Designing Public Information for Strategic Advice
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: We're looking at the paper: "Designing Public Information for Strategic Advice".
Warren: Here's the gist of the paper. A principal chooses an action using advice from an expert who privately knows local conditions, and this paper investigates how a decision maker should design public information when it also shapes the incentives of a later adviser Model Setup and Information Structure The model involves two players, a principal P (headquarters) and an expert E (the regional manager), where the payoff-relevant state has two independent components The expert privately observes θ ∼ U
0, 1: , the quality of local implementation A second state s (future demand) is unknown to both players and has a finite second moment The payoffs are defined as uP (a, θ, s) = − (a − θ − wP s) and uE(a, θ, s) = − (a − θ − wEs) The principal commits to a public study π(z s), and the report z is seen by both players before the expert speaks The timing involves Nature drawing (θ, s), the principal committing to the study, observing z, and then the expert sending a costless message m Communication Protocols and Value of Advice The value of advice depends on how the communication is organized through three protocols Under optimal mediation, the value of advice is convex in conflict, making full revelation optimal Under commitment to an action rule, the principal may prefer no information, while informative optimal studies include a report certifying high conflict Under unmediated cheap talk, changes in the number of equilibrium messages create upward kinks in the value of advice The optimal study can place the expert in two communication regimes far apart, skipping every regime in between, strictly outperforming both full revelation and no information Optimal Study Design and Key Results The principal's problem is to choose a distribution of posterior means µz with mean µ0 that maximizes E
W(µz): The optimal study can place the expert in two communication regimes far apart, skipping every regime in between, while beating both full revelation and no information Blurring is common, occurring for about four priors in ten, while skipping is rarer The gains are moderate and largest near babbling Implications for Policy and Governance The model has a direct political reading where the principal is a government setting the budget, and the expert is a spending agency The results give five implications for fiscal governance For example, publishing the official forecast in broad categories can help budgeting A treasury that keeps its own detailed forecast can let the fiscal council publish coarse categories at no cost to its own budget decisions Commitment and information design are substitutes, as a cap on discretion handles the bias directly The protocol decides whether and how to inform depends on how the subsequent conversation is organized The paper demonstrates that the protocol determines the optimal study, revealing that under commitment, a cap handles small conflicts well and the distinctive report certifies high conflict The most valuable news under cheap talk is news that removes the conflict, because it restores full communication Data governance shows that keeping raw data allows for selective public disclosure without sacrificing forecasting precision The protocol dictates the design, revealing that reveal under mediation, blur under cheap talk, and stay quiet unless conflict is large under commitment are the three main options The implications for fiscal councils and spending agencies show how precisely the official forecast should be published The protocol decides whether and how to inform depends on how the subsequent conversation is organized Page 50 --- References Agastya, Murali, Parimal Kanti Bag, and Indranil Chakraborty.
Warren: “Communication and Authority with a Partially Informed Expert.” RAND Journal of Economics 45(1): 176–197 Alonso, Ricardo, and Niko Matouschek.
Paper summary: Warren: So wrapping up this paper "Designing Public Information for Strategic Advice," it boils down to three main options based on that protocol we talked about. Under optimal mediation, you reveal everything because the value of advice is convex in conflict, so full revelation works best.
Iris: Then under commitment to an action rule, you might prefer no information at all because a cap on discretion handles small conflicts well and the value there is concave.
Ruth: And finally, under unmediated cheap talk with no study, changes in the number of equilibrium messages create upward kinks in the value of advice. This means that even just tweaking how many messages are sent can affect how valuable the expert's advice is when there's uncertainty about what they know and want to say.
Warren: The most valuable news under cheap talk, it seems, is actually news that removes the conflict because it restores full communication. That’s a key finding on where to focus your information design efforts.
Iris: And for governance, the results show that keeping raw data allows for selective public disclosure without sacrificing forecasting precision because you can choose how much detail to share based on the protocol you've set up.
Ruth: It really shows that the protocol dictates whether you reveal under mediation, blur under cheap talk, and stay quiet unless conflict is large under commitment. The paper demonstrates that the structure of communication is what determines the optimal public study design.
Conclusion: Ruth: So we've looked at how the structure of communication changes what advice is actually worth under different scenarios in this paper "Designing Public Information for Strategic Advice."
Warren: Yeah, I mean, the authors are looking at this principal and expert setup where local conditions matter, and they're trying to figure out how much information the principal should give when that information also shapes how much the expert will actually do.
Iris: Exactly. The core idea is that if you change *how* you deliver the information—whether it's a full report, or just a summary—it totally changes the value of that advice depending on what kind of communication environment we're in, like cheap talk versus when commitments are involved.
Ruth: It really boils down to the protocol itself determining the study design. If you're dealing with mediation, full disclosure is usually best because the conflict between what you know and what they do matters a lot.
Warren: But if it's just cheap talk, the paper suggests that revealing too much detail actually hurts things; blurring that information can be better because it restores some kind of communication flow.
Iris: And under commitment, where there's a cap on how much the expert can deviate from what they say, keeping things very tight and not giving too much detail seems to work pretty well for handling those small conflicts.
Ruth: So the big implication for me is that when we think about setting budgets or forecasting things out loud, the way we publish that information—whether it's coarse categories or a very detailed forecast—has a direct effect on how much the decision maker actually trusts and uses that advice.
Warren: It means if you want good outcomes, you need to design your public information structure around those incentive rules, not just throw out the most data possible.
Iris: And that leads us to thinking about what happens when we try to actually implement this in a real government setting, because the results suggest a pretty specific way things should be published.
Vasudha Jain, Jaideep Roy
Department of Economic Sciences, Indian Institute of Technology Kanpur · Department of Economics, University of Bath
econ.TH
Submitted: 2026-11-05
Updated: 2026-11-05
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 83/100
The gist: A principal chooses an action using advice from an expert who privately knows local conditions, and this paper investigates how a decision maker should design public information when it also shapes…
Key concepts
- Principal (P)
- The principal is the decision-maker, like a government setting a budget. They choose an action based on advice from an expert who has private local knowledge and must consider how public information affects that expert's subsequent behavior.
- Expert (E)
- The expert is the regional manager who provides advice. They privately know some local conditions ($ heta$) and have incentives shaped by the public information provided by the principal, which can change based on how communication is structured.
- Communication Protocols
- These are different ways information can be organized: optimal mediation (where value is convex in conflict), commitment to an action rule (where no info might be best), and unmediated cheap talk. The choice of protocol dictates the best way for the principal to design public information.
- Optimal Study Design
- This refers to choosing the distribution of posterior means ($μ_z$) that maximizes the value of advice. The optimal study can place the expert in two distant communication regimes, skipping intermediate ones, which outperforms both full revelation and no information.
Terminology
Summary
A principal chooses an action using advice from an expert who privately knows local conditions, and this paper investigates how a decision maker should design public information when it also shapes the incentives of a later adviser
Model Setup and Information Structure
The model involves two players, a principal P (headquarters) and an expert E (the regional manager), where the payoff-relevant state has two independent components The expert privately observes θ ∼ U[0, 1], the quality of local implementation A second state s (future demand) is unknown to both players and has a finite second moment The payoffs are defined as uP (a, θ, s) = − (a − θ − wP s) and uE(a, θ, s) = − (a − θ − wEs) The principal commits to a public study π(z s), and the report z is seen by both players before the expert speaks The timing involves Nature drawing (θ, s), the principal committing to the study, observing z, and then the expert sending a costless message m
Communication Protocols and Value of Advice
The value of advice depends on how the communication is organized through three protocols Under optimal mediation, the value of advice is convex in conflict, making full revelation optimal Under commitment to an action rule, the principal may prefer no information, while informative optimal studies include a report certifying high conflict Under unmediated cheap talk, changes in the number of equilibrium messages create upward kinks in the value of advice The optimal study can place the expert in two communication regimes far apart, skipping every regime in between, strictly outperforming both full revelation and no information
Optimal Study Design and Key Results
The principal's problem is to choose a distribution of posterior means µz with mean µ0 that maximizes E[W(µz)] The optimal study can place the expert in two communication regimes far apart, skipping every regime in between, while beating both full revelation and no information Blurring is common, occurring for about four priors in ten, while skipping is rarer The gains are moderate and largest near babbling
Implications for Policy and Governance
The model has a direct political reading where the principal is a government setting the budget, and the expert is a spending agency The results give five implications for fiscal governance For example, publishing the official forecast in broad categories can help budgeting A treasury that keeps its own detailed forecast can let the fiscal council publish coarse categories at no cost to its own budget decisions Commitment and information design are substitutes, as a cap on discretion handles the bias directly The protocol decides whether and how to inform depends on how the subsequent conversation is organized
Conclusion
The paper demonstrates that the protocol determines the optimal study, revealing that under commitment, a cap handles small conflicts well and the distinctive report certifies high conflict The most valuable news under cheap talk is news that removes the conflict, because it restores full communication Data governance shows that keeping raw data allows for selective public disclosure without sacrificing forecasting precision The protocol dictates the design, revealing that reveal under mediation, blur under cheap talk, and stay quiet unless conflict is large under commitment are the three main options The implications for fiscal councils and spending agencies show how precisely the official forecast should be published The protocol decides whether and how to inform depends on how the subsequent conversation is organized
--- Page 50 ---
References
Agastya, Murali, Parimal Kanti Bag, and Indranil Chakraborty. 2014. “Communication and Authority with a Partially Informed Expert. p. 2 A large literature improves communication by limiting what
the expert knows about the state she reports. Fischer and Stocken (2001) show that communication
can be best when the sender is imperfectly informed.Where the paper says it
Alonso, Ricardo, and Niko Matouschek. 2008. “Optimal Delegation. p. 3 Goltsman et al. (2009) compare mediation,
arbitration and negotiation at a known bias; the arbitration value coincides with optimal delegation
(Holmström, 1977, 1984; Melumad and Shibano, 1991; Alonso and Matouschek, 2008).Where the paper says it
Ambrus, Attila, and Luke Zhao. 2026. “Optimal Information Structures for Strategic Communication. p. 0 Under unmediated cheap talk, changes in the number
of equilibrium messages create upward kinks in the value of advice. Optimal studies avoid
interior communication thresholds and can induce two distant communication regimes while
skipping every regime in between, strictly outperforming both full revelation and no information. Email: vasudhaj@iitk. ‡Department of Economics, University of Bath.Where the paper says it
Where the paper says it
Antić, Nemanja, and Nicola Persico. 2020. “Cheap Talk with Endogenous Conflict of Interest. p. 0 The study and the manager’s advice concern different uncertainties, yet the study
changes the conversation. The reason is a conflict of interest that depends on demand.Where the paper says it
Arieli, Itai, Ivan Geffner, and Moshe Tennenholtz. 2025. “Optimal Information Design in SenderReceiver Cheap Talk Interactions. p. 0 Under commitment
to an action rule, the principal may prefer no information, while informative optimal studies
include a report certifying high conflict. Under unmediated cheap talk, changes in the number
of equilibrium messages create upward kinks in the value of advice.Where the paper says it
Beetsma, Roel, and Xavier Debrun. 2016. “Fiscal Councils: Rationale and Effectiveness. p. 3 We take
these values as inputs and ask how each protocol shapes the optimal public information about the
bias. The political application builds on the informational view of legislative committees (Gilligan
and Krehbiel, 1987), on budget-maximizing bureaucracies (Niskanen, 1971), and on the role of
independent fiscal councils (Beetsma and Debrun, 2016).Where the paper says it
Bergemann, Dirk, and Stephen Morris. 2019. “Information Design: A Unified Perspective. p. 0 A study can therefore help headquarters
forecast demand while making it harder to learn about implementation. This paper asks how a
decision maker should design public information when it also shapes the incentives of a later adviser.Where the paper says it
Chen, Ying. 2012. “Value of Public Information in Sender–Receiver Games. p. 3 Which public beliefs should the receiver
induce when they also set the value of a
third party’s advice? Controlling the sender’s
knowledge
How much should the sender know
about the state she reports?Where the paper says it
Chen, Ying, Navin Kartik, and Joel Sobel. p. 1 The principal should reveal nothing unless the conflict may be large, and then
certify when it is (Section 7). • Under unmediated cheap talk (Crawford and Sobel, 1982), which is our main case, the value
of advice equals the mediation value minus one “scallop” for each communication regime
(Lemma 3. Antić and Persico (2020) show how some conclusions about endogenous
conflict can be reached without selecting an equilibrium; doing so here is left open. All equilibrium
statements concern monotone pure-strategy partition equilibria.Where the paper says it
Where the paper says it
Crawford, Vincent P., and Joel Sobel. 1982. “Strategic Information Transmission. p. 0 Designing Public Information for Strategic Advice∗
Vasudha Jain† Jaideep Roy‡
October 2026
Abstract
A principal chooses an action using advice from an expert who privately knows local conditions. Before receiving advice, she can commission a public study of an independent state, such as
future demand, that changes the expected conflict between their preferred actions.Where the paper says it
Deimen, Inga, and Dezső Szalay. 2019. “Delegated Expertise, Authority, and Communication. p. 0 Before receiving advice, she can commission a public study of an independent state, such as
future demand, that changes the expected conflict between their preferred actions. We show
that the communication protocol determines the optimal study.Where the paper says it
Fischer, Paul E., and Phillip C. Stocken. 2001. “Imperfect Information and Credible Communication. p. 0 The report says nothing about supplier readiness, but it raises what the
manager gains if headquarters believes readiness is high. Her account of local conditions becomes
less credible exactly when the market outlook is good.Where the paper says it
Frones, Luis. 2026. “Speak or Wait: Strategic Delay and the Value of Public Information. p. 0 Designing Public Information for Strategic Advice∗
Vasudha Jain† Jaideep Roy‡
October 2026
Abstract
A principal chooses an action using advice from an expert who privately knows local conditions. Before receiving advice, she can commission a public study of an independent state, such as
future demand, that changes the expected conflict between their preferred actions.Where the paper says it
Gilligan, Thomas W., and Keith Krehbiel. p. 3 We take
these values as inputs and ask how each protocol shapes the optimal public information about the
bias. The political application builds on the informational view of legislative committees (Gilligan
and Krehbiel, 1987), on budget-maximizing bureaucracies (Niskanen, 1971), and on the role of
independent fiscal councils (Beetsma and Debrun, 2016). We take
these values as inputs and ask how each protocol shapes the optimal public information about the
bias. The political application builds on the informational view of legislative committees (Gilligan
and Krehbiel, 1987), on budget-maximizing bureaucracies (Niskanen, 1971), and on the role of
independent fiscal councils (Beetsma and Debrun, 2016).Where the paper says it
Where the paper says it
Goltsman, Maria, Johannes Hörner, Gregory Pavlov, and Francesco Squintani. 2009. “Mediation, Arbitration and Negotiation. p. 3 Goltsman et al. (2009) compare mediation,
arbitration and negotiation at a known bias; the arbitration value coincides with optimal delegation
(Holmström, 1977, 1984; Melumad and Shibano, 1991; Alonso and Matouschek, 2008).Where the paper says it
Holmström, Bengt. 1977. “On Incentives and Control in Organizations. p. 0 A study can therefore help headquarters
forecast demand while making it harder to learn about implementation. This paper asks how a
decision maker should design public information when it also shapes the incentives of a later adviser.Where the paper says it
Holmström, Bengt. 1984. “On the Theory of Delegation. p. 3 Goltsman et al. (2009) compare mediation,
arbitration and negotiation at a known bias; the arbitration value coincides with optimal delegation
(Holmström, 1977, 1984; Melumad and Shibano, 1991; Alonso and Matouschek, 2008). Table 1: Where the paper fits
Literature Existing question Our question
Bayesian persuasion Which beliefs should a designer induce
in a receiver who then acts? Which public beliefs should the receiver
induce when they also set the value of a
third party’s advice?Where the paper says it
Where the paper says it
Ivanov, Maxim. 2010. “Informational Control and Organizational Design. p. 3 We take
these values as inputs and ask how each protocol shapes the optimal public information about the
bias. The political application builds on the informational view of legislative committees (Gilligan
and Krehbiel, 1987), on budget-maximizing bureaucracies (Niskanen, 1971), and on the role of
independent fiscal councils (Beetsma and Debrun, 2016).Where the paper says it
Ivanov, Maxim, and Alex Sam. 2022. “Cheap Talk with Private Signal Structures. p. 2 Ivanov (2010) and Ambrus and Zhao (2026)
let the receiver garble the sender’s signal; Ambrus and Zhao (2026) show that nonmonotone garbling
can help even in the uniform–quadratic model. Lyu and Suen (2026) let an uninformed sender
commit to an experiment whose outcome she then reports by cheap talk, and Deimen and Szalay
(2019) let the expert choose what to learn.Where the paper says it
Jain, Vasudha. 2018. “Bayesian Persuasion with Cheap Talk. p. 3 Table 1: Where the paper fits
Literature Existing question Our question
Bayesian persuasion Which beliefs should a designer induce
in a receiver who then acts? Which public beliefs should the receiver
induce when they also set the value of a
third party’s advice?Where the paper says it
Kamenica, Emir, and Matthew Gentzkow. 2011. “Bayesian Persuasion. p. 3 Table 1: Where the paper fits
Literature Existing question Our question
Bayesian persuasion Which beliefs should a designer induce
in a receiver who then acts? Which public beliefs should the receiver
induce when they also set the value of a
third party’s advice?Where the paper says it
Krähmer, Daniel. 2021. “Information Design and Strategic Communication. p. 2 The design step is a concavification (Kamenica and Gentzkow, 2011); see
Bergemann and Morris (2019). Our contribution is to show what the object being concavified looks
like when it is the value of strategic advice under different protocols, and what this implies for the
design.Where the paper says it
Kreutzkamp, Sophie, and Yichuan Lou. 2025. “Persuasion without Ex-Post Commitment. p. 3 Table 1: Where the paper fits
Literature Existing question Our question
Bayesian persuasion Which beliefs should a designer induce
in a receiver who then acts? Which public beliefs should the receiver
induce when they also set the value of a
third party’s advice?Where the paper says it
Li, Ming. 2004. “To Disclose or Not to Disclose: Cheap Talk with Uncertain Biases. p. 0 Under commitment
to an action rule, the principal may prefer no information, while informative optimal studies
include a report certifying high conflict. Under unmediated cheap talk, changes in the number
of equilibrium messages create upward kinks in the value of advice.Where the paper says it
Li, Ming, and Kristóf Madarász. 2008. “When Mandatory Disclosure Hurts: Expert Advice and Conflicting Interests. p. 0 Optimal studies avoid
interior communication thresholds and can induce two distant communication regimes while
skipping every regime in between, strictly outperforming both full revelation and no information. When the principal retains the raw data, selective public disclosure can improve advice without
sacrificing forecasting precision.Where the paper says it
Lyu, Qianjun, and Wing Suen. 2026. “Information Design in Cheap Talk. p. 0 Under commitment
to an action rule, the principal may prefer no information, while informative optimal studies
include a report certifying high conflict. Under unmediated cheap talk, changes in the number
of equilibrium messages create upward kinks in the value of advice.Where the paper says it
Melumad, Nahum D., and Toshiyuki Shibano. p. 3 Goltsman et al. (2009) compare mediation,
arbitration and negotiation at a known bias; the arbitration value coincides with optimal delegation
(Holmström, 1977, 1984; Melumad and Shibano, 1991; Alonso and Matouschek, 2008). Proofs and further results are in the appendices. 1Other related settings allow an irreversible action to be delayed while public news arrives (Frones, 2026), or let
the designer later communicate the information herself (Jain, 2018; Kreutzkamp and Lou, 2025).Where the paper says it
Where the paper says it
Moreno de Barreda, Inés. 2024. “Cheap Talk with Two-Sided Private Information. p. 0 Under unmediated cheap talk, changes in the number
of equilibrium messages create upward kinks in the value of advice. Optimal studies avoid
interior communication thresholds and can induce two distant communication regimes while
skipping every regime in between, strictly outperforming both full revelation and no information.Where the paper says it
Morgan, John, and Phillip C. Stocken. 2003. “An Analysis of Stock Recommendations.” RAND Journal of Economics 34(1): 183–203
Niskanen, William A. 1971. Bureaucracy and Representative Government. p. 2 The same structure describes fiscal policy-making. A government sets
the budget of a programme run by a spending agency or a regional government that knows local
implementation conditions.Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
System can design public information protocols based on whether it aims to maximize communication value (mediation), minimize loss (commitment), or induce two distant communication regimes (blurring).
-
The system can dynamically select between different information design strategies—full revelation, no information, or two-report studies—depending on the expected conflict level and the cost parameter c.
-
It can implement a strategy of skipping every intermediate communication regime between two optimal reports to strictly outperform both full revelation and silence when using cheap talk.
-
The system can optimize the public study by choosing a distribution of posterior means that maximizes the value function W, which is concave in conflict and convex around specific thresholds (kinks).
-
The system can distinguish between information design as a tool for communication versus information design as an investment decision by comparing the values V and WD.
-
The system can adjust its disclosure strategy based on whether it retains raw data; when raw data are kept, disclosure serves communication alone, whereas when retained, it maximizes the value of advice V alone.
-
The system can determine the optimal information to certify: news that removes conflict under cheap talk and news that certifies large conflict under commitment.
-
For fiscal policy applications, the system can advise on whether a treasury should keep its own detailed internal forecast or publish coarse categories, depending on the institution's budget rule (committed vs. retained data architecture).
Abstract
A principal chooses an action using advice from an expert who privately knows local conditions. Before receiving advice, she can commission a public study of an independent state, such as future demand, that changes the expected conflict between their preferred actions. We show that the communication protocol determines the optimal study. Under optimal mediation, the value of advice is convex in conflict, making full revelation optimal. Under commitment to an action rule, the principal may prefer no information, while informative optimal studies include a report certifying high conflict. Under unmediated cheap talk, changes in the number of equilibrium messages create upward kinks in the value of advice. Optimal studies avoid interior communication thresholds and can induce two distant communication regimes while skipping every regime in between, strictly outperforming both full revelation and no information. When the principal retains the raw data, selective public disclosure can improve advice without sacrificing forecasting precision. We discuss implications for fiscal forecasts and spending agencies.