Mediated Trust

arXiv:2511.02436 · econ.TH · Submitted 2025-11-04 · Read on arXiv

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: "Mediated Trust".

Warren: Here's the gist of the paper. Dynamic correlation is introduced as an incentive instrument to address moral hazard by correlating continuation strategies across private histories, which relaxes binding incentive constraints and strengthens effort incentives.

Ruth: The paper formalizes this mechanism in a setting where a long-lived worker faces short-lived clients, showing that optimal mediation induces a nonstationary correlated information structure that transitions from personalized to standardized communication.

Warren: By using private communication to correlate continuations, the firm relaxes otherwise binding incentive constraints and strengthens effort incentives.

Ruth: The Benchmark Model The benchmark model describes a setting where a long-lived worker faces a sequence of short-lived clients, where in each period the client decides whether to accept or reject the worker, and if accepted, the worker privately chooses either to exert effort or to shirk.

Paper summary: Ruth: We’ve just covered the core idea of "Mediated Trust" and how dynamic correlation acts as an incentive device to manage moral hazard by linking continuation strategies between a long-lived worker and short-lived clients >

Warren: It really boils down to the firm mediating interactions, which means they don't have to tell the worker what to do every single time, but they use this correlation mechanism instead >

Iris: The central claim is that optimal mediation induces a nonstationary correlated information structure that shifts from personalized communication early on to a standardized one later on >

Ruth: This shift from personalized to standardized communication is presented as being consistent with the empirical pattern of organizational socialization, which is how firms actually manage incentives over time >

Warren: The authors argue that by using private communication to correlate continuations across private histories, they can relax the worker's otherwise binding incentive constraints and actually strengthen their effort incentives >

Iris: They also characterized all optimal communication devices under fixed discounting and found that this correlation structure is nonstationary, meaning it changes depending on how much time you look ahead in the process >

Ruth: The paper shows that optimal mediation expands the Pareto frontier, but this expansion comes with a distributional conflict between what benefits the worker versus what benefits the average client >

Warren: And they clarify that this outcome is Pareto-improving only when you consider how patient you are as a worker because their value stays below the first best for all discount factors under mediation >

Iris: Furthermore, they point out that there are still inefficiencies from moral hazard that remain present even with this dynamic correlation structure, meaning it doesn't completely solve the problem of moral hazard >

Ruth: So to summarize, "Mediated Trust" introduces dynamic correlation as a way to link future plans between workers and clients through private communication, leading to a shift in information structure and some distributional trade-offs >

Warren: It’s about using this correlation over continuation strategies rather than just actions, which is what sets it apart from static correlation approaches >

Iris: And the key finding is that the optimal structure is nonstationary, moving from personalized to public communication as the interaction history grows >

Conclusion: Ruth: Thinking about "Mediated Trust" and those authors, Allen Vong et al., it’s really about how firms can manage incentives in complex settings where timing matters >

Warren: It suggests that even with this dynamic correlation structure, you’re still not achieving the perfect outcome you might get in a simpler relationship between a long-lived worker and a long-lived client >

Iris: The main thing is that they model how private communication can be used strategically to correlate plans across time, which is relevant because it mirrors how organizations evolve their communication methods over years >

Ruth: The implication for me is that we need to look at these incentive mechanisms not just in static equilibrium but through the lens of dynamic correlation, especially when you have long-lived players interacting with short-lived ones >

Warren: And from a trading perspective, it highlights that the cost of implementing this dynamic correlation strategy needs to be balanced against those potential benefits, because it's not just about maximizing one payoff >

Iris: So while they show a mechanism for strengthening effort incentives through this correlation structure, they also emphasize the distributional conflict between worker and client payoffs as a necessary trade-off >

National University of Singapore

econ.TH

Submitted: 2025-11-04

Updated: 2026-11-05

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 84/100

The gist: Dynamic correlation is introduced as an incentive instrument to address moral hazard by correlating continuation strategies across private histories…

Key concepts

Dynamic Correlation
This is an incentive instrument where the firm correlates future continuation strategies based on private histories. By linking these strategies, it relaxes strict incentive constraints and makes it easier for workers to exert more effort.
Optimal Mediation
This refers to the best way for a firm to communicate with a long-lived worker facing short-lived clients. The optimal strategy is not just about what the worker does now, but how communication evolves over time, leading to personalized then standardized information sharing.
Nonstationary Correlated Information Structure
This describes how the information available to the firm changes as time progresses. Initially, communication is private and tailored to each worker's history; later, it becomes more public and standardized across all workers.
Moral Hazard
Moral hazard occurs when one party takes risks because another party bears the cost of those risks. In this context, the firm uses dynamic correlation to address moral hazard by aligning continuation strategies with effort incentives.

Terminology

Summary

Dynamic correlation is introduced as an incentive instrument to address moral hazard by correlating continuation strategies across private histories, which relaxes binding incentive constraints and strengthens effort incentives. p. 1

Where the paper says it

Unlike static correlation, which operates over actions, dynamic correlation operates over continuation strategies. I show that, by communicating with workers in a manner hidden from clients, firms correlate continuation strategies of workers and clients across private histories, thereby relaxing workers’ otherwise binding incentive constraints and strengthening effort incentives.

The paper formalizes this mechanism in a setting where a long-lived worker faces short-lived clients, showing that optimal mediation induces a nonstationary correlated information structure that transitions from personalized to standardized communication. p. 0
Where the paper says it

A firm mediates interactions between a long-lived worker and short-lived clients. I show that optimal mediation induces a nonstationary correlated information structure that transitions from private to public communication, consistent with the empirical shift from personalized to standardized communication in organizations.

By using private communication to correlate continuations, the firm relaxes otherwise binding incentive constraints and strengthens effort incentives. p. 0
Where the paper says it

I show that optimal mediation induces a nonstationary correlated information structure that transitions from private to public communication, consistent with the empirical shift from personalized to standardized communication in organizations. By using private communication to correlate continuations, the firm relaxes otherwise binding incentive constraints and strengthens effort incentives.

The Benchmark Model

The benchmark model describes a setting where a long-lived worker faces a sequence of short-lived clients, where in each period the client decides whether to accept or reject the worker, and if accepted, the worker privately chooses either to exert effort or to shirk. The output is publicly observable, and in a one-shot interaction without mediation, all equilibria take a carrot-and-stick form with rejections as punishment. p. 1

Where the paper says it

The value of mediation—and thus of correlation—is inherently dynamic: in a one-shot interaction, the client rejects, anticipating the worker’s strict incentive to shirk upon acceptance. Without mediation, all equilibria maximizing the worker’s value take a carrot-and-stick form: they begin with a carrot phase featuring acceptance and effort, with bad outputs either 1

The worker’s average payoff is characterized by the discount factor δ, and the unique Nash equilibrium outcome features rejection. p. 6
Where the paper says it

Throughout, it is useful to interpret δ as the probability in each period that the worker’s employment continues to the next period. Effort incentives are dynamic: in a one-shot interaction, upon acceptance, the worker strictly prefers to shirk; the unique Nash equilibrium outcome features rejection.

Characterization of Optimal Communication Devices

The paper recursively characterizes all optimal communication devices under fixed discounting, showing that optimal mediation yields a nonstationary correlated information structure: continuations are initially personalized to the worker and eventually become standardized, consistent with the empirical pattern of organizational socialization. Optimal mediation is distinct from public correlation because it operates over continuation strategies rather than actions, and it expands the set of implementable outcomes via two channels. p. 2

Where the paper says it

It is unlike public correlation, which is standard in dynamic games and convexifies feasible payoffs. By conditioning continuations on private messages, the firm reduces surplus-depleting rejections needed to motivate effort and expands the set of implementable outcomes via two channels.

The worker is not punished when allowed to shirk during secret randomizations, and the possibility of unpunished shirking against accepting clients strengthens effort incentives. p. 2
Where the paper says it

First, the worker is not punished when allowed to shirk during secret randomizations. Second, the possibility of unpunished shirking against accepting clients strengthens effort incentives; this strict benefit emerges because mediation bypasses the indifference condition underlying mixed strategies.

Welfare Implications and Pareto Optimality

The paper shows that optimal mediation generates a distributional conflict by enlarging the set of Pareto-efficient outcomes, meaning it is not necessarily Pareto-improving for both the worker and the average client. The outcome under optimal mediation depends on the weight the firm places on the worker’s payoff relative to the average client’s, and it is Pareto-improving if and only if the worker is sufficiently patient. p. 3

Where the paper says it

The outcome under optimal mediation depends on the weight the firm places on the worker’s payoff relative to the average client’s. Thus, optimal mediation need not be Pareto-improving; it is so if and only if the worker is sufficiently patient.

The worker's value is bounded away from the first best for all discount factors under mediation, establishing an anti-folk result.

Dynamic Correlation Structure

The optimal communication device induces a nonstationary correlated information structure: the firm uses private communication in early periods to create asymmetric information about future continuations, interspersed with periods of effectively public communication, and eventually relies only on public communication to induce PPE continuations. This provides a rationale for the empirical pattern that continuations are initially “personalized to the worker,” in the sense that communication is private at some history in these continuations, and eventually become “standardized,” in the sense that communication is effectively public at all histories in these continuations. p. 2

Where the paper says it

My first main result recursively characterizes all optimal communication devices and their dynamics under fixed discounting. Optimal mediation yields a nonstationary correlated information structure: continuations are initially personalized to the worker and eventually become standardized, consistent with the empirical pattern of organizational socialization.

The Role of Discounting and Inefficiencies

The paper establishes a critical role for the discount factor δ, showing that optimal communication is nontrivial if and only if the discount factor is sufficiently high, as indicated by Lemma 2. Furthermore, it demonstrates that despite the power of mediation in exploiting dynamic correlation, mediation does not replicate an unmediated relationship between a long-lived worker and a long-lived client where a folk theorem would apply. p. 3

Where the paper says it

Dynamic correlation, nonetheless, does not fully replicate intertemporal payoff transfers in sustaining incentives. While the latter underpins the folk theorem (Fudenberg and Levine, 1994), I establish an anti-folk result: under mediation, the worker’s equilibrium value is bounded away from the first best for all discount factors.

The paper also shows that nonnegligible inefficiencies from moral hazard remain irrespective of the discount factor, meaning mediation is not as powerful as intertemporal payoff transfers. p. 9
Where the paper says it

The moral hazard cost c is independent of δ, capturing nonnegligible rejections irrespective of discounting, because short-lived clients do not engage in intertemporal transfers of payoffs. 6 Therefore there is no folk theorem: all equilibrium payoff vectors are bounded away from the frontier of the feasible and individually rational payoff set uniformly over all discount factors.

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

  1. Dynamic correlation as an incentive device can be used to address moral hazard by inducing a nonstationary correlated information structure that transitions from private to public communication, consistent with empirical shifts in organizational communication styles.

  2. The model allows for characterizing optimal mediation dynamics under fixed discounting, showing how firms can use private communication to correlate continuations of workers and clients across private histories, thereby relaxing binding incentive constraints and strengthening effort incentives.

  3. AI systems can be designed to utilize dynamic correlation by employing a communication device that sends private messages based on past messages and outputs, which the firm chooses to maximize the worker's value to the firm.

  4. Optimal mediation can generate a distributional conflict between workers and average clients, meaning AI-driven organizational structures can be designed such that some outcomes favor the worker while others favor the average client, depending on how much patience is required from the worker.

  5. The system can learn to identify when dynamic correlation helps explain pervasive organizational reward-punishment dynamics in settings where existing models of organizational relationships do not apply, particularly by exploiting i-deals for high performers and marginalizing low performers.

  6. AI systems can be trained to distinguish between communication types—private versus effectively public—based on firm histories, allowing them to adapt their incentive strategies based on whether the current continuation is personalized or standardized.

Abstract

I study how mediation facilitates trade between a long-lived seller and short-lived buyers. Without mediation, the seller's moral hazard bounds equilibrium payoffs away from his favorable part, and possibly all, of the efficiency frontier of the set of feasible and individually rational payoffs for the seller and the average buyer, even as he becomes arbitrarily patient. In contrast, with mediation, all interior payoffs in this set can be attained in equilibrium as the seller becomes patient. Mediation achieves these gains by privately informing the seller which continuation applies. Realizing these gains in full requires nonrecursive mediation.

Related papers