Public Good Provision with a Governor

This paper analyzes a public good game involving N citizens and a Governor who balances material gains from embezzlement against belief-dependent reputational concerns, fully characterizing the resulting equilibria to demonstrate how penalties, audits, and psychological incentives jointly determine the levels of citizen contribution and public good provision.

The Gist

Imagine a neighborhood trying to build a community garden. Everyone needs to pitch in, but there's a twist: there's a Garden Manager (the Governor) who holds the keys to the shed, and a Random Inspector (Nature) who occasionally checks if people are actually doing their part.

This paper is a mathematical story about how this garden gets built (or doesn't) when the Manager is tempted to steal the tools for themselves, but also cares about what the neighbors think of them.

Here is the breakdown of the game using simple analogies:

The Cast of Characters

1. The Neighbors (Citizens): They can either Plant a Seed (contribute money) or Slip Away (shirk/free-ride). They don't know who else is planting seeds.
2. The Random Inspector (Audits): Every round, the Inspector picks a few neighbors at random. If a "Slipper" is caught, they have to plant a seed plus pay a heavy fine.
3. The Garden Manager (Governor): This person collects all the seeds (money). They decide how much to put into the garden (Public Good) and how much to pocket for themselves (Embezzlement).
   • The Catch: The Manager isn't just greedy; they are also image-conscious. If they don't give the neighbors what they expect to see, the Manager feels a sharp pang of guilt or loses their reputation.

The Three-Act Play

Act 1: The Neighbors Decide
Everyone secretly decides: "Do I plant a seed, or do I hope someone else does it?"

• If they plant, they lose a seed but gain a share of the garden later.
• If they slip away, they keep their seed, but risk getting caught by the Inspector and paying a fine.

Act 2: The Inspector Checks
The Inspector picks a few people. If a "Slipper" is caught, they are forced to plant a seed and pay a fine. This money goes into the common pot.

Act 3: The Manager Decides
The Manager sees the total pot of money.

• The Greedy Option: Keep it all.
• The Guilt Option: If the Manager thinks the neighbors expected a big garden, and the Manager gives them a tiny one, the Manager feels terrible (reputational loss). So, the Manager might give back enough to match expectations to avoid the guilt.

The Big Discovery: How the Garden Survives

The authors found that the garden's fate depends on a delicate balance between Fear of Fines and Fear of Guilt.

1. The "No-Go" Zone (Total Failure)

If the fines for slipping away are tiny, and the Manager doesn't care about their reputation, everyone slips away.

• Result: The pot is empty (or just has the forced contributions from the few caught). The Manager keeps everything. The garden is a dirt patch.

2. The "Heroic" Zone (Total Success)

If the fines are high enough, or the Manager is very sensitive to reputation, everyone plants seeds.

• The Magic of Reputation: Here is the coolest part. Even if the fines are zero, the garden can still be perfect!
  • Why? Because every neighbor knows that if they slip away, the total pot might drop just below what the Manager expected.
  • If the pot drops below expectations, the Manager (feeling guilty) decides, "Well, if I can't meet the expectation, I might as well keep everything."
  • So, the neighbors plant seeds not because they fear a fine, but because they are pivotal. They know that their single seed is the difference between the Manager building a garden or stealing everything.

3. The "Maybe" Zone (Mixed Strategies)

Sometimes, the situation is so tricky that no one is sure what to do.

• Some neighbors plant seeds 70% of the time; others do it 30% of the time.
• It's like a coin flip. The system finds a weird balance where the risk of getting caught and the hope of a garden cancel each other out perfectly. This only happens in a very specific "Goldilocks" zone of parameters.

The Key Takeaways (The "So What?")

• Reputation is a Superpower: You don't always need a police force (high fines) to get people to cooperate. If the person in charge cares enough about their image and the community's expectations, that psychological pressure can be just as strong as a fine.
• The "Pivotal" Citizen: In a group, every single person matters. If you think "my one dollar doesn't matter," you are wrong. In this model, your one dollar is the switch that turns the Manager from a thief into a hero.
• Expectations are Self-Fulfilling: If the community expects a great garden, the Manager must deliver to avoid guilt, which encourages everyone to contribute. If the community expects nothing, the Manager steals, and everyone stops contributing.

The Bottom Line

This paper proves that human behavior in public projects isn't just about math and money. It's about psychology. A leader who fears being seen as a thief (guilt aversion) can sometimes do a better job of managing public funds than a leader who is just scared of getting caught by the police.

It's a reminder that in a community, trust and expectations are just as valuable as the money in the bank.

Technical Summary

Here is a detailed technical summary of the paper "Public Good Provision with a Governor" by Anwar, Matros, and SenGupta.

1. Problem Statement

The paper addresses the challenge of public good provision in environments characterized by delegation to a self-interested intermediary (a Governor) and imperfect monitoring. Specifically, it investigates how the interplay between:

1. Citizen contribution decisions (voluntary vs. compelled via audit),
2. Formal enforcement mechanisms (audits and penalties for shirking), and
3. Belief-dependent reputational concerns (the Governor's psychological cost of failing to meet public expectations)

shapes equilibrium outcomes. The core tension lies in the Governor's trade-off between material gains from embezzlement and the psychological/reputational loss (guilt aversion) incurred if the level of public goods provided falls short of what citizens expect.

2. Methodology

The authors employ a three-stage game of imperfect information within the framework of Psychological Game Theory (PGT).

• Players: $N$ identical citizens and one Governor ($G$).
• Stage 1 (Citizens): Each citizen $i$ simultaneously chooses to contribute ($c_i=1$) or shirk ($c_i=0$) to a common fund.
• Stage 2 (Auditing): Nature randomly selects $k$ citizens for an audit.
  • If a shirker is audited, they are compelled to contribute and pay a penalty $z$.
  • If a contributor is audited, they pay the contribution (no penalty).
  • The audit outcome is private to the citizen; the Governor does not observe individual actions or the audit profile.
• Stage 3 (Governor's Allocation): The Governor observes the total realized fund $X$ but not the composition of voluntary vs. compelled contributions. She chooses a provision level $g \in [0, X]$. The remainder $X-g$ is embezzled.
• Utility Functions:
  • Citizens: Linear utility based on monetary payoff: $u_i = -\text{contribution} - \text{penalty} + a \cdot g$, where $a$ is the marginal per capita return.
  • Governor: Utility consists of material payoff ($X-g$) minus a belief-dependent reputational cost.
    $$u_G = (X - g) - \sigma \cdot \mathbb{1}\{g < \tau(p)\}$$
    Here, $\tau(p)$ is the Governor's belief about the expected common fund (based on citizens' contribution probability $p$), and $\sigma$ is the sensitivity to falling short of expectations. The cost is a step-loss function: a fixed penalty $\sigma$ is incurred only if $g < \tau(p)$.

Solution Concept: Symmetric Subgame Perfect Equilibrium (SSPE). The analysis uses backward induction, holding the Governor's beliefs fixed based on the citizens' strategy profile (standard in PGT).

3. Key Contributions

1. Integration of PGT into Public Goods: The paper embeds belief-dependent preferences (guilt aversion/reciprocity) into a standard public goods game with a corruptible intermediary, bridging the gap between enforcement literature and psychological game theory.
2. Complete Equilibrium Characterization: The authors provide a full characterization of the parameter space, identifying conditions for:
   • Pure-strategy equilibria (Universal Free-Riding vs. Full Contribution).
   • Mixed-strategy equilibria.
   • Asymmetric equilibria (shown to be non-robust "knife-edge" cases).
3. Substitution Effect: The model demonstrates that reputational incentives can substitute for formal penalties. High reputation sensitivity ($\sigma$) can sustain efficient provision even when penalties ($z$) are low or zero, provided citizens are pivotal.
4. Pivotality Mechanism: The analysis highlights how the Governor's threshold rule creates a "pivotal" dynamic. In a full-contribution equilibrium, a single citizen's deviation (if not audited) can cause the fund to drop below the Governor's expectation, triggering a collapse of public good provision ($g=0$). This threat sustains cooperation.

4. Key Results

A. Pure-Strategy Equilibria

The existence of pure-strategy equilibria depends on the penalty $z$ and the Governor's reputation sensitivity $\sigma$.

1. Free-Riding Equilibrium ($p=0$):

   • Exists if the penalty is low: $z \leq \frac{N-k}{k}$.
   • Outcome: Citizens shirk. The Governor provides $g=k$ (the compelled amount) if $\sigma \geq k$ (to meet the low expectation of $k$), or $g=0$ if $\sigma < k$.
   • Insight: Low penalties make shirking optimal regardless of the Governor's reputation, as a unilateral deviation does not change the provision level.

2. Full-Contribution Equilibrium ($p=1$):

   • Exists if penalties are high enough OR if the Governor is highly reputation-sensitive.
   • Condition for Reputation-Sensitive Governor ($\sigma \geq N$): $z \geq \frac{N-k}{k}(1 - aN)$.
     • If the public good is efficient ($aN > 1$), the RHS is negative, meaning efficient provision can be sustained even with zero penalties ($z=0$). This occurs because citizens are pivotal; shirking risks triggering zero provision.
   • Condition for Reputation-Insensitive Governor ($\sigma < N$): $z \geq \frac{N-k}{k}$.
     • Here, high penalties are required to deter shirking, but the Governor still embezzles everything ($g=0$) because she does not care about expectations.

3. Coexistence (Region E):

   • When $\sigma \geq N$ and $z < \frac{N-k}{k}$, both Free-Riding and Full-Contribution equilibria coexist. This highlights the role of coordination and expectation management.

B. Mixed-Strategy Equilibria

Mixed-strategy equilibria arise when citizens are indifferent between contributing and shirking.

• Reputation-Insensitive Governor: A continuum of mixed equilibria exists if $z = \frac{N-k}{k}$, but the Governor always provides $g=0$.
• Reputation-Sensitive Governor:
  • The authors define a Net Gain function $f(p)$ representing the difference between the expected marginal benefit of contributing (triggering provision) and the marginal cost.
  • $f(p)$ is continuous but piecewise polynomial, with "kinks" where the expected fund crosses an integer threshold.
  • Existence: A mixed-strategy SSPE exists if $f(0^+) < 0$ and $f(1^-) > 0$. This requires specific ranges of $a$, $z$, and $\sigma$.
  • Multiplicity: Depending on parameters, there can be a unique mixed equilibrium or a continuum of them.

5. Significance and Policy Implications

• Mechanism Design: The paper suggests that in environments with weak formal enforcement (low $z$), strengthening reputational mechanisms (increasing $\sigma$ via transparency or social accountability) can effectively sustain public good provision.
• Pivotality: The model reveals that the threat of a collapse in provision (due to the Governor's threshold rule) is a powerful disciplining device. Citizens contribute not just to increase the fund, but to ensure the fund meets the expectation required to trigger the Governor's provision.
• Corruption Control: It offers a theoretical basis for "grievance redress" and social accountability. Even without electoral threats, the fear of triggering a reputational loss (guilt) can constrain embezzlement.
• Experimental Predictions: The model generates sharp, testable predictions regarding how the probability of contribution varies with audit intensity, penalty severity, and the visibility of the Governor's actions.

In summary, the paper demonstrates that belief-dependent preferences act as a critical complement to formal enforcement. When the Governor cares about meeting expectations, the strategic interaction shifts from a simple cost-benefit analysis to a coordination game where the "pivotal" nature of individual contributions can sustain efficient outcomes even in the absence of harsh penalties.