Creating and managing an impossibly large CGE database that is up-to-date

Large-scale multi-regional CGE models of Australia, such as MMRF and TERM, underlie most CoPS consulting work. The regional detail, modelled in bottom-up fashion, greatly interests policy makers and is often needed to answer questions like: how would less rainfall in southern Australia affect the economy? To support this work, we have devised a variable disaggregation master database for any combination of over 1,400 statistical local areas (SLAs). Such a database may represent 172 sectors in over 100 regions. It would be slow to run simulations with so much detail, so we routinely aggregate the database before using it. Each aggregation is tailored to preserve the regional and sectoral detail that is pertinent to a particular policy issue. This paper describes the procedure used to generate a master database, starting from a published national 2001-2 input-output table, and various international trade and census data from other dates. A levels adjustment program is used to update the published input-output table from 2001-02 to 2005-06. This makes it easier to use regional data from national accounts and the 2006 census. The whole procedure is automated via a series of programs. This forms a basis for documentation and also allows us to repeat the whole procedure with different inputs. For example, we could reaggregate the SLAs to distinguish zones within each capital city, if desired. Some of our methods may be useful in global CGE models as practitioners modify databases to deal with policy issues. We have adapted the methodology of variable disaggregation to develop the first ever bottom-up representation of the Australian economy by its 150 federal electorates.