Robinson: Multisectoral Models

Robinson, Sherman. (1988). "Multisectoral Models". In Hollis Burnley Chenery, T. N. Srinivasan & Jere R. Behrman (Eds.), Handbook of development economics (pp. 3 v. in 4). Amsterdam ; New YorkNew York, N.Y., U.S.A.: North-Holland ; Sole distributors for the U.S.A. and Canada, Elsevier Science Pub. Co.

There are three main ways that Robinson identifies that can be used to classify different CGE models. The first is by highlighting the mathematical structure of the model. The second is by what policy is the focus of the model. The third is by the theoretical foundation of the model (886).

The author firstly explains how CGE models grew out of linear modeling work completed before the 1970s. “In the early 1970s, work was started on a new type of non-linear, multisectoral model which sought to simulate the workings of a market economy, solving for both market prices and quantities simultaneously. These computable general equilibrium (CGE) models can be seen as a natural outgrowth of input-output and LP models, adding neoclassical substitutability in production and demand, as well as an explicit system of market prices and a complete specification of the income flows in the economy. The first developing country application was Adelman and Robinson (1978), which started an extensive literature” (888).

Robinson then goes on to identify four important solution techniques that were developed in the late 70s.

“One can classify models along a continuum running from analytic to applied. Analytic models are designed to explore the implications of various sets of theoretical postulates, with as few assumptions as possible about the magnitudes of parameters” (891). “Typically, stylized numerical models are more complex than analytic models, since wider applicability is desired and simplicity is no longer required. Stylized numerical models nonetheless tend to stay close to their underlying analytic models. Since the goal is to explore particular causal mechanisms, simplicity is desirable. Stylized models are still a long way from models which seek to capture in a realistic way the variety of important effects that might impinge on a particular policy problem facing a particular country” (892). “Applied models differ from stylized models in two important ways. First, they broaden further the range of stylized facts that are incorporated. Second, they seek to capture important features of a particular economy or situation. For example, while a styli9zed model can represent a number of similar countries…an applied model would be built for a particular country. By including more institutional detail, applied models are more specific and narrow” (892).

“In general, analytic models do not yield specific policy recommendations” (893).

“While multisector models applied to developing countries are Walrasian and neoclassical in spirit, most modelers quickly abandoned many of the strong assumptions of neoclassical theory when faced with the problem of capturing the stylized facts characterizing these economies. The assumptions of perfect competition, perfectly functioning markets with flexible prices, and free mobility of products and factors are not sustainable in actual economies. Instead, modelers have incorporated a variety of ‘structuralist’ rigidities into their models that seek to capture non-neoclassical behavioral relations, macro imbalances, and institutional rigidities characteristic of developing countries” (894).

“One problem is that applied modelers often seek to draw on strands of theory outside the paradigm of Arrow-Debreu general equilibrium theory. The concept of equilibrium imbedded in the neo classical general equilibrium model underlying all multisector models its that of flow equilibrium in product and factor markets. There are additional equilibrium concepts that one might want to capture in a model, reflecting different underlying analytical theories. A second concept is that of equilibrium in aggregate ‘financial’ or ‘nominal’ flows, which define3d a notion of macro equilibrium—the heart of Keynesian macroeconomics. A third concept is that of equilibrium in asset markets, defining another form of macro equilibrium. Fourth, there is inter-temporal equilibrium involving expectations—adaptive, rational, consistent, or whatever—in an explicitly dynamic framework. The four equilibrium concepts are not independent, and it is an open question how adequate are theoretical and empirical models that only include one or two of them.” (895).

At the end of 895-6, Robinson addresses the issue of non-Walrasian models and equilibrium seeking components. He concludes that they may be just as equilibrium seeking, but that there is just a different equilibrium being sought that is no more or less rigorous than the Walrasian alternative.

“The CGE framework requires a complete specification of both the supply and demand sides of all markets, including gall the nominal magnitudes in the circular flow” (906). On page 907, Robinson outlines the simplest structure of a Walrasian CGE. It starts, at its most basic level, with producers and households. These two actors then are assumed to maximize utility or profits. They respond to signals, in the most basic case, to prices. Finally, the structure of the interaction of the units must be specified. This can become much more complex as models grow, and other actors can be imposed.

This, however, does not fully make a CGE. “With the specification of the agents, their motivation and the institutional constraints under which they interact, a general equilibrium model is still not completely determined. One must also define ‘equilibrium conditions’ which are ‘system constraints’ that must be satisfied, but that are not taken into account by any agent in making his decisions. Formally, an equilibrium can be defined as a set of signals such that the resulting decisions of all agents jointly satisfy the system constraints. The signals represent the equilibrating variables of the model. For example, a market equilibrium in a competitive model is defined as a set of prices and associated quantities such that all excess demands are zero. In a market economy, prices are the equilibrating variables that vary to achieve market clearing.

“As discussed earlier, the definition of equilibrium conditions is a fundamental property of a model. The specification of equilibrating variables and of system constraints that characterize and equilibrium can be seen as a simplifying device that providers a way to describe the results of the workings of an actual economy. For example, instead of specifying prices as equilibrating variables to achieve market clearing, one could instead try to model price determination explicitly, specifying ‘disequilibrium’ price adjustment rules to describe how prices change over time. Such a specification is theoretically very difficult to implement—indeed, event to define—and completely unnecessary if one is willing to accept the market-clearing system constraints under flexible prices as a reasonable description of the final result of such a process within the time period described by the model” (908).

A neoclassical, closed-economy CGE is described.

“Within the framework of the CGE model, one can distinguish three kinds of structuralist models” (913). These different structural aspects, or additions, to traditional neoclassical CGE models are to correct for variance and error. The first remains theoretically close to neoclassical modeling. In this structural addition, there are substitution elasticities that are imposed and is termed “elasticity structuralist” (914). Secondly, there is the “micro-structuralist” version of the model which, “assumes that various markets do not work properly or are not present at all. Instead, there are assumed to be restrictions on factor mobility, rigid prices, rationing and neoclassical disequilibrium in one or more important markets” (914). Thirdly, there is the “macro-structural” model (914). This approach emphasizes, “…achieving equilibrium among various macro aggregates; in particular, savings and investment, exports and imports and government expenditure and revenue” (914).

UPDATE:

“One problem is that applied modellers often seek to draw on strands of theory outside the paradigm of Arrow-Debreu general equilibrium theory. The concept of equilibrium imbedded in the neoclassical general equilibrium model underlying all multisectoral models is that of flow equilibrium in product and factor markets. There are additional equilibrium concepts that one might want to capture in a model, reflecting different underlying analytical theories. A second concept is that of equilibrium in aggregate ‘financial’ or ‘nominal’ flows, which defines a notion of macro equilibrium—the heart of Keynesian macroeconomics. A third concept is that of equilibrium in asset markets, defining another form of macro equilibrium. Fourth, there is intertemporal equilibrium involving expectations—adaptive, rational, consistent, or whatever—in an explicitly dynamic framework. The four equilibrium are not independent, and it is an open question how adequate are theoretical and empirical models that only include one or two of them” (895).

“There is as yet no acceptable reconciliation of micro and macro theory, and the Walrasian model is an uneasy host for incorporating macro phenomena” (896).