RDP 8901: The World Economy from 1979 to 1988: Results from the MSG2 Model 3. The MSG2 Model
April 1989
- Download the Paper 1,399KB
The MSG2 model can be described as a dynamic general equilibrium model of a multi-region world economy. In the present paper the regions modelled are the United States, Japan, Germany, the rest of the EMS (denoted REMS), the rest of the OECD economies (denoted ROECD), non-oil developing countries (LDCs), and the Oil Exporting Countries (OPEC). The model is of moderate size (about three dozen behavioral equations per industrial region). It is distinctive relative to most other global models in that it solves for a full intertemporal equilibrium in which agents have rational expectations of future variables. In theoretical conception, therefore, the model is close in design to intertemporal dynamic models of fiscal policy in Lipton and Sachs (1983) and Frenkel and Razin (1988). Those studies, like the present model, examine fiscal policy in an intertemporal perfect-foresight environment, with considerable attention given to intertemporal optimization and intertemporal budget constraints.
The MSG2 model relies heavily on the assumption that economic agents maximize intertemporal objective functions. This idea is very similar to the class of models known as Computable General Equilibrium (CGE) models[2] except that the concepts of time and dynamics are of fundamental importance in the MSG2 model. The various rigidities that are apparent in macroeconomic data are taken into account by allowing for deviations from the fully optimizing behavior. As with any modelling project that purports to describe reality, the tradeoff between theoretical rigor and empirical regularities is unavoidable.
The model has a mix of Keynesian and Classical properties by virtue of a maintained assumption of slow adjustment of nominal wages in the labor markets of the U.S., Germany, the REMS and the ROECD (Japan is treated somewhat differently, as described below).
The model is solved in a linearized form, to facilitate policy optimization exercises with the model, and especially to use linear-quadratic dynamic game theory and dynamic programming solution techniques[3]. We have experimented with the full non-linear model and found that the properties of this model correspond closely to those of the linearized model, particularly over the initial years of any shocks. The global stability of the linearized model can be readily confirmed by an analysis of the model's eigenvalues.
In fitting the model to macroeconomic data we adopt a mix of standard CGE calibration techniques and econometric time series results. In CGE models, the parameters of production and consumption decisions are determined by assuming a particular functional form for utility functions and production functions and by assuming that the data from an expenditure share matrix or an input-output table represent an equilibrium of the model. For example, if utility is assumed to be a Cobb-Douglas nesting of the consumption of different goods, then the parameters of the utility function and therefore the demand functions for different goods are given by the expenditure shares found in the data. In this example, the demand function for each good in the system will have price and income elasticities of unity. In most cases the data will determine the parameters of the model although in some cases additional econometric analysis is required. The question of calibrating the model is discussed further in McKibbin and Sachs (1988b).
The model has several attractive features which are worth highlighting. First, all stock-flow relationships are carefully observed. Budget deficits cumulate into stocks of public debt; current account deficits cumulate into net foreign investment positions; and physical investment cumulates into the capital stock. Underlying growth of Harrod-neutral productivity plus labor force growth is assumed to be 3 percent per region. Given the long-run properties of the model, the world economy settles down to the 3 percent steady-state growth path following any set of initial disturbances.
A second attractive feature is that the asset markets are efficient in the sense that asset prices are determined by a combination of intertemporal arbitrage conditions and rational expectations. By virtue of the rational expectations assumption and the partly forward-looking behavior of households and firms, the model can be used to examine the effects of anticipated future policy changes, such as the sequence of future budget deficit cuts called for by the Gramm-Rudman legislation in the U.S. Indeed, one of the difficulties of using the MSG2 model is that every simulation requires that the “entire” future sequence of anticipated policies be specified. In practice, forty year paths of policy variables, or endogenous policy rules, must be specified.
A third attractive feature of the model is the specification of the supply side. There are several noteworthy points here. First, factor input decisions are partly based on intertemporal profit maximization by firms. Labor and intermediate inputs are selected to maximize short-run profits given a stock of capital which is fixed within each period. The capital stock is adjusted according to a “Tobin's q” model of investment, derived along the lines in Hayashi (1979). Tobin's q is the shadow value of capital, and evolves according to a rational expectations forecast of future post-tax profitability.
Another point of interest regarding the supply side is the specification of the wage-price dynamics in each of the industrial regions. Extensive macroeconomic research has demonstrated important differences in the wage-price processes in the U.S., Europe, and Japan, and these differences are incorporated in the model. In particular, the U.S. and the ROECD (including Canada) are characterized by nominal wage rigidities arising from long-term nominal wage contracts based on the work of Taylor (1980). In Japan, on the contrary, nominal wages are assumed to be renegotiated on an annual, synchronized cycle, with nominal wages selected for the following year to clear the labor market on average. In the ROECD, nominal wages are assumed to be more forward looking than in the U.S., though real wages adjust slowly to clear the labour market. In Germany and the REMS we assume a form of “hysteresis” where a rise in unemployment leads to a rise in the natural rate of unemployment which persists for a substantial period of time.
Consumption is determined partly by wealth and partly by current disposable income where wealth includes human wealth which is defined as the present value of expected future after-tax labor income. This approach is consistent with the empirical evidence in Hayashi (1982) and Campbell and Mankiw (1987).
A more detailed derivation of the model can be found in McKibbin and Sachs (1988b).