On the modeling and testing of groundwater resource models

Economists have been attempting to take on the optimal management of groundwater for many decades, initially through static models, and since the 1970's through a dynamic framework. Since then, several attempts have been made to test dynamic models through laboratory experiments. Yet formulating and testing these models raises several challenges that we attempt to tackle in this study by testing a very simple dynamic groundwater extraction model in a laboratory experiment. We propose a full characterization of the theoretical solutions, taking into account economic constraints. In the experiment we mimic continuous time by allowing subjects to make their extraction decisions whenever they wish, with an actualization and updating the data (resource and payoffs) every second. The infinite horizon is simulated through the computation of payoffs, as if time were endless. To get around the weaknesses of the widely used Mean Squared Deviation (MSD) statistic and classify individual behavior as myopic, feedback or optimal, we combine the MSD with Ordinary Least Squares (OLS) regressions and time series treatments. Results show that a significant percentage of agents are able to adopt an optimal extraction path, that few agents should be considered truly myopic, and that using the MSD alone to classify agents would be misleading for about half of the study participants.