"GTAP Model Version 6.0"
by Hertel, Thomas, Robert McDougall and Ken Itakura
Abstract
Version 6 of the GTAP Model may be used with the 2001 release of the GTAP Data Base (version 5).
Changes from v.5.0 to v.6.0
 New Regional Household Demand System
In a recent GTAP technical paper, Robert McDougall (2000) proposes a new regional household demand system for the GTAP model. This demand system fixes a problem with the original demand system stemming from the inconsistency of the constant budget share assumption with the nonhomothetic, CDE expenditure function for private consumption. McDougall points out that the cost of private utility in this model varies according to the amount "purchased" by the regional household. If the regional household's utility maximization problem is reformulated to take this into account, then the optimal share of expenditure devoted to private, public and future savings consumption varies as a function of per capita expenditure. While the empirical difference between this model and the original one is small for many applications (see Hertel (2001) for some examples). This is not always the case, and the conceptual difference is substantial. Incorporating McDougall's theory has required a significant rewrite of the regional household module. In so doing, McDougall (2000) also introduces a new approach to fixing any of the components of final demand via preference shifts. So, for example, if the user chooses to fix real government spending, she now does so by swapping a preference shift variable, dpgov(r), with ug(r). Preferences are accordingly altered in order to accommodate the desired pattern of expenditure in the new equilibrium. This preference shift also has implications for the welfare decomposition, to which we now turn.
 New Welfare Decomposition
The version 6.0 GTAP.TAB file incorporates some significant changes, which are also reflected in DECOMP.TAB. The first of these involves eliminating the term relating to nonhomothetic preferences. In the new welfare decomposition proposed by Hanslow (2000) and McDougall (2000) the influence of nonhomothetic preferences  as evidenced in the cost elasticity of utility  is embodied in the common scaling factor. Therefore it is no longer present as a separate term. McDougall also corrects the computation of Equivalent Variation (EV) so that it is no longer an approximation, but rather now provides an exact measure of the EV associated with a given policy simulation. His new formulation also has the virtue of being invariant to the scaling of the CDE parameters. Two further changes to the welfare decomposition include: (a) explicit treatment of preference changes and (b) normalization with respect to population. The first of these means that when a nonstandard closure, such as fixing real government spending, is employed, the welfare decomposition identifies the "contribution" of this change in preference to the overall welfare change. The fact that the decomposition is now done on a per capita basis means that if population is shocked in the model simulation, this will show up as a separate "contribution" to aggregate regional welfare.
 Iceberg Trade Costs "ams" importaugmenting "tech change" variable
The parameter "ams(i,r,s)" has been introduced to handle bilateral services liberalization as well as other efficiencyenhancing measures that serve to reduce the effective price of goods and services imports. Shocks to ams(i,r,s) represent the negative of the rate of decay on imports of commodity or service i from region r imported by region s. When ams(i,r,s) is shocked by 20%, then 20% more of the product becomes available to domestic consumers  given the same level of exports from the source country. In order to ensure that producers still receive the same revenue on their sales, effective import prices (pms) fall by 20%. The introduction of this variable also facilitates simulation of efficiency improvements such as customs automization or ecommerce.
 Baldwintype capital accumulation
Francois, et al. (1996) explores the interaction between trade policy and capital accumulation in the GTAP model. They follow Richard Baldwin in arguing that in the standard static setting CGE models undervalue the positive relationship between trade, investment and growth due to the absence of capital accumulation effect. A simple, onesector neoclassical growth model is used to illustrate the basic idea behind this argument. An efficiencyenhancing reform such as trade liberalization shifts upwards the economywide production function. The same amount of capital and labor now can produce more than before, thus increasing income. This is the static effect of the reform. Under the assumption of fixed saving rates, part of the increased income is saved and invested to form new capital, which results in further income gains. This mediumrun effect is missing in the standard comparative static GTAP model. The version 6.0 GTAP model introduces an equation named BALDWIN to incorporate the multiplier effect by feeding increase in gross investment back to capital services available in economy, following the approach outlined in Francois, et al. (1996).
 Uniform consumption tax, "tp"
A new tax variable, tp(r), is introduced to permit implementation of a uniform adjustment to all private consumption taxes in a particular region. The variable can be swapped with the variable DTAXR(r), the change in ratio of taxes to regional income, to generate a tax replacement scenario, whereby taxes remain a constant share of regional income with the adjustment (increase or decrease in revenue) fully absorbed by this consumption tax. This is commonly used to replace lost tariff revenue under trade liberalization scenarios.
 Correction on equations for the ratio of taxes to income
The variables of the ratio of taxes to income were introduced in the version 5 of GTAP.TAB model file. The idea was to look at the ratio of taxes to income in order to preserve homogeneity in prices. We can also look at changes in tax revenue, but then a uniform price increase would change those variables. Obviously a simple percent change variable doesn't work, since many taxes are initially zero. The basic logic of this approach is as follows:
Let R be the ratio of taxes to income: R = T/Y, then:
dR = d(T/Y) = R(t y)
where t = dT/T and y = dY/Y, and multiply through by Y to get:
YdR = dT  yT (1)
This ratio change is computed for each tax type and for total taxes.
Then the change in tax revenue itself may be computed as:
dT = YdR + yT
in order to determine regional income.
In the v.5.0 GTAP.TAB file the equations were written as in the form of (1). As a rule used in the GTAP model, however, small letters are defined as percent change so the equations in the form of (1) should look as seen in the equation,
TOUTRATIO:
100.0 * INCOME(r) * TOUTR(r)
= sum(i,PROD_COMM, VOA(i,r)*[to(i,r)]) + sum(i,PROD_COMM, PTAX(i,r) * pm(i,r) + qo(i,r)])
 TOUT(r) * y(r)
