Evaluating the role of behavioral factors and practical constraints in the performance of an agent-based model of farmer decision making
Introduction
Agri-environmental policies are developed to influence environment through farmers' actions. However, such policies are often ineffective and result in unintended consequences (Malawska et al., 2014). Modeling farmers' decision making on land-use can increase the effectiveness of these policies by providing insights into possible outcomes of specific regulations. In particular, models that can grasp the complexity and heterogeneity of farmer decision making are needed. Agent-based modeling has been widely used for that purpose since it is well suited for representing individual human behavior (Filatova et al., 2013).
Farmer decision making agent-based models (ABMs) use various approaches to represent human decision making, including two opposing theories on rationality of decision makers, i.e. perfect rationality and bounded rationality (Simon, 1955). While the decision making representations based on full rationality assume agents maximize utility which is mostly confined to economic value, the representations based on bounded rationality vary with respect to behavioral assumptions (An, 2012, Parker et al., 2003) and might be subjective (Janssen et al., 2006). Therefore, such assumptions can be a source of uncertainty in a model (Hare and Pahl-Wostl, 2001, Holtz and Nebel, 2014), and the specification of farmer decision making might be decisive for model outcomes. Thus, it is important to assess if the assumptions related to agent rationality and specific implementations in case of bounded rationality have a strong impact on model outcomes. The sensitivity of models using alternative decision making specifications should be tested (Filatova et al., 2013); more specifically, a comparative analysis of the model under different specifications of agent rationalities should be performed (Holtz and Nebel, 2014).
Another means of assessing the robustness of the behavioral assumptions representing the bounded rationality theory is a comparison of the ‘traditional’ model, assuming rational agents, with a corresponding model implementing boundedly rational agents (Holtz and Nebel, 2014). However, studies that focus on the comparison between different rationalities/decision making specifications in agri-environmental models are not common. Lindgren and Elmquist (2005) use four classical models of decision-making that represent different levels of knowledge, i.e. rational, bounded rational, incremental and pure chance driven decision maker. Jager et al. (1999) carried out a test of different behavioral rules in the Consumat approach where agents are equipped with different decision strategies. The authors showed that assumptions about which behavioral strategies agents engage in will determine the macro-level outcomes of the models. Holtz and Nebel (2014) compared two models with different farmer rationality, i.e. rational utility maximizer versus satisficing approach based on the bounded rationality theory. All of these studies find that the specification of agent rules of behavior significantly affects model results. Thus, such tests appear to be an important part of model assessment.
Here we present a farmer decision making ABM able to represent both a detailed, empirical data-based profit maximization as well as other decision making strategies and farmer types based on goals. The model was developed in a staged-based procedure from an existing model based on microeconomic optimization. Such model development, i.e. starting with an economic model as a base and introducing gradual changes, enables assessment of each change in the model specification. In particular, developing a model by adding additional factors one at a time allows testing if the added factor changes model outcomes significantly, and thus, if it can be ignored (Edmonds, 2012). Since the model development included the change in the agents' rationality, it was possible to compare two versions of the model: first assuming perfectly rational profit maximizers, and second, assuming boundedly rational agents.
The aim of this study is to ascertain how the modifications introduced in the model including changes in practical/agronomic constraints and in the behavioral assumptions affected the model dynamics and its outcomes, i.e. if they actually improved the model performance. Moreover, the goal was to compare the relative impact of the different types of factors added at subsequent model development stages. This was tested by assessing each development stage in terms of their ability to reproduce real world data on crop composition, fertilizer and pesticide usage in the modeled region, and by performing a sensitivity analysis on the final model version.
In the next section we present the general modeling strategy. This is followed by a description of changes introduced at each stage of the model development and the reasoning behind the modifications. Next, we present the results of the assessment of the model development stages, and discuss them in the context of the model's design, limitations and applicability to policy impact assessment. We conclude with general recommendations for designing fine scale farmer decision making ABMs.
Section snippets
Modeling goals and strategy
One of the strategies for creating decision making models is further development of an existing model. This is in accordance with the TAPAS (Take A Previous model and Add Something) approach suggested by Frenken (2006), who argues that models developed using incremental modeling strategies are faster to build, easier to communicate, and thus, easier for others to understand. We therefore used an existing microeconomic optimization model (Fonnesbech-Wulff et al., 2010) to develop a farmer
Model development stages
The existing microeconomic farm model (Fonnesbech-Wulff et al., 2010), here referred to as a ‘farm optimization model’ (FOM), used as a starting point in the model development includes decisions at a farm level on crop composition, fertilizer and pesticide usage per crop and is parameterized for a 10 × 10 km area of Bjerringbro in Jutland, Denmark in 2005; which is a source of the real world data here referred to as a baseline. Four farm types are distinguished: pig, cattle, arable and other; each
Stage 1 vs. FOM
Both the value of SDCA (Fig. 3) and SM (Fig. 4) for stage 1 were lower than the corresponding values for the FOM. The difference in the fertilizer usage relative to the baseline was much lower (+ 0.52% vs. − 11.36%, Fig. 5), while the difference in the pesticide usage (TFI) was slightly larger (− 31.24% vs. − 27.31%, Fig. 5). Stage 1 model had a lower value of SDCA for four farm type/soil type/farm size categories, the same value for nine categories, and a higher value for five farm categories. The
Farmer decision making ABM vs. other farmer ABMs
One of the aims of the model presented here was to bridge the gap between the two typical farmer decision making categories of ABMs: models with complex economic optimization (e.g. Happe et al., 2011, Lobianco and Esposti, 2010, Schreinemachers and Berger, 2011), and behavioral models dealing with relatively simple decision problems and using simplified economic modeling. Models in the latter category often apply the bounded rationality approach (Parker et al., 2003). However, in these models
Conclusions
Using a spatially-explicit ABM of farmer's decision making informed by real-world information on farm units and practical agronomic constraints it was possible to accurately replicate cropping patterns, fertilizer and pesticide usage. The resulting behavioral representation of decision making is more dynamic than the original purely economic approach – from which the ABM was derived – that was limited by non-mechanistic constraints. This is important from a policy assessment perspective since a
Acknowledgments
This research was supported by the ECOGLOBE project funded by Aarhus University. We are grateful to Helle Ørsted Nielsen for guidance on the bounded rationality theory and comments on earlier paper drafts. We thank Berit Hasler and Anders Fonnesbech-Wulff for providing details of the original microeconomic model. We also thank Tommy Dalgaard, Nicholas Hutchings, Poul Henning Petersen and Annemette Pagter for agricultural consultations.
References (68)
- et al.
Assessing vulnerability of selected farming communities in the Philippines based on a behavioural model of agent's adaptation to global environmental change
Glob. Environ. Chang.
(2008) Modeling human decisions in coupled human and natural systems: review of agent-based models
Ecol. Model.
(2012)- et al.
Empirical models of farmer behaviour using psychological, social and economic variables. Part I: linear modelling
Agric. Syst.
(1998) Agent-based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis
Agric. Econ.
(2001)- et al.
An agent based model to simulate structural and land use changes in agricultural systems of the argentine pampas
Ecol. Model.
(2011) - et al.
Simulating agricultural land rental markets by combining agent-based models with traditional economics concepts: the case of the Argentine Pampas
Environ. Model. Softw.
(2015) - et al.
Simulation of ecological, social and economic factors in agricultural systems
Agric. Syst.
(1995) - et al.
Towards an improved understanding of farmers' behaviour: the integrative agent-centred (IAC) framework
Ecol. Econ.
(2010) - et al.
Spatial agent-based models for socio-ecological systems: challenges and prospects
Environ. Model. Softw.
(2013) - et al.
Integrated modelling of farm adaptation to climate change in East Anglia, UK: scaling and farmer decision making
Agric. Ecosyst. Environ.
(2008)
Pattern-oriented modelling in population ecology
Sci. Total Environ.
Modelling the interactions between regional farming structure, nitrogen losses and environmental regulation
Agric. Syst.
Behaviour in commons dilemmas: Homo economicus and Homo psychologicus in an ecological-economic model
Ecol. Econ.
An exploratory integrated model to assess management of lake eutrophication
Ecol. Model.
Chapter 30 governing social-ecological systems, handbook of computational economics
Landscape changes in agrarian landscapes in the 1990s: the interaction between farmers and the farmed landscape. A case study from Jutland, Denmark
J. Environ. Manag.
The Regional Multi-Agent Simulator (RegMAS): an open-source spatially explicit model to assess the impact of agricultural policies
Comput. Electron. Agric.
FSSIM, a bio-economic farm model for simulating the response of EU farming systems to agricultural and environmental policies
Agric. Syst.
New woodlands in Denmark: the role of private landowners
Urban For. Urban Green.
Why do we need to integrate farmer decision making and wildlife models for policy evaluation?
Land Use Policy
Categorising farming values as economic, conservation and lifestyle
J. Econ. Psychol.
An agent-based model for analyzing land use dynamics in response to farmer behaviour and environmental change in the Pampanga delta (Philippines)
Agric. Ecosyst. Environ.
A Bayesian sensitivity analysis applied to an agent-based model of bird population response to landscape change
Environ. Model. Softw.
Agricultural landscapes as places of production and for living in owner's versus producer's decision making and the implications for planning
Landsc. Urban Plan.
How farming styles influence biodiversity maintenance in Austrian agricultural landscapes
Agric. Ecosyst. Environ.
An agent-based simulation model of human-environment interactions in agricultural systems
Environ. Model. Softw.
Towards a cross-sectoral analysis of land use decision-making in Scotland
Landsc. Urban Plan.
ALMaSS, an agent-based model for animals in temperate European landscapes
Ecol. Model.
Landscape structure and management alter the outcome of a pesticide ERA: evaluating impacts of endocrine disruption using the ALMaSS European Brown Hare model
Sci. Total Environ.
A method to define a typology for agent-based analysis in regional land-use research
Agric. Ecosyst. Environ.
Vis. Stud.
Farmers' landscape decisions: relationships between farmers' values and landscape practices
Sociol. Rural.
Cited by (45)
A multi-method simulation model to investigate the impact of sunflower seed segregation on silos
2024, Simulation Modelling Practice and TheoryAn integrated socio-economic agent-based modeling framework towards assessing farmers’ decision making under water scarcity and varying utility functions
2023, Journal of Environmental ManagementCoupling agent-based modeling with territorial LCA to support agricultural land-use planning
2022, Journal of Cleaner Production