|
Memory RAM for full RunGTAP
I appreciate if anyone can give me hint regarding the amount of memory RAM we need to run the full RunGTAP. We're thinking in upgrading our computer systems and we have to consider the needs that models as RunGTAP and GTAPDyn impose. In particular we know that, for instance, the full RunGTAP model does not run in my PC with one giga memory -I get a message "unable to allocate sufficient memory. (I0STAT is 912)". Does anyone know how much memory is needed to run the full regional and sectoral specification model? In the case of GTAPDyn, how much is needed? Do you think it is advisable a more powerful computer like those with dou-processors? Are GEMPACK and RunGTAP compatible with UNIX? I have heard that is not feasible to expand the memory RAM further than 2 GB using the standard XP Professional, is that right? Overall, what do you advise: upgrade the memory RAM of the current PCs and run the models with UNIX or any upgrade WINDOWS or the acquisition of a type "server" system?
View Answer
This is an interesting question and topic.
To run full GTAP (GTAP.TAB, comparative static, all regions and commodities) requires more than 2GB of RAM. To do the same for dynamic GTAP (GDYN.TAB) would require more memory. I say more about getting that much memory later.
However it may not be a good idea to try to solve the fully disaggregated versions of these models. GTAP provides excellent facilities for aggregating the models. Usually you can get just as good results for any specific policy question by making a suitable aggregation. And that will probably solve within 1 or 2GB of RAM, and solve much more quickly than the fully disaggregated model would. Trying to solve the fully disaggregated model is difficult because of memory problems and takes a very long time even if you have the memory etc.
Accessing more memory with GEMPACK ---------------------------------- > I have heard that is not feasible to expand the memory RAM further > than 2 GB using the standard XP Professional, is that right? That is correct.
But recently there are Windows PCs with 64-bit processors and 64-bit XP operating system. These break the 2GB memory limit. In order to run GEMPACK on such a PC and access more than 2GB for each simulation, you need to install GEMPACK using a 64-bit Fortran compiler. We will support the Intel Fortran compiler (which runs on 32-bit and 64-bit PCs) with Release 10 of GEMPACK (due latish in 2007). We have a beta test version of Release 10 available to selected customers now.
You can read more about these issues in the paper "Running Simulations Faster on Multi-Processor or 64-Bit PCs via GEMPACK" by Mark Horridge and me. This paper is available at http://www.copsmodels.com/elecpapr/c15.htm
Dual processor machines ----------------------- These are now becoming commonly available for 32-bit and 64-bit PCs. And Release 10 of GEMPACK provides a way of speeding up simulations on such machines - see the paper referred to above. But that does not help with memory problems - indeed even more memory is need to sensibly make use of 2 or more processors. For example, if you want to use 2 processors to speed up a simulation and if that simulation needs 3GB of RAM, you will need that much memory for each processor (that is, total of 6GB).
Unix? ---- Yes, GEMPACK is available on Unix machines. But I would recommend going the 64-bit route (where your familiar other word-processing etc utilities are available).
Most importantly I would urge you and others to consider whether you can achieve your policy objectives by aggregating the model suitably. If so you can avoid the complications (and costs) of high memory PCs for the present - such machines will become mainstream soon but at present they are a bit at the cutting edge.
|
How do I open HAR files?
How do I open HAR files?
View Answer
HAR files can be opened with the ViewHAR program, a free software produced by Mark Horridge at the Center of Policy Studies, Monash University.
|
Difference with GDP numbers in GTAP
Why are the GDP (or other macro data) numbers in GTAP so much different from my country’s official statistic?
View Answer
There are several reasons why this can happen. For example, our source for GDP data, the World Bank dataset, may have different definitions of a reference year for different countries. These could vary quite a bit across countries and obtaining harmonized information for all countries for 12 months in a year would be daunting task, requiring expertise from different countries as well as harmonization efforts. Therefore, we accept the World Bank numbers as harmonized enough. What is more important from GTAP’s point of view, for CGE modeling are the cost and sales shares, which are relatively more stable. What we strive to provide here is a global database, which has IO tables of different countries, linked with trade, tariff and other policy data, in a consistent and balanced manner. In that process, we may not have reached complete consistency or harmonization in one or more components of our data, to the extent that our sources have not ensured the same. It is for this reason that we request our users to be careful and exercise judgment while using the data for specific purposes. You can match the exact GDP you want it to have via Altertax simulations. So, we urge our users to put in a bit more effort into improving the representation of the sectors and countries important for them, using Altertax, given that they have access and more knowledge of their economies. See for example, the following study, wherein the authors had done the same for a study on Egypt: https://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=2466You can learn more about Altertax at: https://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=315
|
Log-file warnings and re-ordering of data when I convert GTAP 8 Data Base in to GAMS format (GDX file)
I see some log-file warnings and re-ordering of data when I convert GTAP 8 Data Base in to GAMS format (GDX file). Is this a problem? How can I avoid this issue?
View Answer
In GTAP 8 Data Base, the new countries Oman (omn) and Israel (isr) share the same codenames with the sectors Other Mining (omn) and Insurance (isr), respectively. This seems to cause a minor problem in GDX files: it rearranges the data such that Oman is the first country, while the rest of the data remains in the same order as in the original data. You have two alternative options to exercise: - Use the same data file to do your analysis, while recognizing the re-ordering issue mentioned above.
- Re-label these two countries in the HAR file to begin with, by assigning the names that you may be more comfortable with. For example, you can rename omn as c_omn and isr as c_isr. This is easy to do, if you follow these steps using the Viewhar program that you use to see the HAR file:
- On the Viewhar menu, please click Sets...View Set Library. If you do not see "Sets", click File...use advanced editing menu, to see it.
- You will see a list of sets and their elements. Click on PROD_COMM to see the 58 sectors (including CGDS).
- Below the elements you will see different options: click on Edit elements.
- Now, you can change the sectors omn and isr to their new names and then click OK. Repeat the same steps (2-3) for TRAD_COMM. Alternatively, you could change the names of regions in set REG.
- Click on close, once you are done with the changes. Now, Viewhar will prompt you with a question like this: "Set labels on the following RE headers are inconsistent....... Do you want to change the labels on these headers?". Please say "Yes" and you will realize that you have changed the set element names all over the file. Now you can save the file and then convert it to GAMS.
|
GTAP 7 vs. GTAP 7.1 regions
Why does GTAP 7.1 Data Base have only 112 regions, though GTAP 7 Data Base has 113 regions?
View Answer
We found some serious data quality issues involved in the contributed Myanmar table, which was included in version 7 and removed later in version 7.1.
|
Decomposing trade effects of an FTA simulation
Is it possible to decompose the trade effects of an FTA simulation into trade diversion and trade expansion effects?
View Answer
There is no official way of decomposing the trade diversion and trade creation effects of an FTA in GTAP. An extensive discussion of trade creation and diversion is available in GTAP Working Paper #26. There are two things you can examine to help you get at trade diversion and trade creation type effects, however. 1) You can decompose the IMPORTDEMAND equation into the expansion and substitution terms. Equation IMPORTDEMAND # regional demand for disaggregated imported commodities by source (HT 29) # (all,i,TRAD_COMM)(all,r,REG)(all,s,REG) qxs(i,r,s) = -ams(i,r,s) + qim(i,s) - ESUBM(i) * [pms(i,r,s) - ams(i,r,s) - pim(i,s)]; Ignoring ams, the expansion term is (qim(i,s) ). This term shows you the overall expansion effect on trade of the FTA (a trade-creation-like effect). The substitution term (- ESUBM(i) * [pms(i,r,s) - pim(i,s)]) shows how we substitute between imports from different sources (a trade-diversion-like effect). Note that the overall impact of FTA on trade, is qim(i,r). If you sum the trade-diversion-like effects (qxs(i,s,r) - qim(i,r)) you will get zero (since the weighted sum of qxs should = qim). So trade is greater than qim in some regions and less in others but overall trade increases by qim. 2) The second approach is to use the allocative efficiency effects in the welfare decomposition to show trade creation and diversion. These are volume changes combined with the tariff revenues - so are dollar amounts and therefore can be aggregated. As they are efficiency effects, you are getting the welfare gain/loss from the trade creation/diversion. To get trade creation and diversion one might aggregate the allocative efficiency gains (particularly those related to MTAX) over sources within the FTA and outside of the FTA. The increase in trade/allocative efficiency gains within the FTA and the decrease in allocative efficiency with countries outside of the FTA, respectively, are again trade-creation and trade-diversion-like effects that give an indication of the extent of trade diversion and creation.
|
Metadata description of GTAP data
Is there a metadata description of the GTAP data?
View Answer
|
Translating GTAP base data into CSV format
Is it possible to translate the GTAP base data into CSV format for loading into a different application, such as MS Access?
View Answer
Yes it is possible to do this and there are different ways to get CSV files from HAR files in GTAP Data Base: 1. You could open a header in the HAR file and then do Export..save screen as CSV file. You will have to do this for every other header to complete this task. Sometimes this task may be daunting given that some headers have more than 2 dimensions! 2. You could download some programs from GEMPACK website, mainly the following: http://www.copsmodels.com/gpmark9.htm. Har2csv is the program that will help you convert the HAR files to CSV format.
|
Costs not matching with sales in GTAPVIEW file
Why don't the costs (computed from CM01) match with sales (computed from SF01) in the GTAPVIEW (gsdview.har) file?
View Answer
SF01 includes OUTAX, which is the output taxes. This is not included in CM01, the cost side value. You have to add up the taxes and the base from the cost side and remove investment (CGDS) from CM01. Then, the balance condition will match if you add the output taxes to the CM01 total.
|
RORG
I try to impose a shock on rorg or globalcgds, but they are not in the default exogenous variable list. I checked equation RORGLOBAL, rorg should be exogenous when RORDELTA=1 or globalcgds is exogenous when RORDELTA=0. But they are both endogenous in the default closure. How can I exogenize rorg or globalcgds?
I understand that total investment is constrained by total savings, but I could not find such an equation in GTAP tablo. The equation WALRAS would impose such constraint, but when ‘walraslack’ is endogenized in the default closure, this constraint breaks down. Could please tell me which equation shows the constraint of savings over investment so the ‘walraslack’ will be zero (when it is endogenized)?
View Answer
1) You cannot increase global investment without increasing global savings (unless you want to break walras law, which I won't answer here). You can increase some countries investment but then other countries' investment will have to fall so that global savings remains equal to global investment. So if you want to increase global investment then you need is to increase the savings of one or more regions, thereby increasing global saving and hence allowing global investment to increase.
You then decide how you want this global increase allocated across regions:
RORDELTA = 1, would then allocate that increased global savings according to rates of return. In this one rorg is determined endogenously as the rate at which global savings equals global investment. It would be driven down as the increased global savings tries to find new investment opportunities.
RORDELTA=0 allocates the increase proportionately across all countries. In this case this proportion is determined by global saving and hence global saving equals global investment is maintained. The rorg is then just endogenously determined as the weighted average of the rates of return afterwards.
2) If income equals expenditure in each region and all markets are in equilibrium, then the last one is in equilibrium automatically. This is walras law. You don't need the equation. Walraslack should equal zero automatically and this works as a test on the model.
|
How to use GTAP Data Base for modeling applications?
How to use GTAP Data Base for modeling applications?
View Answer
|
Working with HAR files
How to open, read, convert or copy the contents from a HAR file
View Answer
- Download and install a program named viewhar.exe from: http://www.copsmodels.com/gpwingem.htm
- Open the HAR file, with this program viewhar.exe.
- You can now see the data on different headers, by doubleclicking each of them.
- You can copy the numbers from the HAR file to Excel by Export..Save Screen as CSV or Copy Screen to Clipboard and then pasting into Excel.
- If you want to convert this data into Excel readable form in entirety, you will have to use a program named HAR2CSV, which is also freely available at: http://www.copsmodels.com/gpmark9.htm
Particularly look at bonus.txt for more details on them: http://www.copsmodels.com/ftp/gpextra/bonus.txt
You need to be a bit familiar with DOS prompt to use this utility, so if you find step 4 convenient, please do it that way. To learn DOS environment just enough to handle this stuff, you might want to consider going through parts of the following document and exercises. Though these are meant for a different purpose and audience, I’ve often found them useful for anyone who wants to learn to work in DOS environment for some fairly simple tasks: http://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=2850
|
GTAP Data Base Values
Are the values in the GTAP Data Base data or estimation results by GTAP Model?
For example, I am checking GView, and there is a table called "NVFA". The table decomposes a total output into various inputs. Do the values in the table come directly from I-O table data? Or are they results of GTAP Model’s estimation?
In this sense, I am not very sure about a relationship between GTAP Data Base and GTAP Model. Is there GTAP documentation which explains this point?
View Answer
The GTAP Data Base is a result of reconciliation of data from various I-O tables, international bilateral trade data, tariffs, domestic support, export subsidies, macro-economic data, econometrically estimated value-added shares, elasticity parameters, etc. So what you see in the files like Gview.har is constructed, compiled and adjusted from various sources. We would suggest you go through our detailed documentation of the data base available online, to know more about how the data was constructed: https://www.gtap.agecon.purdue.edu/databases/v7/v7_doco.asp
You will notice from the documentation that none of the values in this data are estimated/calculated from the GTAP Model itself. However, the accounting relationships and linkages that are imposed in the data, while reconciliation, are all seen in the model as well. This is what we can think of about the relationship between the model and data.
For example, the fact that GDP comprises C,I,G, X-M is captured in the data as well as in the model, but that does not mean that these values are estimated/determined by the model. Of course, once you run simulations using the model, the updated data would contain numbers determined by the model.
|
Variable "to" defined with regards to "POS"
I am a new user of the standard GTAP model, which I am using to simulate a removal of EU argicultural subsidies on different EU agricultural sectors. Would it be possible for you to answer a couple of short questions that I have?
1) How is the variable "to" defined with regards to "POS" found in chapter 10C of the GTAP 7 documentation?
2) Is there a variable that corresponds directly to "POS"?
3) How is it that "to" is positive for certain agricultural subsectors, such as wheat in EU_25?
View Answer
Please note that chapter 10.C corresponds to GTAP 7 Data Base, but if you are using GTAP 7.1 Data Base you should refer to 10G for EU agricultural subsidies. The variable "to(I,r)" that you see in the GTAP model is the % change in the power of subsidy (power = 1+ rate of subsidy).
To see the levels of subsidy rates, you can look at the header rTO in gsdtax.har in the database.
This is calculated by dividing the output subsidies by total output for each sector. POS mentioned in the documentation gives the total power of support in all categories. However, rTO may be obtained merely by the following steps:
1. Multiply the (POS-1) by total pre-tax output VOA (you can get this from CM04 of gsdview.har). This will give you the total subsidies in the sector.
2. Multiply the total payment obtained above by the %share of output subsidies divided by 100. This will give you actual output subsidy payments.
3. Divide the output subsidy payments by total output and multiply by 100 to get rTO.
There is no variable or coefficient in the model or database that corresponds to POS. It is merely to give the users an idea about the total support in the database, but holds less significance in terms of simulations, where you’d actually want to target output subsidies, factor subsidies or input subsidies.
TO is positive for many sectors because there are subsidies in them. You may want to check with the initial data presented in chapter 10G to see the level of subsidies in the contributed data.
|
Final/Intermediate Demand and Producer Taxes
There are several kinds of taxes included in the GTAP database. In most cases the tax base seems to be quite straight forward. However for several taxes we are uncertain which taxes were used as basis for the data given. Therefore my question:
Is it indeed the case that the commodity taxes on:
1. intermediate usage on domestic products,
2. household consumption on domestic products,
3. investment usage and
4. government usage of domestic products
represent a combination of VAT and e.g. consumption taxes on beer/tobacco? If so, how can we split those?
‘Domestic firm and investment tax’ would that be VAT only or does that include anything else as well?
We understood that ‘ptax’ is a sort of producer tax. What does that exactly include? Excise tax?
View Answer
Yes, it is. Note that VAT deductions are netted out from VAT payments, so that VAT recorded on sales for intermediate usage is typically zero.
You cannot split those, without resort to other information sources.
Domestic firm and investment tax includes any tax that creates a wedge between price paid by purchasers and price retained by producers.
ptax, not excise tax, which is a commodity tax, but any non-product indirect tax paid by enterprises. Examples include payroll tax, property taxes, motor vehicle taxes, and business licenses.
|