Towards A Generic CGE Model for South Asian Trade Liberalization

6.1: What is the "appropriate" class of CGE models for dualistic economies?

In order to discuss how to incorporate poverty analysis in a CGE model, we need a clear understanding of the structure of CGE models as such. As a first step in understanding the CGE models, we can start with the Walrasian "fundamentalist" approach to general equilibrium. Essentially, the problem here is to find a set of prices (a price vector) that will clear all markets.¹⁷

The producers maximize profit and the consumers maximize utility. All markets including futures markets must exist and all uncertainty must be subject to actuarial calculation of risk¹⁸. It is clear that while theoretically elegant and analytically impressive, the conditions in many actual economies do not approximate this theoretical model.

In the Keynesian type macroeconomic models at any rate, there can also be underemployment equilibrium. There is thus a tension between such macroeconomic models and the Walrasian general equilibrium models where full price flexibility ensures full employment at market clearing wage level.

As Robinson (2003) observes: The literature on CGE models is replete with debates about the macro properties of these models, and a number of different schools of thought have emerged concerning how, or indeed whether, one should incorporate macro features into these SAM-based models. No clear consensus has emerged, which is hardly surprising since the debate really concerns the theoretical dividing line between Walras and Keynes, and the micro foundations of macro models--- or the lack thereof. (p. 1)

It is not relevant here to outline the contours of this debate, except to keep in mind that neither the fully Walrasian nor the standard Keynesian model is likely to correspond exactly to the actualities of a developing economy. However, we need to keep firmly in mind that a CGE model in its origin--- and initial historical development--- is Walrasian in spirit.

At the applied level, a CGE model incorporates all the flow variables that can be captured in a SAM.¹⁹ The origins of social accounting can be traced as far back as Gregory King’s efforts in 1681, but more recent work stems from the attempts by Richard Stone, Graham Pyatt, Erik Thorkbecke and others.²⁰ The relkevant SAM accounts must be specified clearly for the particular CGE model one wishes to construct and implement.

These accounts will usually include production activities, factorial income distribution and household income distribution among other variables. The importance of both the factorial income distribution and household income distribution for poverty analysis in a CGE model are intuitively obvious. However, proper modelling strategy for these distributions in a CGE model is far from obvious. Later, we will have an occasion to deal with the issues that arise in this context in some concrete examples of CGE models for poverty analysis, and finally to formulate an appropriate model for South Asia incorporating the dualistic structure of South Asian economies.

As implied before, the Walrasian spirit of a CGE model is shown in its determination of only relative prices, with some price index being chosen as the numeraire. The model also incorporates the assumption of 'no money illusion'--- all supply and demand equations are homogeneous of degree zero with respect to prices.²¹ If all prices are multiplied by a fixed number, the equilibrium quantities do not change at all.

As a matter of historical record, it has been a standard practice of CGE modelling to specify fixed supplies of factors of production such as various types of labor and capital, or aggregate indexes of these, and carry through the implications of the assumption that all markets must clear. These "classical" CGE models calibrate wage and rental rates to employ all of the exogenously specified labor and capital. In many "applications", the guiding idea has been to introduce distortions to the 'equilibrium price vector' and calculate the resulting inefficiencies. In this sense, CGE models have been used as a normative check for distortions and their costs against the benchmark of a Walrasian market clearing price system.²²

There is also much discussion in the CGE modelling literature about the various “closure rules” for the models. The discussion about macro-closures, initiated by Sen (1963), was revived by Taylor and Lysy (1979) who found that the choice of macro-closure to a large extent affected the policy simulation results obtained with a CGE model. As the previous discussion already indicates, the macroeconomic modelling is forced to depart from the Walrasian assumptions embodied in a “fundamentalist” CGE model. This also leads to the “closure rule problem”. Because the short-run macro CGE models do often deviate from the Walrasian closure, a separate literature has grown up around the various alternatives.

There are mainly two ways to interpret and define the closure rule problem. In mathematical terms, the problem boils down to the simple notion that the model should consist of an equal number of equations and endogenous variables.²³ Thus, the closure rule problem is the decision the model builder has to make on which variables are endogenous and which variables are exogenous. Alternatively, if the model is built in the Walrasian tradition and all decisions are based on optimizing behavior, the closure rule problem involves the introduction of macroeconomic constraints that impinge upon the microeconomic behavior of individual agents. One then needs to introduce additional balancing equations. (Ginsburgh and Keyzer, 1997). In general, a closure rule is determined by the theoretical preferences of the model builder and, in her view, empirically the most plausible adjustment processes.

In the early works that used CGE models for development policy work, much time was spent in finding ways to model the various distortions in the foreign trade sectors. Thus, modelling exports, imports, balance of trade and balance of payments became important items on the modelling agenda during the 1980s. After trying various approaches, a general consensus was reached. The consensus approach admits imperfect substitutability between imported goods and their domestic counterparts. The Armington assumption is invoked by almost all modelers.²⁴ The Armington assumption regarding imperfect substitutability has been extended to the modelling of exports as well. The most common approach now is to specify sectoral constant elasticity of substitution(CES) import demand functions, export transformation functions that assume constant elasticity of transformation(CET) and aggregation functions based on these.²⁵

We may recall that starting with Hume and his price-specie flow mechanism, the classically inspired trade theories have implied a trade balance of zero in equilibrium. But in the real world data the trade balance is rarely zero. Does this mean that the equilibrium assumption is somehow violated? The most widely practiced way of handling this nonzero trade balance is to make it exogenous. Typically, trade imbalances find their counterpart in the saving-investment imbalance.

Looked at in this way, trade imbalances can be treated as foreign saving flowing in with a trade deficit, and of savings flowing abroad when trade balance is positive. However, this does raise the question of why people at home or abroad would be willing to save and lend--- a question that can only be answered in an explicitly intertemporal model. Thus, static CGE models which treat trade balance as exogenous are, in fact, compressions at a point in time of a more fully specified intertemporal equilibrium model.

There is also the related issue of how to bring in balance the traded with the nontraded sector, and the domestic economy with the rest of the world. This is done by making flexible another relative price. This is the relative price of traded and nontraded goods, or under the purchasing power parity and small country assumption, the real exchange rate. Naturally, modelers tend to specify an implicit functional relationship between the real exchange rate and the trade balance. Increased flow of foreign savings raises the relative price of nontraded goods which is equivalent to an appreciation of the real exchange rate in these models (Devarajan, Lewis and Robinson, 1993). There is a shift of production away from exports goods producing sectors to nontraded goods and services. Consumers shift demand to cheaper imports and the new trade balance equals the exogenous flow of higher foreign savings.²⁶

This is perhaps a good place to shift our attention from foreign savings to domestic savings and investment, with the role of the government as a key macroeconomic entity. In a flow description of the economy via the SAM accounts the savings-investment account collects savings and spends money on investment goods. The flow equilibrium condition is that savings must equal investment. Some mechanism is clearly needed to achieve this balance, as our previous discussion of the closure rules already indicated.

The common strategy here is to specify savings parameters by household types. These fixed parameters map income to savings. A fairly common (neoclassical) assumption is also to assume that all savings are spent on investment. Thus under this closure rule there is no “paradox of thrift”. Either through loanable funds markets or a more direct allocation rule( this is often the case), savings are translated into investment. However, this is not the only way to relate savings and investment, and even here, as the reference to the loanable funds markets hints, the full specification of a ‘savings-driven’ model on the financial side is often missing. Important questions regarding the saving-investment links need to be raised. These include: why save at all ? Why spend on investment rather than on consumption? Who owns the new capital stock? Do actors have and care about an asset portfolio? Introduction of proper dynamics is necessary to answer these and other similar questions.

The question of private savings is also related to that of public savings and dissavings, as the case may be. But the government does more than simply generating savings or dissavings. It collects taxes, makes transfer payments and purchases goods and services. Through all these activities it can affect the flow of income and consumption of all or at least some socioeconomic groups. Hence, an intuitive link between government’s actions and poverty is justified. Later, we will see how this link can be made more explicit in a causal sense. For the moment, let us simply observe that in most CGE models government is a rules-based (but not necessarily a utility maximizing) actor. Typically, the monetary side is absent or sketchy. Usually, there is a flow-of-funds specification, but no consideration of how the government finances its deficit. There is simply a crowding out of private investment.

Thus, the trade balance, private saving-investment balance and the public sector balance are all treated in a somewhat ad hoc fashion, but in a way this treatment broadly respects the relative price flexibility in the Walrasian spirit. However, the previous discussion also raises the question of including dynamic considerations explicitly. In particular asset endowments, markets and expectational dynamics may need to be included. Opening up the model in this way, also carries the danger of making it less tractable. This explains why dynamic CGE models to this day are not as well developed as a reasonable theoretical critique would demand. It would seem reasonable, for example, to expect that an "ecumenical" approach could postulate the possibility of unemployment, informal labor markets, financial markets for various assets and their relation to the real sectors. Such a "realistic" model could better capture the location and dynamics of poverty among other things. Better policy analysis prospects may be an important motivation for searching for such models. However, this is beyond the scope of this paper. What can not be ignored in an exercise in poverty analysis even if it is technically limited to comparative statics is the distributional side I now turn to a consideration of distribution within CGE models leading towards the formulation of an appropriate model for South Asia.

6.2 Income Distribution, poverty and dualism:

The seminal contribution by Adelman and Robinson(1979) had used an implicit SAM to capture both factorial and household income distribution in a disaggregated manner. At about the same time the work of Lysy and Taylor(1980) focused on Brazil and made distributional aspects a part of the overall analysis. Dervis, De Melo and Robinson(1982) also addressed distributional issues in the general equilibrium modelling context. However, real concern with distribution and poverty analysis started towards the end of 1980s, after a decade of structural adjustment policies. Under the aegis of the OECD, Thorbecke (1991) for Indonesia, de Janvry, Sadoulet and Fargeix( 1991) for Ecuador, Morrison(1991) for Morocco and Chia, Wahba and Whalley for the Ivory Coast are some modelling examples from this "second generation" of CGE models for developing countries that addressed income distribution and welfare issues in greater detail than before.A number of papers by Bourguignon and others also contributed to this stream.²⁷

We can summarize the main analytical developments in modelling distribution upto this point by noting that these first and second generation models relied on a representative household assumption and fixed distributional coefficients for the household income distribution. Therefore, the analysis of poor households was necessarily coarse. No information about intra representative household income distribution and poverty was sought or used. The multiplier decomposition models of Thorbecke and Jung(1996) for poverty analysis in Indonesia and Khan(1999) for South Africa also share this weakness.

However, by utilizing the information in household income and expenditure surveys, it is now possible to generate intrahousehold groups income distribution and poverty profiles. It is also possible to use these profiles as part of the initial calibrating exercise in CGE models. A set of recent modelling efforts have been directed in precisely this direction.²⁸ Here, the paper by Decaluwé, Bernard, A. Patry, Luc Savard, and Erik Thorbecke (1999) is a pioneering piece. Another paper by Decaluwé, Dumont and Savard (1999) tests the relevance of intrahousehold distributional information for poverty analysis. Based on an archetypal economy with four areas of activity (agriculture, industry, marketable and nonmarketable services), three factors of production (capital, skilled and unskilled labor) and four types of agents(rest of the world, government, firms and households), their approach is to isolate the contribution of average income variations, poverty line changes, and income distributional changes and then to look at the effect of these variations on various poverty indicators. Their results are unambiguous. They clearly highlight the relevance and significance of intrahousehold group information. Of the three influences they discuss, the changes in poverty line in a price-endogenous model accounts for most of the changes in poverty. Therefore, both intra-household group information and price endogeneity that allows us to compute a new nominal poverty line after each policy change are important. Azis (2002) is an example of the use of this approach for analyzing poverty after the Asian financial crisis.²⁹ Another set of papers exemplified by Cogneau and Robillard (2000) and Cororaton (2003) utilizes the household expenditure survey results to carry out microsimulations. Here each household is treated effectively as an individual economic agent and its decisions are modeled directly.

Since the purpose of this paper is to see if there are "generic" models of poverty analysis within the CGE family of models applicable to South Asia, I now turn to a detailed discussion and evaluation of a generic model which is a slight modification of Stifel and Thorbecke(2003) and present the empirical results from my work on South Asia.

Download this Discussion Paper [ PDF 431.2KB| 61 pages ].

Post a Comment

We welcome your feedback on this publication. Post a comment. ADBI is not obliged to acknowledge or publish comments and may abridge or edit them before web posting.

Comment(s)

There are [0] comment(s) for this entry. Post a comment.