Simulating Heterogeneous Farmer Behaviors
the information treatment 2 , we give participants information on the technology adoption rate and average production in their group in the last decision . This is similar to a policy that provides information on what others in the neighborhood are doing and has a self-evolving nature . Since each decision is independent and each participant has a unique dominant Nash strategy , theoretically the information treatments should not change participants ’ decisions . However , as noted before , human decisions often demonstrate bounded rationality and may follow simple heuristics or ad hoc rules .
Experiment Procedure
The economic experiment consists of 12 sessions conducted in late 2016 , involving 192 participants recruited at a large public university in the northeastern United States .
Agent-based Model Setup
In this part , we discuss the ABM setup and initialization . We design the ABM to capture key elements of the economic experiment and an actual watershed while avoiding including unnecessary assumptions and processes . We first set the ABM to a spatially explicit context based on the Murderkill watershed located in the southeast part of Kent County , Delaware ( Figure 1 ). The Murderkill 1 watershed is chosen mainly because it consists primarily agricultural land use and it is a typical coastal plain . Besides , it has promulgated TMDL regulations and has research efforts on the estuarine portion of the watershed . Moreover , the watershed is comprised of 68,000 acres of land , which is large enough to generate meaningful conclusions , but not too large to create computational obstacles .
Figure 1 . Murderkill River Watershed , Delaware , United States . Source : delawarewatersheds . org .
1
Note that the origin of the name , Murderkill , has a Dutch origin as “ moeder ” means mother and “ kill ” means river or creek in Dutch . Thus , the rough translation of the name is “ Mother River ”, and not a reference to a bloody past .
170