Hugh Kelley

Spatially explicit experiments for the exploration of land‐use decision‐making dynamics

We explore the special outcomes of decision‐making through two laboratory‐based experiments, one with a homogenous land suitability surface and another with a heterogeneous suitability surface. Subjects make resource allocation decisions on an abstract landscape and are given a monetary incentive to maximize their revenue during the experiment. These experimental results are compared with simulation output from an agent‐based model run on the same abstract landscape that uses a utility‐maximizing agent. The main findings are: (1) landscapes produced by subjects result in greater patchiness and more edge than the utility‐maximization agent predicts; (2) there is considerable diversity in the decisions subjects make despite the relatively simple decision‐making context; and (3) there is greater deviation of subject revenue from the maximum potential revenue in early rounds of the experiment compared with later rounds, demonstrating the challenge of making optimal decisions with little historical context. The findings demonstrate the value of using non‐maximizing agents in agent‐based models of land‐cover change and the importance of acknowledging actor heterogeneity in land‐change systems.

Measuring the Impact of Behavioral Traders in the Market for Closed-end Country Funds from 2002 to 2009

This work investigates whether traders’ state dependant expectations biases can account for anomalous country fund price movements for a seven year period spanning the 2007-08 banking crises. We provide a multiple agent asset-pricing model that includes both rational traders and traders who display biases in expectations formation following market states with large amounts of fundamental value variance or CNN financial news. Importantly, traders’ biased behavior is based on evidence of state-dependant over- or under-reaction biases observed in asset price forecasting experiments. Closed-form solutions from the multi-agent pricing model predict a multiple driver property of fund prices. Empirical tests for these drivers’ influence in field data finds that a significant amount of out-of-sample country fund discount variance can be explained by dummies representing the occurrence of behavioral bias trigger states.