Policy and Complex Systems
submitted to them every year based on their multiple criteria expected value functions ( shown in Table 1 ). The parameters for the seven RPC criteria are calibrated to the observed probability distributions shown in Table 2 . The RPC agents rank the roadway projects and send them to the state agent . Under the reference scenario , the state agent assigns 20 % weight to the regional priority . A project that is ranked number 1 by an RPC is assigned 20 points by the state agent , a project ranked number 2 is assigned 18 points , and so on until a project that is assigned a rank of 10 or lower is given zero point by the state agent . Furthermore , the state agent calculates its expected value for each submitted project according to its default criteria ( 40 % highway system , 20 % cost per vehicle mile , and 20 % to the project momentum ) and ranks the projects by assigning rank no . 1 to the project with the highest expected value , 2 to the next highest , and so forth . The federal agent is not explicitly modeled as an agent in the current ABM . Rather , the federal agent is represented through a parameter that sets the available funding in a given year ( shown as dummy agent in Figure 3 ). The ABM model user can define and vary this parameter . The default value is set to 15 %, which was estimated during the calibration process of the ABM by minimizing the difference between observed and simulated funding allocations . The VTrans agent calculates the total cost of the ranked projects and selects top ranked projects for which funding is available in a given year . The state agent rejects the rest of the projects in a given annual cycle . The rejected projects are sent back to RPCs and local towns for evaluation in the next period . New projects are added ( though a user-defined parameter ) every year that are evaluated along with the projects that were rejected in the previous cycle . Model calibration process also showed that 60 new projects ( consistent with observed data ) are , on average , added every year .
Overall , there are 27 input parameters in the model as shown in Table 3 . The majority of the input parameters in the model have either uniform or triangular probability distributions , which means that the ABM is stochastic at its base and each model run is a unique realization chosen from the random probability distributions . The probability distributions for the RPC parameters shown in Table 3 were estimated from the observed roadway project prioritization data ( e . g ., Table 2 ). Experimental simulations of the intergovernmental decision-making process could be run by varying the probability distributions of these input parameters , as explained in the findings section below .
Table 2 . Comparison of convergent cross mapping coefficients
61