Efficient Public Good Provision Between and Within Groups

1. Problem Statement

The paper addresses the dual challenge of public good provision when agents are organized into groups (e.g., nations in a federation):

1. Inter-group Free-Riding: Groups have an incentive to free-ride on the contributions of other groups.
2. Intra-group Coordination Failure: Individual players within a group lack incentives to cooperate, leading to suboptimal simultaneous contributions.

Existing literature often suggests that without a "grand coalition" or external enforcement mechanisms, efficient provision is impossible. This paper investigates whether position uncertainty (where groups do not know their exact rank in a sequence) combined with observational learning can sustain full cooperation among self-interested players in a finite-horizon game, even without enforcement.

2. Methodology and Model Setup

Game Structure:

• Players: A set of $N$ players divided into $b$ groups of size $n$ (where $N = b \times n$).
• Sequence: Groups play sequentially ($t = 1, \dots, b$).
• Position Uncertainty: Groups are assigned to the sequence randomly. Crucially, a group does not know its exact position $t$.
• Information (Observational Learning):
  • Groups observe a sample $\zeta = (\zeta', \zeta'')$ of their immediate predecessors.
  • $\zeta'$ is the number of groups sampled ($m$).
  • $\zeta''$ is the total number of contributors in that sample.
  • If a group is early in the sequence ($t \le m$), they observe all predecessors. If $t > m$, they observe a fixed sample of size $m$.
• Actions: Within each group, players choose simultaneously and independently to Contribute ($C$) or Defect ($D$).
  • $C$: Cost of 1 unit.
  • $D$: Cost of 0.
• Payoffs: The total contributions form a common fund redistributed to all $N$ players with a rate of return $r$.
  • $u_i(C) = \frac{r}{N}(G_{-i} + 1) - 1$
  • $u_i(D) = \frac{r}{N}(G_{-i})$
  • Assumption: $1 < r < N$ (Contributions are socially desirable but individually costly).

Solution Concept:
The authors analyze the game using Sequential Equilibrium (Kreps & Wilson, 1982), which accounts for players' beliefs about the history of play given their limited observations.

3. Key Contributions

1. Generalization of Position Uncertainty: The paper extends the model of Gallice and Monzón (2019), which applied position uncertainty to individual players, to a setting where groups make simultaneous decisions.
2. Intra-Group Pivotality: It demonstrates that position uncertainty transforms the standard simultaneous public goods game within a group into a threshold public goods game. Because a group's contribution influences the beliefs of all subsequent groups, every individual member becomes pivotal.
3. Enforcement-Free Cooperation: The paper challenges the necessity of a "grand coalition" or external enforcement. It proves that efficient provision is achievable with self-interested players solely through the mechanism of conditional cooperation driven by position uncertainty, provided there are more than two groups.

4. Main Results

The results are divided based on the sample size $m$ (number of predecessor groups observed).

A. Case 1: Large Samples ($m > 1$)

• Pure Strategy Equilibrium: A "grim trigger" style strategy exists where groups contribute as long as no defection is observed in the sample. If a defection is observed, all subsequent groups defect.
• Condition for Existence: Full cooperation is sustainable if the rate of return $r$ is sufficiently high:
  $$r \ge \frac{2N}{N - n(m + 1) + 2}$$
• Insight: The condition is stricter than in the individual-player model (Gallice & Monzón) because the group size $n$ increases. A unilateral defection within a group allows the defector to free-ride on their own group members with certainty, raising the bar for cooperation.

B. Case 2: Single Observation ($m = 1$)

This is the more realistic and challenging case where groups only see their immediate predecessor.

• Contrast with $n=1$: In the individual-player model ($n=1$), a pure strategy equilibrium fails for high $r$ because a single player can unilaterally restore cooperation, removing the credibility of punishment.
• Result for Groups ($n > 1$): A pure strategy equilibrium exists for all $r \ge \frac{2N}{N - 2(n-1)}$.
  • Reasoning: In a group, a single player cannot unilaterally restore a history of full cooperation. Even if one player contributes, if the group as a whole is perceived to have defected (or if the sample shows a defection), the group cannot credibly signal a return to cooperation without unanimous action. This makes the threat of punishment credible.
• Mixed Strategy Equilibria: For sufficiently large $r$, there exist two distinct mixed strategy equilibria where players forgive defections with a probability $\gamma \in (0,1)$.
  • The function $\Delta(\gamma)$ (difference in utility between contributing and defecting after a defection) is negative at $\gamma=0$ and $\gamma=1$ but positive in between.
  • By Rolle's Theorem, there are at least two roots where $\Delta(\gamma) = 0$, corresponding to two stable mixed equilibria.
  • Group Size Effect: As group size $n$ increases, the equilibrium probability of forgiveness $\gamma$ increases. Larger groups require a higher probability of forgiveness to make individuals indifferent between contributing and defecting, because the probability of the entire group contributing after a defection decreases with size.

5. Significance and Implications

• Theoretical: The paper bridges the gap between observational learning literature and public goods games. It shows that group structure acts as a stabilizing force; the inability of a single individual to unilaterally "fix" a broken chain of cooperation actually strengthens the equilibrium by making punishment credible.
• Policy (Climate Change): The findings suggest that building a "grand coalition" of all countries may not be strictly necessary to achieve efficient CO2 reduction. If countries are organized into sequential groups with uncertainty about their position, and if they can observe the contributions of predecessors, self-enforcing cooperation is possible without external enforcement mechanisms.
• Practical Application (Crowdfunding): The model offers a blueprint for crowd-funding public goods. Organizers can split donors into sequential groups, inform them of predecessors' success, and leverage position uncertainty (not revealing if a group is the last one) to induce high contribution levels.

Conclusion

The paper establishes that position uncertainty combined with simultaneous group decision-making creates a unique equilibrium environment where self-interested players fully cooperate. The key mechanism is that every player is pivotal to the group's signal to future groups. This allows for efficient public good provision in finite horizons without the need for a grand coalition or external enforcement, provided the rate of return is sufficiently high and groups are of size $n > 1$.