Trusting Your Sources∗

The paper reports on experiments designed to determine the effect of the source of information about the history of play on cooperation in a simple trust game. In the game, the investor can invest an endowment; if she invests, the allocator can keep or split the returns to investment. In the equilibrium, the investor does not invest, foreseeing that the allocator would keep all the returns of investment. In the control treatment, the investor has to take her decision without knowing the allocator’s history. In contrast, in each of the four treatments, the investor receives information about the allocator’s previous action. The first treatment replicates previous findings: When a reliable authority informs the investor about the allocator’s last choice, allocators ’ trustworthiness is enforced and investors ’ trust is built. In the second treatment, the allocator sends a message about his past play to the investor. This treatment is ineffective in promoting cooperation. Allocators are selfish and deceive investors about their opportunistic actions; investors seem to understand that messages are not truthful, become suspicious and cease to invest. In the third treatment, the investor sends a message about the allocator’s last response to the allocator’s next investor. This treatment is effective in improving efficiency. Investor messages are truthful. Allocators seem to believe the reports. Trust is built and trustworthiness is enforced. In the fourth treatment, a third party — another participant who plays the role of an observer — forwards information about the allocator’s history to the investor. Surprisingly, this treatment is effective in improving cooperation: Observers tend to reveal truthfully allocator behavior. Both investors and allocators seem to anticipate honest reporting. Hence, both cooperate. Keywords: trust game, cheap talk JEL: C92, D62, H42 ∗I thank Fabrizio Adriani for his valuable comments. I am grateful to the British Academy for funding the experiment. I have benefited from suggestions from participants at seminars (Max Planck Institute of Economics, University of Lon-don, and University of Bath) and conferences (Workshop on Behavioral Game Theory at Stony Brook and Gesellschaft fü