GTAP Resources: Frequently Asked Question Details

Fluctuating agricultural production shares are a problem for Australian data (we have droughts).
An example of the "typicalization" procedure you suggest is described in the book: Adaptation and Survival in Australian Agriculture, Peter J. Higgs, Oxford

I dont think Altertax will help you much. With Altertax, the world becomes very COBB-DOUGLAS-like, so shares, in general, are preserved. You want to change the shares.

I had an idea about this.
The idea involves constructing a special TAB file as follows: For each flow data item, a corresponding %change variable, which is used to update that flow.

Equations which relate flow variables; for example, wages used in a sector follow value of output of that sector The constant of proportionality would be an exogenous SHIFT variable.

Accounting equations as in GTAP so that change in flows add up.

Equations defining summary variables, eg: total rice exports of Thailand

Then, to create a database with larger Thai rice exports, you would exogenize and shock the [Thai rice export] variable, whilst endogenizing some matching shift term (eg, a taste change term which increased each users value share of Thai rice in all rice usage).

The updated data base would satisfy accounting restrictions. The procedure is similar in numerical effect to a RAS or FIT procedure but very flexible with regard to instruments and targets. It works on values only, not on prices or quantities separately.

Recently, I did a similar thing for an older single country indonesian database.
In that case I adapted an existing model. I used a special closure: factor prices were exogenous, factor usage endogenous. In my simulations, prices did not change so substitution did not happen. I exogenized and shocked the chief expenditure side components of GDP [C I G X M]. Endogenous were matching shift variables. The result was a database which matched recent National Accounts data, whilst maintaining as nearly as possible the share structure of the original data.

I believe a framework of the type described above could be practicable and useful for GTAP.
Maybe someone will construct it one day. In fact it would not surpise me to learn that someone else already has !