115
ways in which these rules may be incorporated into FEW systems models. Here we
briefly summarize a handful of approaches for developing such models. This is not
intended as an in-depth or exhaustive review. For further details see, for example,
Irwin and Wrenn (2014) and Robinson et al. (2007).
Primary data can be used to estimate the key parameters of individual decisions.
Zhang et al. (2016) use survey data and discrete choice modeling to estimate the
parameters of farmers’ decisions to adopt a specific best management practice.
Alternatively, secondary data may be used to estimate a key decision- making
parameter. For example, data on vehicle purchases and registration can be used to
learn about how changes in gas prices affect consumer demand for fuel- efficient
vehicles.
Because of data limitations, it is often not possible to fully estimate the relevant
set of parameters for the population or context of interest. In these cases, descriptive
data analysis may be used to identify patterns among one group that can be applied
to different segments of a population. Alternatively, parameters that have already
been estimated in the literature may be used. For example, economic simulation
models commonly specify plausible demand and supply elasticity parameters based
on the estimates that have been reported in the literature.
Once the decision-making rule has been specified, simulation models can be
used to incorporate these decisions into FEW system models. Different types of
simulation models are possible, but two of the most common approaches are structural economic models and agent-based models. As Irwin and Wrenn (2014) discuss
in more detail, structural economic models are primarily focused on modeling consumption and production decisions and price feedbacks for one or more key sectors
of the economy. While these models focus on economic outcomes, they may also
account for key environmental externalities, such as carbon emissions or natural
land degradation.
These models assume some kind of optimizing behavior (e.g., profit or utility
maximization or cost minimization) subject to resource and other constraints. They
have become increasingly sophisticated by incorporating multiple types of uncertainty and environmental tipping points that can lead to large discontinuous
changes in the coupled human-natural system.
3
However, these models are computationally intensive, making it difficult to incorporate many sources of behavioral
heterogeneity.
Agent-based models offer an alternative that, because they do not rely solely on
optimization, are more flexible and do not require the same approach to “solving”
the model. Instead of assuming market equilibrium in which prices instantaneously
adjust to the cumulative actions of individuals, these models are ad hoc in their
treatment of the price mechanism, requiring them to specify alternative assumptions
about how buyers and sellers interact and how prices emerge from these interactions. While this is often viewed as a limitation of these models, their advantage is
3 See Irwin et al. (2016b) for more discussion of these and other dynamic coupled models of
human-natural systems.
4 Human Behavior and Adaptation
ways in which these rules may be incorporated into FEW systems models. Here we
briefly summarize a handful of approaches for developing such models. This is not
intended as an in-depth or exhaustive review. For further details see, for example,
Irwin and Wrenn (2014) and Robinson et al. (2007).
Primary data can be used to estimate the key parameters of individual decisions.
Zhang et al. (2016) use survey data and discrete choice modeling to estimate the
parameters of farmers’ decisions to adopt a specific best management practice.
Alternatively, secondary data may be used to estimate a key decision- making
parameter. For example, data on vehicle purchases and registration can be used to
learn about how changes in gas prices affect consumer demand for fuel- efficient
vehicles.
Because of data limitations, it is often not possible to fully estimate the relevant
set of parameters for the population or context of interest. In these cases, descriptive
data analysis may be used to identify patterns among one group that can be applied
to different segments of a population. Alternatively, parameters that have already
been estimated in the literature may be used. For example, economic simulation
models commonly specify plausible demand and supply elasticity parameters based
on the estimates that have been reported in the literature.
Once the decision-making rule has been specified, simulation models can be
used to incorporate these decisions into FEW system models. Different types of
simulation models are possible, but two of the most common approaches are structural economic models and agent-based models. As Irwin and Wrenn (2014) discuss
in more detail, structural economic models are primarily focused on modeling consumption and production decisions and price feedbacks for one or more key sectors
of the economy. While these models focus on economic outcomes, they may also
account for key environmental externalities, such as carbon emissions or natural
land degradation.
These models assume some kind of optimizing behavior (e.g., profit or utility
maximization or cost minimization) subject to resource and other constraints. They
have become increasingly sophisticated by incorporating multiple types of uncertainty and environmental tipping points that can lead to large discontinuous
changes in the coupled human-natural system.
3
However, these models are computationally intensive, making it difficult to incorporate many sources of behavioral
heterogeneity.
Agent-based models offer an alternative that, because they do not rely solely on
optimization, are more flexible and do not require the same approach to “solving”
the model. Instead of assuming market equilibrium in which prices instantaneously
adjust to the cumulative actions of individuals, these models are ad hoc in their
treatment of the price mechanism, requiring them to specify alternative assumptions
about how buyers and sellers interact and how prices emerge from these interactions. While this is often viewed as a limitation of these models, their advantage is
3 See Irwin et al. (2016b) for more discussion of these and other dynamic coupled models of
human-natural systems.
4 Human Behavior and Adaptation