📈 economics

Information for nothing and authority for free

This paper demonstrates that in a setting where a principal and an agent possess private information about project costs and benefits respectively, and monetary transfers are unavailable, the principal's optimal mechanism either disregards the agent or grants them full decision authority accompanied by free information, thereby eliminating the need for the agent to report.

Original authors: Deniz Kattwinkel, Alexander Winter

Published 2026-08-11

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Deniz Kattwinkel, Alexander Winter

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Art of the Deal Without Money

Imagine you are trying to solve a puzzle where two people need to agree on a decision, but they are playing a game with a very strict rule: no money can change hands. This is the world of "mechanism design," a branch of economics that studies how to set the rules of a game so that people, acting in their own self-interest, end up doing what is best for the group. Usually, if one person wants to convince another to do something, they offer a bribe or a reward. But in many real-life situations—like a corporate boss deciding whether to fund a risky division, or a regulator approving a new drug—bribes are illegal, impossible, or just bad practice.

In these scenarios, the person making the final call (the "Principal") often doesn't know the full picture, and the person with the information (the "Agent") might be a little too eager to say "yes" because they care more about the project's success than the cost. The big question is: How can the Principal get the right decision without cash? The answer lies in two powerful tools: information and delegation. Think of information as a flashlight the Principal can turn on or off to show the Agent what the costs look like, and delegation as handing the Agent the steering wheel. The challenge is figuring out exactly when to shine the light and when to hand over the wheel to get the best result.

The Paper's Big Discovery

In their paper, Deniz Kattwinkel and Alexander Winter tackle this exact problem. They ask: If a boss (the Principal) knows the cost of a project but the employee (the Agent) knows the benefit, and they can't pay each other, how should the boss structure the decision-making process? The authors find that the best strategy isn't a complex contract or a secret handshake. Instead, the optimal solution is a clever mix of managing expectations and giving away power.

Here is the magic trick they discovered: The Principal should sometimes hide the exact cost and sometimes reveal it, depending on how high the cost is.

When costs are high: The Principal simply says, "No, we aren't doing this," without even asking the Agent. This saves time and prevents the eager Agent from trying to convince them otherwise.
When costs are low: The Principal reveals the cost (or at least a range of possible costs) and then says, "You decide." The Agent will then say "Yes" only if the benefit they see is greater than the cost they expect.

The paper proves that this "disclose-and-delegate" strategy is the best way to handle the situation. It turns out that the Principal can manipulate the Agent's expectations just as effectively as if they were paying them. By hiding the cost in certain situations, the Principal can make the Agent think the project is more expensive than it might be, which stops the Agent from approving bad projects. This is what the authors call "expectation management."

The "Privacy" Surprise

One of the most surprising findings in the paper is that the Principal never needs to know the Agent's true feelings to make the best decision. In many other economic problems, the boss has to guess what the employee is thinking or try to trick them into revealing their secrets. Here, the authors show that the Principal can design a system that works perfectly without ever asking the Agent, "What is your benefit?" The Agent simply reacts to the information the Principal chooses to show them. The paper calls this "privacy-preserving" because the Agent's private thoughts remain private, yet the Principal still gets the right outcome.

When Hiding Information is Better

The paper also digs into a specific scenario where the "perfect" solution isn't just about showing the cost or hiding it completely. Sometimes, the best move is to show the Agent that the cost is "somewhere between X and Y" without giving the exact number. The authors call this "interval censorship."

Imagine a game where the Principal says, "The cost is somewhere between $10 and $20." The Agent has to decide based on that range. The paper shows that by carefully choosing these ranges, the Principal can filter out the "bad" decisions more effectively than if they just gave the Agent the exact number. This works because it changes the "price" the Agent feels they are paying. If the Agent thinks the cost might be high, they are less likely to say "yes" to a project that is only okay, but they will still say "yes" to a project that is amazing.

What the Paper Rules Out

The authors are very clear about what doesn't work. They show that you cannot just let the Agent decide everything (full delegation) if the Agent is too eager, because they will approve projects that cost too much. They also show that you cannot just let the Principal decide everything (no delegation) because the Principal doesn't know the benefit. The "sweet spot" is a hybrid: the Principal sets the rules, hides or reveals information strategically, and then lets the Agent make the final call only when it's safe to do so.

Furthermore, the paper proves that the Principal doesn't need to learn the Agent's type (their specific benefit) to make this work. Some might think the Principal needs to interrogate the Agent to get the truth, but the math shows that a simple rule based on cost intervals is enough. The paper also rules out the idea that the Principal needs to learn the Agent's type before designing the mechanism; the mechanism must be designed before anyone knows the costs or benefits, and it must work for all possibilities.

How Sure Are They?

The authors don't just guess; they use rigorous mathematics to prove these results. They treat the problem like a puzzle with strict rules and show that their solution is the only one that fits. They use a method called "duality," which is like solving a maze by looking at it from the exit instead of the entrance, to prove that their "disclose-and-delegate" strategy is the absolute best possible outcome. They also check their math against different types of probability distributions (like uniform or exponential) to make sure their findings hold up in various real-world scenarios. While they don't run computer simulations to test every single possibility, their mathematical proofs are considered "solved" within the framework of their model.

The Takeaway

In the end, Kattwinkel and Winter show that in a world without money, information is the currency. By carefully controlling what information is revealed and when, a decision-maker can guide an eager partner to make the right choice without ever handing over a dollar. It's a bit like a parent telling a child, "You can have dessert, but only if you finish your vegetables," but with a twist: the parent might sometimes hide the fact that the vegetables are actually very hard to eat, just to make sure the child doesn't get too excited about the dessert. The result is a system where everyone gets what they need, and the Principal gets the best possible outcome, all while keeping the Agent's secrets safe.

Technical Summary: Information for Nothing and Authority for Free

1. Problem Statement
The paper addresses a principal-agent problem characterized by two-sided private information and the absence of monetary transfers. A principal (e.g., corporate headquarters, a regulator) must decide whether to implement a project. The principal privately observes the cost $s \in [0, 1]$ , while an agent (e.g., a division manager, an applicant) privately observes the benefit $t \in [0, 1]$ . The agent's payoff is $t - \alpha s$ and the principal's payoff is $t - s$ , where $\alpha \in (0, 1)$ . The parameter $\alpha$ implies that the agent is biased toward implementation relative to the principal (i.e., the agent internalizes the cost less than the principal does).

The principal possesses full commitment power and designs a mechanism before learning her own cost $s$ . The mechanism maps the agent's report to a probability of implementation. Crucially, because transfers are unavailable, the principal cannot use payments to align incentives. Instead, the mechanism must rely on information disclosure and the allocation of decision rights to influence the agent's behavior. The objective is to maximize the expected social surplus $E[x(s, t)(t - s)]$ subject to the agent's incentive compatibility constraints.

2. Methodology and Analytical Framework
The authors employ a two-pronged analytical approach, first establishing a benchmark under ex-post incentive compatibility and then solving the general Bayesian incentive compatible case.

Ex-Post Benchmark: The authors first analyze mechanisms that remain incentive compatible even if the agent learns the cost $s$ before reporting. They characterize these mechanisms as "disclosure-delegation" schemes where the principal reveals $s$ and offers a menu of implementation probabilities.
Myersonian Transposition: For the general Bayesian case, the paper transposes the no-transfer bilateral private information problem into a standard mechanism design problem with transfers and one-sided information (specifically, a Myersonian auction design problem).
- The principal's problem is reformulated in terms of the allocation rule $X(t) = E[x(s, t)|t]$ and a virtual payment rule $S(t) = E[s x(s, t)|t]$ .
- The absence of transfers imposes a new feasibility constraint: the induced payment $S(t)$ must lie within the bounds of the lowest and highest possible expected costs compatible with the allocation $X(t)$ .
- This transforms the problem into maximizing a linear functional of $X$ subject to the standard envelope condition (incentive compatibility) and the new interim feasibility constraints.
Duality and Convex Analysis: The authors utilize convex analysis to solve the relaxed version of the problem (ignoring the upper feasibility bound). They establish a strong duality result using the Fenchel-Rockafellar duality theorem. The dual problem is identified as a Bayesian persuasion problem where the "sender" (principal) chooses a distribution of posteriors to maximize a payoff function, subject to the constraint that the induced distribution is a mean-preserving contraction of the agent's preferred allocation rule.
Interval Mechanisms: A key structural result is that any mechanism (incentive compatible or not) is interim-equivalent to an "interval mechanism," where the project is implemented if and only if the cost $s$ falls within a benefit-dependent interval $[a(t), b(t)]$ .

3. Key Results

Optimal Ex-Post Mechanism: Under regularity conditions (specifically, if the function $V(\alpha s)$ , representing the principal's interim expected payoff from full-information delegation, is single-crossing), the optimal ex-post incentive compatible mechanism is a capped delegation mechanism with full disclosure. The principal reveals the cost $s$ and delegates the decision to the agent if and only if $s$ is below a certain cap $c$ . If $s > c$ , the principal unilaterally rejects the project.
Optimal Bayesian Mechanism: When the single-crossing condition fails, the optimal mechanism may involve selective disclosure (interval censorship). The principal may withhold specific cost information, revealing only that $s$ lies within an interval. This allows the principal to manipulate the agent's expected cost (acting as a "price") to screen out low-benefit types more effectively than full disclosure would.
Privacy Preservation: Remarkably, the optimal mechanism never requires the principal to learn the agent's type $t$ . The mechanism is "privacy preserving" because the agent's report is not needed to determine the outcome; the outcome depends solely on the cost realization and the pre-committed information structure.
Connection to Persuasion: The solution to the principal's problem is linked to a Bayesian persuasion problem. The optimal allocation rule $X$ is constructed from the solution to a persuasion problem where the "prior" is the agent's preferred mechanism and the sender's preferences are defined by the non-increasing envelope of the principal's hypothetical payoff.
Conditions for Full Transparency: Full transparency (no censorship) is optimal if a strengthening of Myerson regularity holds (specifically, if the generalized virtual value function satisfies certain monotonicity properties). In the absence of these conditions, selective disclosure is optimal.

4. Significance and Contributions
The paper claims to provide a tractable framework for mechanism design in settings with two-sided private information and no transfers, a setting previously considered difficult due to the lack of standard transfer instruments.

Novel Mechanism Structure: The authors demonstrate that "expectation management" via information disclosure can substitute for monetary transfers. By controlling the agent's expectations of the cost, the principal can effectively manipulate the agent's incentives in a manner analogous to a transfer.
Unification of Fields: The work bridges the literature on delegation, informed principal problems, and Bayesian persuasion. It shows that the optimal mechanism in this no-transfer setting can be interpreted as a cost-dependent stochastic posted price in an equivalent auction design problem.
Structural Simplicity: Despite the complexity of two-sided information, the paper proves that the optimal mechanism is simple: it is always an interval mechanism. Furthermore, it shows that the principal does not need to elicit the agent's private information to implement the optimal outcome.
Robustness: The results hold under general distributional assumptions, with specific conditions (like $\lambda$ -regularity and $\beta$ -tilt monotonicity) provided to determine when full transparency is optimal versus when selective censorship is required.

The paper concludes that in environments where side payments are impossible and agents are biased toward implementation, the principal can achieve optimal outcomes by strategically designing information structures and delegation rules, often without ever needing to know the agent's private valuation.