

In many settings, the outcome of a mechanism is not certain when an agent makes her decision. An auction, for instance, might be won or lost. A contract might pay a bonus that depends on a random performance measure. How do we assume agents evaluate such gambles? The standard mechanism design framework assumes risk neutrality.

A risk‑neutral agent cares only about the expected value of her payoff, not about its variability. If she faces a 50–50 chance of getting 100$0$50\tilde{w}\mathbb{E}[\tilde{w}]$.

Risk neutrality: The property that an agent’s utility from an uncertain prospect is equal to the expected value of the monetary outcomes. She does not demand a premium for bearing risk, nor does she pay to avoid it.

This assumption pairs naturally with quasi‑linearity. Together, they mean that an agent’s objective is simply to maximize the expected value of . The designer can then write the agent’s expected utility directly, without worrying about the shape of a utility‑of‑wealth function. It also means that the designer can use lotteries or randomized mechanisms without introducing an extra cost of risk, which simplifies the search for optimal rules.

Why is this a reasonable starting point? Many economic actors—large firms, diversified investors, or governments—can spread risk so well that they behave approximately as if they are risk‑neutral. For individuals, it is a simplification, but one that lets us isolate the effects of private information. Later, we can relax it and study how risk aversion changes the design, but the core insights often survive.

📝 Section Recap: Risk neutrality means agents evaluate uncertain outcomes solely by their expected monetary value. Combined with quasi‑linear utility, it makes the designer’s problem about expected gains and losses, ignoring the extra complications of risk attitudes.

The Designer’s Commitment: Promises That Can’t Be Broken#

A mechanism is a set of rules announced in advance: “If you report this, I will do that; if you report that, I will do this.” But what if, after the agents have revealed their private information, the designer would like to change the rules? If she can, the whole arrangement unravels. So we assume full commitment: the designer can bind herself to follow the announced rules, no matter what.

This is a strong assumption. In the real world, governments pass laws and later amend them; firms promise a return policy and then make it hard to claim. But without commitment, mechanism design becomes a much murkier game of promises and credibility. By assuming full commitment, we can focus on the pure problem of designing the best rules under the constraint that agents will play strategically.

Full commitment: The assumption that the mechanism designer can irrevocably commit to the entire mapping from agents’ messages to outcomes, and cannot alter it after seeing the agents’ reports.

Why does commitment matter so much? Consider a simple example. A seller wants to sell a single item to a buyer whose value is either 10$20$20$20$10$20$ in the first place. Full commitment is what makes a posted price credible and lets the designer extract information through a once‑and‑for‑all mechanism.

In the models we build, the designer moves first, announcing a mechanism. Then agents observe the rules, decide whether to participate, and send messages. Finally, the outcome is implemented exactly as the rules specified. No renegotiation, no surprises. That is the world of full commitment.

📝 Section Recap: Full commitment means the designer can tie her hands to the announced rules. This credibility is essential for agents to trust the mechanism and reveal their private information truthfully.

The Revelation Principle: Why We Can Focus on Direct Mechanisms#

Now we come to the single most powerful idea in mechanism design—the revelation principle. It is a brilliant shortcut. In principle, a designer could dream up any complicated game: agents send messages back and forth, make bids, renegotiate, and so on. But the revelation principle says: for any outcome that can be achieved by some mechanism (no matter how complex), there is a simple, direct mechanism that achieves the same outcome, and in which it is optimal for every agent to report her true type honestly.

Let’s unpack that. A direct mechanism is one where the only thing each agent does is send a single message—a claimed type—to the designer. The designer then uses a pre‑announced rule to map the vector of reported types to an outcome (allocation and transfers). There are no multiple rounds, no side conversations, no strategizing over what message to send when. And the revelation principle tells us that, without loss of generality, we can restrict our search to direct mechanisms where truth‑telling is an equilibrium.

Direct mechanism: A mechanism in which each agent’s strategy space is simply the set of possible types, and the outcome function maps reported types directly to allocations and payments.

Revelation principle: If a certain outcome (who gets what and who pays what) can be achieved by some mechanism, then it can also be achieved by a direct mechanism where truthful reporting is an equilibrium.

Why is this true? Imagine a mediator. She says to the agents: “Just tell me your true type. I will then play the original equilibrium strategy on your behalf.” If telling the truth to the mediator and letting her execute the original strategy gives each agent exactly the same outcome as playing the original game herself, and if lying to the mediator would only lead to an outcome the agent could have achieved in the original game by deviating (which was not profitable in equilibrium), then truth‑telling in this direct mechanism must be optimal. The mediator essentially simulates the original equilibrium.

So the designer does not need to consider arbitrary, multi‑stage mechanisms. She can focus exclusively on direct mechanisms and impose the constraint that truth‑telling is an equilibrium. This reduces the problem to finding an allocation rule and a payment rule (where denotes the vector of agents’ reported types) such that no agent wants to misreport her type, and such that the outcome is feasible and desirable.

A classic illustration: an auction. Instead of an auctioneer calling out rising prices (an English auction) or collecting sealed bids and giving the item to the highest bidder at the second‑highest price (a Vickrey auction), the revelation principle says we can simply ask each bidder “What is your value?” and then apply a direct rule: if the reported values are , allocate to the highest reporter and charge her the second‑highest report. In the original Vickrey auction, bidding one’s true value is a dominant strategy; the direct mechanism that asks for values and mimics the Vickrey outcome inherits that property—truthful reporting is optimal.

The revelation principle is not a description of how real institutions work. It is a theoretical tool. It lets us ignore the vast space of possible game forms and concentrate on the compact set of direct, truthful mechanisms. Almost all of mechanism design theory flows from this insight.

📝 Section Recap: The revelation principle shows that any outcome achievable by any mechanism can be replicated by a direct mechanism where agents simply report their types and find it optimal to be truthful. This allows us to study only direct, truth‑telling mechanisms without loss of generality.

Summary#

We have built the foundations of mechanism design from the ground up. We distinguished the two main information problems—adverse selection (hidden types) and moral hazard (hidden actions)—and noted that the classical theory often begins with adverse selection. We then introduced the key modeling choices that make the analysis easier to work with: quasi‑linear utility, which separates money from everything else; risk neutrality, which focuses on expected values; and full commitment, which gives the designer the power to make credible promises. Finally, we met the revelation principle, a remarkable result that reduces the many possible mechanisms into the simple, direct ones where agents are asked to report their types and have no incentive to lie. With these tools, we are ready to ask precise questions about optimal auctions, efficient public projects, and fair voting rules—questions that the coming chapters will take up.