Does It Matter if Trade or Transport Data Are Used in SCGE Modeling?

Does It Matter if Trade or Transport Data Are Used in SCGE Modeling?


W Hansen, Institute of Transport Economics, NO


This paper analyzes the differences between trade and transport data in a SCGE model. P/C zone pair matrices are established from commodity flow survey data for use in the Norwegian SCGE model PINGO.


The recently performed Norwegian Commodity Flow Survey (CFS) conducted by Statistics Norway enables us to construct Production / Consumption (P/C) zone-pair matrices for the trade flows in the SCGE model PINGO. Generating P/C matrices from the CFS data makes it possible to analyze the difference between transport and trade data in freight transport demand modelling.

The Spatial Computable General Equilibrium (SCGE) model for Norway, PINGO, was developed by Ivanova, Vold and Jean-Hansen (2002). Vold and Jean-Hansen (2007) developed a second version of the PINGO model, making it consistent with the year 2003 base case matrices and the logistic model for freight transport in Norway, and further disaggregating sectors and commodity groups. In the PINGO model, the elements of the P/C matrices are established through the use of deterrence functions on the marginals for total input and output per agent, where mode-specific count data are used to calibrate the deterrence function in order to ensure that there are deliveries between all regions, also those that are far apart. The marginals for total input and output per agent are constructed using available statistics. PINGO is a static SCGE model and is part of a neo-classical general equilibrium-modelling tradition assuming constant return to scale and perfect competition in all markets.

In the analysis of the demand for freight transport we usually separate between two different types of zone-pair matrices for goods movements, Production / Consumption (P/C) matrices and Origin / Destination (O/D) matrices. The P/C zone-pair matrices establishes trade links between the production sites and intermediate or final consumption of the goods, and are usually constructed from commodity flow surveys. While the O/D data are usually established from mode-specific count data and represents the physical transport movements of the P/C trade flows. The current flow matrices used in the PINGO model are somewhere in-between these two. When only O/D matrices are available, the commodity flow patterns in the SCGE models are usually established using gravity models. This method provides P/C matrices that reflect the assumptions in the underlying model and hence do not portray the actual trade.

This paper compares and analyses three different ways of constructing P/C zone-pair matrices in a SCGE setting: P/C matrices based on O/D zone-pair matrices of shipments, P/C matrices from the current PINGO model, and P/C matrices generated from commodity survey data.


Association for European Transport