���������������������������������������������������������������
that the system will behave in a similar way as the model . For the case study , the inputs were sets of policies , such as the amount of residential expansion permitted , and assumptions about rainfall patterns . The model estimates the amount of water exiting the planning region under the circumstances described by the inputs . That is , the model allows the user to ask ‘ what if ?’.
II - Framework Overview
If diagrams , equations , games , computer
code , text descriptions , and other model implementations are simply different ways of representing relevant features and relationships of some target system as described in the model design , in which situations would one or the other be preferred ? This question gets to the heart of assessing the quality of a model . A model effectively represents the target system , presenting accurate information about a specific issue . A good model does more than this ; it is a tool to more deeply understand multiple aspects of the issue , stimulate ideas about how to respond to concerns and potentially assess the impact of those ideas . There are three sets of considerations that affect a model ’ s suitability for a particular research project :
• functionality — ensuring that the model can be used to achieve project goals ;
• accuracy — matching the model features and relationships with the characteristics of the system being modeled ; and
• feasibility — whether the required resources are available .
In practice , it is unlikely that one technique will be clearly best for a particular policy model . Instead , one or more techniques may have the most appropriate assumptions for the target system , while others most easily provide the required functionality to achieve the policy objectives . Choosing a technique therefore involves trading off between these aspects , with different choices appropriate for different projects .
In order to enable generalization from the model to the real world issue , the model must represent the target system sufficiently accurately . There are three ways in which modeling incorporates accuracy considerations : abstraction , selection of appropriate modeling technique , and parameterization and testing . Abstraction is typically explicit , with formal discussions and decisions about the model design to select relevant features and relationships of the target system to be represented . Similarly , parameterization and testing are explicit , with measures of model fit ( to data ) and analysis of test datasets routinely included in the model report .
In contrast , modeling techniques may be chosen for convenience or familiarity and the accuracy implications are potentially implicit . Implementing the model design adopts the perspective of the modeling technique used and entails a set of assumptions , which should preferably match the characteristics of the system being modeled . Policy analysts need to be involved in the choice of technique to understand which aspects of the system are included and excluded by each option .
In contrast to accuracy ’ s focus on the content of the model , functionality concerns how the model is developed and used , which is influenced by the type of research question being pursued . For example , some models are particularly effective for developing a shared understanding of a system between people who have different perspectives because they expose the reasoning used by each person , while others can be used to compare policy options . Such functionality
64