Center for Behavior, Institutions and the Environment / CBIE Working Papers / Using Agent-Based Models to Compare Behavioral Theories on Experimental Data: Application for Irrigation Games

Using Agent-Based Models to Compare Behavioral Theories on Experimental Data: Application for Irrigation Games

CBIE_WP-2015-003

Abstract

Behavioral experiments have demonstrated that people do cooperate in commons dilemmas. The traditional theory of selfish rational behavior is clearly falsified. However, we lack agreement on alternative formal models to explain the actions seen in the lab and beyond. In this paper, we will use agent-based models to compare alternative behavioral theories on a series of experimental data of irrigation games. The irrigation dilemma captures situations of asymmetric access to common resources while contributions of all participants are needed to maintain the physical infrastructure. The sets of experimental data we use differ in the way participants are able to communicate. In experiments with full information participants in all positions invest similar levels in the public infrastructure. Investment inequality, however, increases for experiments without full information. When participants can only communicate with direct neighbors downstream participants start to invest less in the public infrastructure. This is not surprising given the limited level of water downstream participants receive in the second phase of each round in the irrigation experiment. Extraction inequality follows a similar pattern: experiments in which full communication is allowed have lower levels of extraction inequality compared to experiments in which communication is limited.  In our model analysis we compare various alternative theories, including naive simple ones like selfish rational actors and altruistic actors. We contrast these with various alternative behavioral models for collective action as well as inclusion of other-regarding preferences. The systematic comparison of alternative models on experimental data from 44 groups enables us to test which behavioral theories best explain the observed effects of communication. We do not find that one theory clearly outperform others in explaining the data.

AttachmentSize
PDF icon cbie_wp_2015-003_0.pdf500.43 KB

Published February 20, 2015

Marco A. Janssen, Arizona State University, School of Sustainability

Jacopo A. Baggio, Utah State University