Cooperation Across Generations: An Experimental Study of Intergenerational Altruism in the Prisoner’s Dilemma
How to cite
Pang, Yu. 2026. "Cooperation Across Generations: An Experimental Study of Intergenerational Altruism in the Prisoner’s Dilemma." AEA RCT Registry. AEARCTR-0018136. https://doi.org/10.1257/rct.18136-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
We design a laboratory experiment to test whether intergenerational altruism can sustain cooperation in a one-shot Prisoner’s Dilemma embedded in an overlapping generations (OLG) structure. In each generation, two players from rival “families” play a Prisoner’s Dilemma; each player’s payoff depends on both their own earnings and those of their successor in the next generation, weighted by an altruism parameter α. A sharp theoretical prediction emerges: mutual cooperation is a subgame perfect equilibrium if and only if α exceeds a threshold determined by the game’s payoffs. We implement a 3×3 factorial design crossing three communication protocols—no communication, partner communication, and parent-child plus partner communication—with three altruism levels (α ∈ {0.4, 1.0, 1.6}) spanning the sub-threshold, boundary, and
supra-threshold regimes. Our design bridges the infinitely repeated games literature (Dal Bó, 2005; Dal Bó and Fréchette, 2018) and the OLG tradition (Samuelson, 1958), offering the first experimental test of cooperation sustained not by the “shadow of the future” but by the “shadow of posterity.”
Intervention
We implement a 3 × 3 factorial design that crosses three communication protocols with three altruism levels. The communication treatments vary the channels through which information and promises can flow: in Treatment 1 (T1), no communication is permitted; in Treatment 2 (T2), partners can exchange messages before playing the Prisoner’s Dilemma; and in Treatment 3 (T3), both parent-child and partner communication are allowed. The altruism treatments vary the weight α that each player places on her successor’s earnings: α = 0.4, which lies below the cooperation threshold; α = 1.0, which sits at the theoretical boundary; and α = 1.6, which lies well above the threshold. This design generates clean comparative statics. If the theory is correct, cooperation should be rare at α = 0.4 regardless of communication, since defection strictly dominates. At α = 1.6, cooperation is an equilibrium outcome, and communication — particularly the combination of parent-child and partner communication in T3 — should help subjects coordinate on it. The boundary case α = 1.0 is especially informative: cooperation is theoretically possible but fragile, making equilibrium selection pivotal and communication potentially decisive.
Experimental design
We design a laboratory experiment to test whether intergenerational altruism can sustain cooperation in a one-shot Prisoner’s Dilemma embedded in an overlapping generations (OLG) structure. In each generation, two players from rival “families” play a Prisoner’s Dilemma; each player’s payoff depends on both their own earnings and those of their successor in the next generation, weighted by an altruism parameter α. A sharp theoretical prediction emerges: mutual cooperation is a subgame perfect equilibrium if and only if α exceeds a threshold determined by the game’s payoffs. We implement a 3×3 factorial design crossing three communication protocols—no communication, partner communication, and parent-child plus partner communication—with three altruism levels (α ∈ {0.4, 1.0, 1.6}) spanning the sub-threshold, boundary, and supra-threshold regimes. Our design bridges the infinitely repeated games literature (Dal Bó, 2005; Dal Bó and Fréchette, 2018) and the OLG tradition (Samuelson, 1958), offering the first experimental test of cooperation sustained not by the “shadow of the future” but by the “shadow of posterity.”
Design details
Cited by
No citation data available for this registration yet.