'Decision making with naive advice' Schotter (2003)26 Jan 2017
After summarizing a recent paper on chatting in matching mechanisms, I decided to dig deeper into Andrew Schotter’s research. ‘Decision making with naive advice’ (2003 AER) caught my eye. In light of all the behavioral economics books I have been reading, my hunch was that taking advice from friends would cause actions to diverge from the predictions of theory with rational agents (Econs as Richard Thaler calls them). Turns out that advice does the opposite. It helps people to act rationally.
Schotter has made extensive use of intergenerational games. Subject A plays a game, and is then substituted by Subject B. The key is that B is allowed to observe the history of the game, and receive advice from A. This process is repeated to create a generation of players. The number of generations is large – between 60, and 80 – and the experimenters plot the advice, and actions over time. The treatments vary if B sees only the advice, only the history of the game, or both. The paper is a summary of results from different variations of intergenerational games.
Advice in ultimatum games lowers the offers, and improves the probability that receivers accepts low offers, which is an intuitive result. In contrast, the coordination games raised an “advice puzzle” for Schotter, and his co-authors. Coordination rates were 49 percent in the advice only treatment, 29 percent in the history only treatment, and 58 percent when subjects see both. Since there are a large number of generations, why does advice help when the successor has nearly the same information as his predecessor? Secondly, why is advice so effective for coordination? At the time he wrote the article, Schotter did not have an answer, and I will need to read his more recent work to see if he solved the puzzle. Schotter notes that there are some hints towards herding behaviour (Banerjee, 1992; Bikhchandani, Hirshleifer & Welch, 1992), because subjects take advice rather than using their own information of the history of the game.
In a different game, where subjects may only observe a single predecessors action, or advice, the advice treatment is much more likely to cause herding behavior. Humans (in contrast to Econs) prefer follow advice, rather than copy actions. Further experiments revealed that Humans think more carefully about a decision when they giving advice than when they make the decision in isolation. To give advice is to learn twice!