Essays on experiments and agent-based modeling
Date
2017
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
[CHAPTER 1] Heterogeneous agents and information nudges in non-point source water pollution management: Non-point source (NPS) water pollution from agricultural runoff is a leading cause of impairment for many water bodies in the United States; however, sources of NPS pollution are difficult to identify because of hidden actions and asymmetric information. Theoretical and experimental research has shown that ambient pollution policies can induce groups to reduce pollution to socially efficient levels, but many of these studies have imposed restrictive assumptions about farmer homogeneity and management choices. In reality, agricultural firms differ in both size and location, and farmers make numerous management decisions that can affect runoff and nutrient loss, including decisions about production intensity and pollution abatement technologies. Researchers have shown that introducing either size or location heterogeneity affects the efficiency of ambient pollution policies, but no research has analyzed policy performance while considering several sources of heterogeneity and multiple management decisions. Furthermore, despite multiple examples in using non-pecuniary interventions to promote environmental conservation, little research has examined how to use information nudges, like social comparisons or information about peer actions, to induce better NPS pollution abatement decisions. ☐ In this study, we designed an economic experiment to test the effects of multiple layers of heterogeneity, information nudges, and an extended decision space on the performance of the classic ambient tax/subsidy policy. Experiment participants (n=192) were recruited from a large public university in the U.S. In the experiment, each individual was assigned a firm and asked to make individual decisions that affected the profitability of his/her firm and ambient water pollution of their group. In each round of the experiment, participants selected their production intensity and chose one of two production technologies—a conventional technology or a more expensive technology that generated less pollution. ☐ Eight within-subject treatments were tested, including two policy variations (no policy and a tax/subsidy policy) and four size/location variations (homogeneous, location heterogeneity, size heterogeneity, and both location and size heterogeneity). Three between-subject information treatments were also tested, including a no information control. In information treatment 1, we tested how individual decisions were affected by information nudges about decisions that similar individuals had made in past sessions. In information treatment 2, participants were provided with information about the average production and technology adoption rate in their group during the last round. A unique dominant strategy Nash Equilibrium was calculated for both the adoption decision and production decision based on location and size. ☐ Our results demonstrate that, without information nudges, more firm heterogeneity reduces the effectiveness of ambient tax/subsidy policies and target pollution levels are achieved less frequently. However, the tax/subsidy policy was effective under different heterogeneity scenarios when information is provided about peer and group decisions in past rounds. Furthermore, information treatment 1 and information treatment 2 generate higher policy efficiency than no information treatment. Lastly, participants are able to find and retain their dominant strategy better in the information 1 treatment, suggesting that providing individually targeted information is more effective than providing information about aggregate group-level decisions. Our findings suggest that traditional ambient pollution policies may be less effective when agents are heterogeneous and make multiple decisions that affect pollution, but information nudges can improve policy performance. ----- [CHAPTER 2] Simulating Heterogeneous Farmer Behaviors under Different Policy Schemes: Integrating Economic Experiments and Agent-Based Modeling: In this paper, we develop an agent-based model that scales up results from economic experiments on technology diffusion and abatement of non-point source water pollution under the conditions of an actual watershed. The results from the economic experiments provide the foundation for assumptions used in the agent-based model. Data from geographic information systems and the US Census of Agriculture initialize and parameterize the model. This integrated model enables the exploration of the effects of several policy interventions on technology diffusion and agricultural production and, hence, on agricultural non-point source pollution. Simulation results demonstrate that information ‘nudges’ based on social comparisons increase ambient based policy performance as well as efficiency, especially individual-level tailored information on what others like them have done in past similar situations. ----- [CHAPTER 3] Auctions versus Posted Price in Experiments: Comparisons of Mean and Marginal Effect: Economic experiments have been widely used to elicit individuals’ evaluation for various commodities and non-market goods. Common elicitation methods include auctions and posted price mechanisms. Experimental auctions are theoretically incentive compatible so are assumed to give an unbiased estimate of individuals’ evaluation including willingness to pay (WTP). However, the vast majority of purchasing decisions are not made in auctions but in market settings, such as grocery stores, where consumers make yes/no decisions in response to a set price. In this research, we carefully design an experiment to compare homegrown-value WTP estimates between an auction and a posted price elicitation format. This design enables us to make both within- and between-subjects comparisons of the mean WTP and marginal effect estimates. Results from 115 adult consumers indicate that WTP estimates obtained from an auction are approximately 32% - 39% smaller than WTP estimates obtained from a posted price mechanism. In addition, we compare the statistical significance and conclude that auctions require a smaller sample size than posted price mechanisms in order to detect the same preference change. Nevertheless, the signs of marginal effects for different product characteristics are consistent in both mechanisms.
Description
Keywords
Pure sciences, Social sciences, Agent-based model, Ambient based policy, Auction, Economic experiments, Non-point source pollution, Posted price