Skip to content



There are various settings where it makes sense to construct a new benchmark from the GTAP data base before shocking the model instead of comparing different shocks against each other where all comprise the same changes against the benchmark. One option to construct a new benchmark changes first entries in the global SAM and next re-balances it. That approach is especially suitable if benchmark changes are motivated by new or updated data. The filtering approach implemented in CGEBox can be seen as a specific variant of that approach.

The option discussed in here uses the simulation model itself. That approach is especially appropriate for changes where the behavioral responses embodied in the equation structure provide a better way to re-adjust the benchmark compared to the more mechanical re-balancing the approaches. Conceptually, that is straightforward as the accounting identities underlying a balanced SAM must also be fulfilled by a simulation with the CGE itself. The main challenge is to map the simulated results back in same format used on the input side of the simulation model. As CGEBox is a SAM based CGE approach, that step is not overly complex. That brief documentation will discuss the basic methodology and technical implementation of AlterTax in CGEBox.

Technical realization

The technical realization re-uses existing elements of CGEBox to implement alterTax:

  1. A predefined configuration which can be chosen from the interface set-ups a simulation model useful for alterTax, i.e. the parameterization of the model is changed such that most substitution elasticities are unity (or close to unity).

The user can select any shock file:

  1. As part of post-processing, the relevant entries for a start of the model define a new benchmark, including the global SAM.

Note: The alterTax output can be also generated from any normal model run, in case the pre-configuration given next is not suitable.


The pre-configuration changes the following elements comparted to the standard GTAP Model:

  • The demand system for private households is changed to a CD.

  • The production nesting using substitution elasticities of unity between VA and the intermediate composite, between primary factors and between intermediate.

  • The Armington elasticities are set close to unity (a CD implementation is currently not supported).

  • All factors are mobile

The use of the CD functional form allows to keep the value shares constant during the alterTax simulations which thus maintains basic relations between transactions in the SAM. If highly detailed global SAMs are subject to AlterTax, the per-solve mechanism can be used as well.

Post-Model processing

The SAM output option is switched on and the results for the shock are stored again as a SAM in a new data file. The same holds for various sub-matrices, for instance:

The update symbols are stored to a new data file:

Note that the AEZ implementation does not yet work properly.

Use of the updated benchmark

The updated benchmark data set is stored in the data directory with the post-fix “alttax” appended and can be used as any other data set: