Policy and Complex Systems
The ABM calculates annualized flow of financial resources from the state government to regional and local town jurisdictions that is contingent upon the project prioritization decision-making undertaken by the intergovernmental network of regional and state agents . Approvals of projects for different regions are characterized on an annualized basis and take into account delays and queues inside the system . Observed project cost distributions for each phase of the project in Vermont were used to calibrate the computer simulation model . The findings from the experimental simulations presented in this study , however , are generated from the model that is calibrated to reflect the institutional design structure of intergovernmental decision-making in the state of Vermont , its RPCs , and local towns . Through a user interface for the ABM , the users ( e . g ., researchers , policymakers , managers , and other stakeholders ) can run reference and alternate scenarios by defining the parameters for different scenario runs as a policy informatics platform .
5 . Findings from Experimental Simulations
All the results reported below are
averages of 1,000 realizations for each model run scenario . Since the ABM is initialized with random parameters as shown in Table 3 , 1,000 realizations for each scenario model run provide a robust set of findings . The model interface also allows decision makers to run up to 5,000 realizations for each scenario but that requires higher computational capacity . There was not much difference between 1,000 and 5,000 realizations of each of the scenarios reported below ; hence , we decided to report 1,000 realization results that could be replicated on any modern computer . Furthermore , there are many possible scenario runs by varying the 27 input parametric values . Here , we focus on reporting the findings from the scenario runs that address the two research questions posed above . In particular , we focus on reporting the funding allocations at RPC and local town level for each scenario run . All scenarios are constrained to a 50-year simulation horizon that could be reduced or expanded if desirable by the stakeholders . Fixed seed runs were imposed to compare the simulation outputs across the scenarios . The reference scenario is run with the default parametric values reported in Table 3 . Figure 5a shows the allocation of funds at the RPC level and Figure 5b shows the allocation of funds at the local town level under the reference scenario . Noticeably , this reference scenario is calibrated against baseline weights between regional and state project prioritization criteria , as reported in Table 1 . Another important component of the model calibration process is driven by the observed probability distributions for the project class on the input parameters reported in Table 3 ( e . g ., probability distributions for the state agent criteria and RPC criteria assessments of the projects submitted by local towns ).
The reference scenario , and all other scenario run outputs discussed below , presents three types of output information : first , as shown in Figure 5a , the time stack charts show simulated allocation of US $ ( Nominal ) for all 11 RPCs in the annualized time-step of the 50-year model simulation horizon . Second , the time plot in Figure 5a ( and all other scenario runs in the Results section ) presents the mean percentage of total projects funded per RPC per year . More successful RPCs in a given year have a 15 % or higher success rate for funding the projects , whereas less successful RPCs have a relatively lower (< 15 %) success rate , i . e .,
63