Refining the Isard Multiregional Input–Output Model: Theory, Operationality and Extensions - Regional Science Matters - page 44

into the transportation cost impact model. The first empirical application of
TransNIEMO, in which geography boundaries are limited to California and
Arizona, is found in Park et al. ( 2011a ).
While NIEMO and its related models are spatially disaggregated only to the state
level, the transportation nodes for freight modes are the major metropolitan areas,
which are the dominant centers of economic activity. Furthermore, in most states
there is one or more major metropolitan area. Hence, the interstate freight flows had
to be expanded to the flows among 114 nodes (specifically “economic centroids”) to
make them comprehensible. The non-metropolitan regions in selected cases also
account for a significant proportion of state gross domestic product and freight O-D
movements. Although local governments are mostly responsible for transportation
infrastructure planning within their jurisdictions, most highways serve areas
beyond their boundaries.
FAF provides a comprehensive data set but not all of the data are directly
applicable to our research problems because services are also included in the annual
flows among NIEMO sectors. Service values need to be excluded from the model.
Therefore NIEMO freight flows are used as freight flow input values. However,
NIEMO does not account for transportation mode. Consequently, truck proportions
from FAF are used to apportion NIEMO-estimated trade flows to obtain truck
shipments. These are then loaded onto the highway network.
We also require data on the shipping costs associated with all flows. Total
shipping costs between states are estimated using data from NIEMO, IMPLAN,
and FAF as follows

ShipCost ij ¼ TruckCostValue i

TTradeValue ij TruckP ij ð 4 : 8 Þ

where
ShipCost ij k are aggregated shipping costs from state i to j by commodity sector k.
TTradeValue ij are total trade values obtained from NIEMO for 49

49 states

(Hawaii and Alaska are omitted for obvious reasons).
TruckP ij are truck proportions of total trade calculated by applying truck output
values divided by total output values obtained from FAF data. Data for
114 MSAs by 114 MSAs are aggregated to 49 states by 49 states.

PurTruckSer i

TruckCostVaule i ¼

TIndValue i are truck costs per value. These are the truck cost
proportions in origin states obtained from IMPLAN. IMPLAN's sectors are
aggregated to 29 USC commodity sectors. PurTruckSer i k are the total value of
purchased services by the trucking sector and TIndValue i k are the total output of
industry sectors.
Increased time and distance proportions are estimated by applying the user
equilibrium (UE) model. Time changes and distance changes are separately
modeled in Eqs. (4.6) and (4.7).