Information Cascade Experiment: Admin Introduction

Overview: This program sets up a sequential decision-making task in which each person obtains a private signal about which of two events has occured ("Red" or "Blue") and must make a public prediction of the event. Predictions are made in sequence, with each person being able to observe the predictions but not the signals of those who preceded in the sequence. Earnings are higher if one's prediction turns out to be correct. This setup allows "information cascades" in which an individual may make a prediction that contradicts their own private signal, which is optimal if the information content of prior predictions exceeds that of one's own signal.

Parameters: The total number of participants in the session may be divided into groups of a size that you specify. Groups are fixed (same in all rounds), and you specify the number of rounds. You also specify the probabilities of each event, the information structure of the signals, and the payoffs for correct and incorrect predictions. The intstructions for participants are automatically configured to conform to the setup parameters that you use. The best way to become acquainted with the various features of this program is to test it (see instructions on the next page).

Discussion: The discussion can focus on the formation of information cascades and on Bayesian information processing. For hints on classroom discussion of these topics, see Anderson and Holt (1996) "Classroom Games: Information Cascades," Journal of Economic Perspectives, Summer, 187-193. A simple ball-counting heuristic used to teach Bayes' rule is presented in Anderson and Holt (1996) "Classroom Games: Understanding Bayes' Rule," Journal of Economic Perspectives, Spring, 179-187. For related research experiments, see Anderson and Holt (1997) "Information Cascades in the Laboratory," American Economic Review, 847-862.

Session Name: This is one of the Veconlab web-based experiment programs, and to use them you will need your personal "session name." If you do not already have one, it can be obtained quickly online from the Guide to Experimenters link on:

http://veconlab.econ.virginia.edu/admin.html

Your session name should be reused, i.e. you will use this session name for each subsequent experiment that you run, and participants will use your session name when they log in. Finally, the "Admin" site shown above should be added to your browser's list of Favorites; it provides information, hints, and links for setting up any of the 35 other Veconlab programs, which include auctions, bargaining, games, decisions, markets, voting, and an on-line survey instrument.

Vecon Lab -