Comparing Regional CGE models

Top-down and bottom-up approaches

Several CGE models with a regional dimension have been developed at the Centre of Policy Studies (CoPS). Each model has its own advantages. Two different approaches are used to add a regional dimension: top-down and bottom-up.

Under the top-down approach, national results for variables such as output, employment, and final demands are disaggregated into 8 states or into more, sub-state, regions. An input-output methodology is used, which recognizes regional variation in quantities but not in prices. Local multiplier effects are recognized, so that, for example, a construction project in Queensland would generate employment there, which would in turn stimulate consumer spending on non-tradeable goods produced in Queensland. The 'top-down' approach is economical both of computer resources and data -- which allows very detailed models(say 120 sectors, 100 regions) to be implemented and solved quite easily. On the other hand, region-specific supply behaviour is not easily modelled, and proximity effects (when growth in one region benefits its neighbours) are usually neglected.

The alternative bottom-up approach consists in linking a series of independent CGE models [one for each region] which interact through trade and primary factor flows. In these multiregional CGE models both prices and quantities may vary independently by region. This type of model makes fewer theoretical compromises -- but imposes high computing and data demands. Consequently a more aggregated model must be used, sacrificing some sectoral or regional detail. In practice, the number of regions plus the number of sectors must not exceed 100.

Hybrid models combine top-down and bottom-up approaches. For example, MMRF-GREEN is a bottom-up model of the 8 Australian states. Top-down methods are used to break down results for each state into results for 10 or more sub-state regions.

Comparison of regional CGE models used at CoPS

Regional detail in the VU-National model (formerly called MONASH) is achieved by top-down methods. VURM(formerly called MMRF) is an 8-region bottom-up model. TERM is another bottom-up model distinguishing up to 57 separate regions. The differences are summarized in the table below.

Model VU-National VURM TERM
Type of regional modelling Top-down Bottom-up Bottom-up
Region-specific prices no yes yes
Region-specific quantities yes yes yes
Typical no. of sectors 113 around 40 around 40
Typical no. of regions 8 states or 57 statistical divisions 8 states
(but see below)
57 statistical divisions
Forecasting (year-to-year dynamics) yes yes usually
Region-specific demand-side shocks yes yes yes
Region-specific supply-side shocks no yes yes
State Government Accounts module no yes sometimes
Additional "top-down" regional detail Top-down breakdown to 8 states or 57 statistical divisions Within-state top-down breakdown to 57 statistical divisions Within-statistical-division top-down breakdown to 1379 statistical local areas
Other features In some versions, inclusion of full bilateral matrices of inter-regional trade allows for growth in one region to spill over into neighbouring regions. Some versions include detailed modelling of CO2 emissions Exports and imports distinguished by port of exit/entry. Sectoral and regional aggregation tailored to particular simulations

Go to VURM page

Go to TERM page

Go to VU-National page