This page shows the steps needed to set up a calibrated Global Trade Analysis Project (GTAP) model, using a GTAP database.
Getting GTAP data from the GTAP database
The GTAP data is licensed for use, and can be acquired from the GTAP website. A user account is required to log in before data can be accessed. There are multiple GTAP database versions.
Note the following conditions:
You will need the data provided as a FlexAgg package;
The data must have the following files in the 'Fujitsu' Header Array FORTRAN format (extension
.har):"gsdfset.har"(GTAP sets),"gsdfpar.har"(GTAP parameters), and"gsdfdat.har"(GTAP data).GlobalTradeAnalysisProjectv7.jl runs version 7 of the GTAP model, and is compatible with version 11 of the GTAP database. The package may be compatible with older versions of the database as well, but it has not been checked for such compatibility at the present time. Since EpiEconShocks.jl relies on GlobalTradeAnalysisProjectv7.jl, it has also only been tested for version 11 of the GTAP database.
Using inbuilt model initialisation
The ModelInit module in EpiEconShocks.jl provides ModelInit.initial_gtap_model to help get a calibrated 'baseline' model against which models incorporating shocks can be compared.
The function simply needs to be pointed at the directory holding GTAP data, and to be provided with a string vector of regions of interest. All other regions, and the EU countries are aggregated into a 'rest of world' and 'EU states' category, respectively. This default behaviour is because the package is currently aimed at a U.K. context.
EpiEconShocks.jl currently aggregates GTAP regions into the United Kingdom, some major global economies, EU states, and all other regions. Countries modelled separately are the USA, China, France, Germany, China, India, and Japan.
EpiEconShocks.jl currently aggregates GTAP commodities based on the International Standard Industrial Classification of All Economic Activities (ISIC) revision 4 into all primary sectors ("allprimary", ISIC AB), manufacturing ("manufac", ISIC C), utilities ("utilities", ISIC DE), construction ("constr", ISIC F), retail ("retail", ISIC F), transport ("transport", ISIC H), hospitality ("hosp", ISIC I), information and computing-related services ("ictprofserv", ISIC JKLMN), publishing and administration ("pubadm", ISIC OPQ, U), and arts, recreation and others ("artsrecother", ISIC RST).
EpiEconShocks.jl currently aggregates endowments into "land", "skilled labour", "unskilled labour", "capital", and "other",
This function returns a GlobalTradeAnalysisProjectv7.model_container_struct object which can be passed to downstream functions.
Manually aggregating GTAP data
To set up your own GTAP model with customised sector and region aggregation, the helper functions Helpers.cluster_regions, Helpers.cluster_commodities, and Helpers.cluster_endowments can help with preparing NamedArrays to be passed to GlobalTradeAnalysisProjectv7.aggregate_data() to aggregate the GTAP data.
Warnings
A non-exhaustive list of issues discovered while using GlobalTradeAnalysisProjectv7.jl are laid out below, to serve as reference points for users of EpiEconShocks.jl encountering the same issues.
- Using the full GTAP dataset (even at the
FlexAggstage) to set up an initial model is prohibitively memory intensive. For example, it fails on a 32 GB RAM machine during local testing. Users are strongly advised to aggregate the data to the geographies of concern using the functionGlobalTradeAnalysisProjectv7.aggregate_data. We suggest that an appropriate aggregation would be to the countries of interest and their major trading partners; all other regions can be aggregated into a 'rest of world' category. Similarly, commodities and endowments should be aggregated to a sensible set.
Please contribute or report other issues encountered so these can be added here.