A Monetary Model of the Korean Economy
Author:
Mr. Ichiro Otani
Search for other papers by Mr. Ichiro Otani in
Current site
Google Scholar
Close
and
Yung Chul Park
Search for other papers by Yung Chul Park in
Current site
Google Scholar
Close

There is a long-standing argument in the literature that in less developed countries (LDCs) the quantity theory framework is more applicable and relevant than are Keynesian effective demand theory models. A number of studies, with a varying degree of reliability, support this view. Among these, Harberger’s (1963) study on the Chilean inflation is the one most often quoted. The framework of analysis Harberger uses is a quantity theory model in which the rate of inflation is explained by the supply of money, real income, and the expected rate of inflation. In a recent study, Vogel (1974) extends Harberger’s analysis to other Latin American countries and also arrives at Harberger’s conclusion that inflation in these countries is largely a monetary phenomenon. While the Harberger and Vogel studies are interesting and throw some light on the process of inflation, they are not without defects.

Abstract

There is a long-standing argument in the literature that in less developed countries (LDCs) the quantity theory framework is more applicable and relevant than are Keynesian effective demand theory models. A number of studies, with a varying degree of reliability, support this view. Among these, Harberger’s (1963) study on the Chilean inflation is the one most often quoted. The framework of analysis Harberger uses is a quantity theory model in which the rate of inflation is explained by the supply of money, real income, and the expected rate of inflation. In a recent study, Vogel (1974) extends Harberger’s analysis to other Latin American countries and also arrives at Harberger’s conclusion that inflation in these countries is largely a monetary phenomenon. While the Harberger and Vogel studies are interesting and throw some light on the process of inflation, they are not without defects.

There is a long-standing argument in the literature that in less developed countries (LDCs) the quantity theory framework is more applicable and relevant than are Keynesian effective demand theory models. A number of studies, with a varying degree of reliability, support this view. Among these, Harberger’s (1963) study on the Chilean inflation is the one most often quoted. The framework of analysis Harberger uses is a quantity theory model in which the rate of inflation is explained by the supply of money, real income, and the expected rate of inflation. In a recent study, Vogel (1974) extends Harberger’s analysis to other Latin American countries and also arrives at Harberger’s conclusion that inflation in these countries is largely a monetary phenomenon. While the Harberger and Vogel studies are interesting and throw some light on the process of inflation, they are not without defects.

One defect is the so-called reduced form approach, which might be subject to a simultaneous equation bias as well as misspecification errors.1 Also, the authors assume that the stock of money is an exogenous variable. This assumption may be invalid in LDCs with a relatively large foreign sector, fixed exchange rates, and an underdeveloped financial sector. A third defect is that real income is determined outside the system in Harberger’s model. That is, Harberger ignores the interaction between money and real income.

The main purpose of this paper is threefold: First, to remedy various aspects of the defects inherent in a single-equation approach to an explanation of a monetary phenomenon. For this purpose, the presently fashionable monetary approach to the balance of payments seems to be a useful starting point for correcting these defects. According to the monetary approach, the supply of money is treated as endogenous because of the feedback from the balance of payments through changes in the net foreign assets position to those in the monetary liabilities of the central bank. At the same time, income is regarded as being influenced by changes in the money supply. Second, to construct a simultaneous-equation system of a quarterly monetary model of the Korean economy and to test the validity of the model against the data for the period 1962-74. Third, to analyze influences on major macroeconomic variables—including real output, prices, and balance of payments—of alternative policy instruments, by way of simulation exercises.

Section I of this paper is devoted to the specification of a quarterly model, emphasizing the relation between the monetary sector, aggregate demand and supply, and the foreign sector, and to the examination of the estimated model. In Section II, a variety of simulation exercises is performed to examine the model’s properties, such as stability, predictive ability, symmetry, and multiplier analyses; policy implications are drawn. Section III presents concluding remarks.

I. The Model

One of the most important characteristics of the model is its emphasis on the role of money in determining prices, real output, and the position of the balance of payments: the model is monetary in that money plays the central role in determining key macroeconomic variables of the economy.

The model has its origin in long-run monetary models of balance of payments developed by Polak (1957), Johnson (1972), Dornbusch (1973), and Swoboda (1974). But it differs fundamentally from these standard models and others developed by monetarists in several respects. First, it is a short-run model in which, unlike long-run models, prices and real output are determined endogenously. Second, money wages and the interest rate are assumed to be exogenously determined by institutional arrangements. Third, the model can explain the interactions between the monetary and real sectors of the economy and, consequently, can distinguish, in a limited sense, effects of the monetary impulse on prices and real output. Fourth, an explicit supply of output function is developed, based on the neoclassical optimizing behavior.

The importance of imported intermediate goods in the supply of real output is also considered in our model. As a result, the quantity of imports directly affects the supply of money and real output.

The model is formulated in the context of a small open economy in that the Korean economy faces a completely price-elastic supply schedule for its imports. However, it is assumed that the prices of its exports are determined in the domestic market rather than in the world market and that international capital flows are not perfectly mobile because of the government controls.

The model is relatively small in size, but it is rather rich in implications for stabilization policies. We can examine the effects of changes in monetary, fiscal, and exchange rate policies on the level of economic activity.2

the market for money

Demand for money: derivation of price equation

The demand for money, including savings and time deposits, is assumed to be a function of real income, the nominal rate of interest on bank savings and time deposits, which is institutionally fixed by the monetary authorities, and the expected rate of inflation.

A log-linear form of the function may then be specified as follows:

log ( T M P ) t d = a 0 + a 1 log Y t + a 2 log R t + a 3 Π t e ( 1 )
article image

It is expected that a1 > 0, a20, and a3 < 0.3

Equation (1) explains a long-run desired stock of money, which may not be satisfied at every point in time. Given equation (1), we can derive a flow demand for money as follows:

Δ log ( T M P ) t = λ [ log ( T M P ) t d log ( T M P ) t 1 ] ( 2 )

where λ is a constant coefficient of adjustment; 0 ≤ λ ≤ 1.

Assuming that the flow demand for money is always equal to the flow supply of money, and substituting equation (1) in equation (2), we obtain

Δ log ( T M P ) t = λ [ a 0 + a 1 log Y t + a 2 log R t + a 3 Π t e log ( T M P ) t 1 ] ( 3 )

It is further assumed that the lagged value of the rate of inflation can be used as a proxy for the expected annual rate of inflation. This is based on the assumption that the formation of the expected rate of inflation is static.4 As a result, equation (3) can be rewritten as follows:

Δ log ( T M P ) t = λ a 0 + λ a 1 log Y t + λ a 2 log R t + λ a 3 [ log P t 1 log P t 5 ] λ log ( T M P ) t 1 ( 4 )

Solving equation (4) for log Pt, subtracting log Pt-4 from both sides of the equation, and rearranging the terms, we have:5

[ log P t log P t 4 ] = a 0 * + a 1 * ( log T M t log T M t 1 ) + a 2 * log Y t + a 3 * log R t + a 4 * log ( T M P ) t 1 + a 5 * ( log P t 1 log P t 5 ) + a 6 * ( log P t 4 log P t 5 ) ( 5 )

In its present form, equation (5) is the price equation of the model that explains the annual rate of inflation. This function can also be interpreted as the aggregate demand function for goods and services.

The supply of money

The actual stock of money, TM, is determined by a traditional money multiplier function,

T M t = m t B t ( 6 )

where m is the actual money supply multiplier, and is defined as

m t = 1 k + ( 1 k ) ( C T M ) t
article image

The desired currency to money ratio (CTM),d is assumed to be a function of real income and the interest rates on bank deposits.

( C T M ) t d = b 0 + b 1 Y t + b 2 R t ( 7 )

where b1 < 0, b2 < 0.

Implicit in equation (7) are the assumptions that the public holds less currency in favor of savings and time deposits as income increases, and that the interest rate on deposits influences the portfolio composition of total money (broadly defined).

Since it is assumed that it will take some time for the public to adjust the currency/money ratio to a desired level, it is postulated that a change in the ratio in a given period is a fraction of the gap between the desired ratio and the actual ratio in the previous period. As a result, our hypothesis is formulated as follows:

Δ ( C T M ) t = δ [ ( C T M ) t d ( C T M ) t 1 ]
o r ( C T M ) t = δ b 0 + δ b 1 Y t + δ b 2 R t + ( 1 δ ) ( C T M ) t 1 ( 8 )

where 0 ≤ δ ≤ 1.

The supply of base money, B, is defined as the sum of the central bank’s net credit to the commercial bank (BC), its net claims on the Government (BG), and net foreign assets (NF).

B t = B C t + B G t + N F t ( 9 )

Among these, BC and BG are exogenously controlled by the monetary authorities; NF is directly related to the balance of payments position of the economy. Thus, the change in NF is defined as

Δ N F t = N F t N F t 1 = ( X P ) t ( I M e P M e ) t ( I M i + P M i ) t + F K t ( 10 )
article image

It is assumed that the Korean economy always satisfies the foreign demand for its exports (X), and that the demand is a function of the foreign income alone.6 The import prices of consumer and intermediate goods (PMc and PMi) are assumed to be determined exogenously, whereas the volumes of imports of these goods (IMc and IMi) are determined endogenously within the model. (See the demand for imports later in Section I.)

In the world of perfect capital mobility, net foreign capital flows could be determined endogenously. However, free capital mobility is far from reality in many developing economies, such as Korea, where there are a variety of controls over capital flows. In the context of the Korean economy, the government authorities have been implementing numerous control measures over capital inflows as well as both short-term and long-term outflows in order to achieve, among other things, sustained capital formation in the process of economic growth. Consequently, it is postulated that net capital flows (FK) is an exogenous variable in the short run.

the supply of output

The supply of output is divided into real output in the primary sector and the nonprimary sector. The former, consisting mostly of agricultural products, is assumed to be exogenous, whereas the latter is determined within the system and is influenced by the price level, real wages, the rate of capacity utilization, and import prices.

In the nonprimary sector, we postulate that the supply of output is subject to the following production function:7

Y m s t = A K t α N t β ( I M i ) t γ ( 11 )
article image

α, β, and γ represent the output elasticity with respect to capital stock, labor, and imported materials, respectively. The conditions for profit maximization require that the marginal product of labor is equal to real wages. That is,

Y m s N = β Y m s N = W P ( 12 )

where W = money wages, which is exogenously determined for the period. Similarly, the marginal product of the inputs of imported raw material is equal to the ratio of the import price to the output prices. As a result we have

Y m s I M i = γ Y m s I M i = P M i P ( 13 )

In reality, firms will probably not satisfy conditions (12) and (13). We are merely arguing that firms attempt to satisfy these conditions, and that these conditions are a first approximation of reality.

From equations (12) and (13), we obtain the expansion path of the economy.

β γ I M i N = W P / P M i P ( 14 )

The cost equation (CT) of the economy is defined as

C T t = ( W N ) t + ( P M i I M i ) t + C F T ( 15 )

where CF is the rental cost of capital stock, which is determined exogenously.

Then equations (11), (14), and (15) give rise to the following form of a supply function of desired output in the nonprimary sector.8

log ( Y m s ) t d = g 0 + g 1 log ( W P ) t + g 2 log ( P M i P ) t + g 3 t ( 16 )

This is a neoclassical supply function of desired real output.

Since it has been argued in the Korean context that the existing capacity of the economy is not fully utilized in every period, and since the utilization ratio depends upon such factors as changes in the availability of credit or in the stock of the real quantity of money, equation (16) can be modified9 as

log ( Y m s ) t d = g 0 + g 1 log ( W P ) t + g 2 log ( P M i P ) t + g 3 t + g 4 log ( C P ) t ( 17 )

where CP stands for capacity utilization of capital stock.10 However, actual output may not be equal to the desired output in every period. In fact, it may lag behind the desired output because it would take some time to reallocate variable inputs, such as the labor force and intermediate imports, given changes in real wages and the price of intermediate imports relative to the general price level. In order to capture the lagged response of actual output to desired output, equation (17) can be modified as follows:

Δ log ( Y m s ) t = φ [ log ( Y m s ) t d log ( Y m s ) t 1 ]
o r log ( Y m s ) t = φ g 0 + φ g 1 log ( W P ) t + φ g 2 log ( P M i P ) t + φ g 3 t + φ g 4 log ( C P ) t + ( 1 φ ) log ( Y m s ) t 1 ( 18 )

where 0 ≤ φ ≤ 1.

the demand for imports

In most LDCs, the bulk of imports consists of intermediate goods and raw materials, which are an important factor of production. Only a minor portion of total imports is for consumption. This characteristic suggests that a meaningful short-run analysis of economic fluctuations must take into consideration the relationship between imports and the supply of output.

As shown in equation (11), we assume that imported intermediate goods enter into the production function as an input. In the context of Korea, this is very important, in that more than 90 per cent of total imports have been for production of goods and services. The lack of imported materials has often created a serious supply bottleneck in the economy.

From the conditions for profit maximization, together with the modified production function (see footnote 9), we can derive a neoclassical demand function for intermediate imported goods.

log ( I M i ) t d = h 0 + h 1 log ( Y m s ) t + h 2 log ( W P M i ) t + h 3 t + h 4 log ( C P ) t ( 19 )

where h1 > 0, h2 > 0, h3 < 0, and h4 < 0.

It is further assumed that even though the world supply schedule of imports for Korea is infinitely elastic, Korea is unable to import all that it wishes within a quarter because of various factors, such as the time necessary for international transportation. Therefore, equation (19) can be rewritten as

Δ log ( I M i ) t = ψ [ log ( I M i ) t d log ( I M i ) t 1 ]
o r , log ( I M i ) t = ψ h 0 + ψ h 1 log ( Y m s ) t + ψ h 2 log ( W P M i ) t + ψ h 3 + ψ h 4 log ( C P ) t + ( 1 ψ ) log ( I M i ) t 1 ( 20 )

where 0 ≤ ψ ≤ 1.

As for the function of imports of consumer goods, we assume that it is based on a consumer demand theory, and that real income and the relative price ratio are the most important determinants of the demand.

log ( I M c ) t d = f 0 + f 1 log Y t + f 2 log ( P M c P ) t ( 21 )

where f1 > 0, f2 < 0.

As with the import function for intermediate goods, we can postulate that the actual imports of consumer goods can be presented as follows:

o r , Δ log ( I M c ) t = log ( I M c ) t = ω [ log ( I M c ) t d log ( I M c ) t 1 ] ω f 0 + ω f 1 log Y t + ω f 2 log ( P M c P ) t + ( 1 ω ) log ( I M c ) t 1 ( 22 )

where 0 < ω < 1.

Yms is the level of real output of the nonprimary sector. Since gross national product (GNP) is a value-added concept, we must subtract the real value of intermediate imports (IMiPMiP) from Yms to obtain the value added in the nonprimary sector Yn.

Y n = Y m s I M i P M i P ( 23 )

Finally, the system is closed by the income identity equation. That is,

Y = Y a + Y n ( 24 )

where Ya = real output (value-added concept) of the nonprimary sector.

the complete model and workings

Complete model

The complete model to be estimated in this paper has been modified slightly to capture special factors of some importance. For example, to account for the existence of seasonal factors, most of the behavioral equations include the seasonal dummies. In addition, NFt1PMc,t is included in the demand function for consumer goods to examine the foreign exchange constraint on imports. Obviously, in any economy the amount of foreign goods and services that it can import will be influenced, to some extent, by the availability of foreign exchange; NFt1PMc,t is a proxy representing this availability. Similarly, NFt1PMi,t is introduced in the import function for intermediate goods. In doing so, it was felt that the capacity utilization variable and the foreign exchange constraint variable in the import demand function for intermediate goods were likely to be highly related to each other. Therefore, we have decided to exclude the former variable. Similarly, to avoid the possibility of strong multicollinearity between the time trend and the output variables, t was also excluded as an explanatory variable from the import demand function for intermediate goods. This demand function includes D5, a dummy covering the period 1963-65, during which the Korean economy experienced a foreign exchange crisis. During the period under consideration, consumer imports have always been severely rationed. Over the period 1963-65, the shortage of foreign exchange was such that the government authorities found it necessary to extend their control even to imports of intermediate goods. (See Brown (1973), pp. 137-39.) Finally, as with the capacity utilization variable, the additional availability of real cash balances seems to be a reasonable proxy variable.

Therefore, log (CP)t was replaced by Δlog(TMP)t in the supply function.

A glossary of the variables appears in Table 1, and the complete model, consisting of five behavioral and five identity equations, is summarized in Table 2.

Table 1.

Korea: List of Variables

article image
Table 2.

Korea: Complete Model

article image

Workings

To elucidate the working of the model, we will consider the effects of an increase in the supply of the central bank’s credit to the Government (i.e., ΔBG > 0). This will directly increase the base money (B), which will lead to an increase in the (nominal) amount of the total money stock (TM). Following these straightforward direct effects of the increase in the central bank credit to the Government, the secondary effects will reach all endogenous variables in varying degrees. The direction of changes is difficult to predict when the coefficients of various endogenous variables are not known, because some changes in endogenous variables can be offset by feedback effects of other endogenous variables. Nonetheless, it may be worthwhile to indicate that an increase in the stock of money would tend to increase the price level initially, which, in turn, is likely to increase the imports of both intermediate and consumer goods as import prices fall relative to the general price level. As a result, real income is likely to increase, and the ratio of currency to the stock of money tends to decline. At the same time, the net foreign assets of the central bank will decline, which will offset the initial increase in base money, owing to the exogenous increase in the central bank’s credit to the Government. The chain of causations is now traced in a full circle, and further change will take place until a new equilibrium is restored again.11

The results of estimation

The model formulated in the previous section is nonlinear in variables but linear in parameters. In addition, all the behavioral equations satisfy the order condition for identifiability,12 and the two-stage least-squares method can be shown to yield consistent estimators for the parameters of the system. (See Kelejian (1971) and Amemiya (1974).) The results of the estimated behavioral equations are summarized in Table 3.

Table 3.

Korea: Regression Results of Behavioral Equations, First Quarter 1962-Fourth Quarter 1973 1

article image

The statistics for D-W are biased toward 2 for the equation that has a lagged dependent variable as an explanatory variable. Therefore, these statistics should not be taken at face value. Figures in parentheses under coefficients are t-ratios.

In the inflation equation (i)′, all the coefficients have the expected signs. The estimated coefficients indicate that the current rate of inflation is heavily dependent on the past rates of inflation, and that it tends to respond very inelastically to the changes in the current variables, such as real income, the interest rate, and the growth of money. Also, the demand schedule for goods and services is rather price inelastic (i.e., the coefficient of the income variable is −0.035) in the short run.

The import function for consumer goods is estimated to be inelastic with respect to relative prices, which is consistent with many other empirical findings for the LDCs. The coefficient of income, whose elasticity is estimated to be greater than unity, is indicative of the fact that most consumer imports consist of items that are considered luxuries in Korea. The coefficient of the lagged dependent variable suggests that the adjustment coefficient is about 0.65, which indicates that it would take about three quarters before 95 per cent of the gap between the desired level of imports in the current quarter and the actual flow in the previous quarter could be eliminated.

The estimated equation for imports of intermediate goods is also found to be elastic with respect to real output in the nonprimary sector, whose coefficient is highly significant. This finding seems reasonable, since the marginal productivity of imported inputs is typically less than unity in the short run, with the implication that one unit of output would require more than one unit of imported inputs, other things remaining unchanged. However, the elasticity of substitution between labor and intermediate imported materials is estimated to be quite low. This seems reasonable in the Korean context, since most of the labor force is more or less permanently employed once they are hired by firms.

As indicative of many developing countries, such as Korea, which has experienced varying degrees of foreign exchange constraints in the past, the coefficient of the real value of net foreign assets tended to have a consistent influence on the flow of imports, even though its magnitude was much less than the output elasticity. The adjustment coefficient of 0.94 indicates the high speed with which the demand for imports of intermediate goods is met by the actual supply of imports within one quarter. This reflects the adoption by the government authorities of flexible import control policies to allow imports of necessary intermediate goods to maintain a certain level of output, other things remaining the same.

All the estimated coefficients in the supply function have the right signs, even though the significance of the coefficients differs considerably. The supply of real output in the nonprimary sector tends to adjust inelastically in response to changes in prices of imported intermediate goods relative to the prices of domestically produced goods and real wages. This finding is consistent with an earlier observation that the substitution between the two factors of production is very low. Furthermore, the low elasticity of supply with respect to real wages tends to reflect the inelastic response of the demand for labor to these changes. The estimated adjustment coefficient of 0.3 is quite low, and it would take about eight quarters to close 95 per cent of the gap between the desired output in the current quarter and the actual level in the previous quarter.

The ratio of currency to the total stock of money is estimated to be significantly influenced by movements in real income and the interest rate, as well as seasonal factors. The estimated coefficients indicate that for the last two years of the sample period (1972 and 1973), the average elasticity of the ratio is estimated to be 0.367 with respect to real income, and 0.155 with respect to the interest rate on time and savings deposits. Therefore, real income tends to have more influence on the ratio, and, accordingly, the value of the money multiplier, than does the interest rate. The estimated coefficient of the lagged dependent variable suggests that the speed of adjustment is rather slow, the coefficient being about 0.17.

II. Simulation Exercises

The value of simulation exercises is well known. (1) They can be used to check whether a model is valid. Once the model is accepted as reasonable in reflecting the economic reality, the simulation can reveal many economic policy control properties of the model. (2) An analysis of simulation results can also show how the effect of a monetary impulse is divided between price level changes and output changes. (3) Simulation results may reveal mathematical properties of the model. (4) They may also suggest problem areas that can be improved for a better specification of the model.

In this section, a variety of simulation exercises are performed to study some of the properties of the model.13 Specific objectives of the simulation exercises are as follows. The most important is testing the validity of the model.14 For this purpose, the stability of the model is examined through the administration of a once and for all exogenous shock to the system to see whether or not the system will revert back to the original equilibrium condition that existed before the shock was given. Second, once the model is accepted as stable, the ability of the model to trace actual movements in the key macrovariables is tested by a fully dynamic simulation both within and beyond the sample period. In addition, an eight-period simulation is repeated in order to analyze projection errors inherent in the model. Once the model’s validity is confirmed, symmetrical properties of the model are analyzed in order to answer such questions as, does an increase in the interest rate on savings and time deposits have the same kind of impact on the economy, but in an opposite direction, as a decline in the interest rate? Next, the multiplier effects are studied by introducing sustained changes in the values of some policy and other exogenous variables, and implications are drawn for stabilization policies.

stability of the model

The stability of the model is extremely important when the model is used to analyze various aspects of the economy. Without ascertaining that a model is stable, any economic implications drawn from the model would lose their significance. For the purpose of examining the stability of the model, a once and for all exogenous shock is given to the system for just one period. In the first case, the value of the central bank’s credit to the Government is reduced by 10 per cent, to W 59.8 billion from the historically observed value of W 66.4 billion in the first quarter of 1970, and from the second quarter of 1970 to the fourth quarter of 1973, the historically observed values are used for simulation. All the actual values of other exogenous variables are left intact. In the second case, the import prices (of both consumer goods and intermediate goods) are raised by 10 per cent, to 1.081, from the actual value of 0.983 in the first quarter of 1970, and the values of all the import prices beyond the first quarter of 1970 and those of other exogenous variables are left intact, as in the first case.

These exercises are compared with the simulation for which all the actual values of the exogenous variables are used for the entire period, the first quarter of 1970 to the fourth quarter of 1973; such simulation has been called the base run by Zellner and Peck (1973). The results are shown in Charts 1 and 2, where the solid lines for each of the major macrovariables represent the ratio of the simulated values with once and for all shock to those of the base run. In Case I, a reduction in the central bank credit to the Government initially reduces the level of price, stock of money, real income, and imports of intermediate goods but increases the net foreign assets position as expected. After about four quarters, the system tends to return to the original equilibrium state that existed before the shock was given. All of the solid lines for the major variables show slight oscillation around the value of unity. This is perhaps due to the nature of the highly nonlinear system of the model. Nonetheless, it is clear that the system indeed returns to the initial equilibrium state; for example, the outputs of the simulation converge to those of the base run. In Case II, a visual inspection suggests that it takes a much longer time, compared with Case I, before the model returns to the original equilibrium state. This seems attributable to the much larger direct impact of import prices on the key macrovariables than that of the credit to the government sector. Thus, the model (and therefore the Korean economy) is quite sensitive to the fluctuation of import prices. The results of these exercises show that the model will eventually return to the original equilibrium with varying degrees of speed of adjustment following a disturbance in the equilibrium state owing to once and for all shock. Therefore, the model is stable.

Chart 1.

Korea: Multipliers of Major Variables (Case I)1

A06crt01
1 Effects of a 10 per cent reduction in central bank credit to the Government in the first quarter of 1970.
Chart 2.

Korea: Multipliers of Major Variables (Case II)1

A06crt02
1 Effects of a 10 per cent increase in import prices in the first quarter of 1970.

prediction by fully dynamic simulation

The question of whether any behavioral equation traces movements of its dependent variable can be answered quite easily by looking at the coefficient of determination or similar statistics for that equation. However, such a method cannot answer satisfactorily whether the model as a whole performs well unless we perform a fully dynamic simulation exercise. In order to answer this question, a simulation was performed for the period from the first quarter of 1970 through the fourth quarter of 1974, of which the entire calendar year 1974 is beyond the sample period; for 1974, some exogenous and endogenous variables are preliminary estimates rather than solid actual data. The results of these exercises are shown in Charts 3 and 4.

Chart 3.

Korea: Prices, Stock of Money, and Net Foreign Assets (Actual and Simulated)

A06crt03
Chart 4.

Korea: Money Multiplier, Real Income, and Imports of Intermediate Goods (Actual and Simulated)

A06crt04

It is clear that most of the major turning points of the selected key macrovariables are picked up quite well for the quarters within the sample period. Particularly, those of real income, imports of intermediate goods, and the money multiplier are accurately traced by the model; to a lesser degree, but still reasonably well, the model shows its tracking ability for the price level, stock of money, and net foreign assets. The results of the model’s predictive ability beyond the sample period are mixed. As far as its ability to pick up major turning points is concerned, the model still performs reasonably well for all the key variables, especially those of real income, imports of intermediate goods, money multiplier, and net foreign assets. However, the predicted and actual values show a tendency to differ considerably for all the major macrovariables. The predicted values of imports of intermediate goods are considerably less than the actual; as a result, the predicted values of the net foreign asset position, base money, and the stock of money all exceed the actual value. At the same time, the underestimation of imports of intermediate goods caused the underestimation of the predicted values of real output in the nonprimary sector, which in turn led to the underestimation of real income. All this is attributable to the underestimation of the price level, which the Korean authorities decontrolled during 1974 as a response to the oil crisis, after realizing the limitations of the price controls that had been imposed since the summer of 1972. Therefore, it appears inevitable that the model whose behavioral equations are estimated on the basis of data during a period of considerably strict price control would not be able to predict accurately the price level, and, consequently, other key macroeconomic variables for 1974. However, the unsatisfactory performance of the model in predicting beyond the sample period should be judged in light of the special circumstances mentioned earlier.

error analysis

To examine the model’s ability to trace movements in the major macroeconomic variables beyond a visual inspection, a series of simulation over the eight-quarter period has been run in successive quarters, the initial period starting with the first quarter of 1965 and ending with the fourth quarter of 1971. Throughout the simulation, the historically observed values of the exogenous variables are used. The three types of error statistics for important macrovariables are summarized in Table 4; they are compiled on the basis of the 28 available simulations over the eight-quarter period.

Table 4.

Korea: Error Analysis for the Sample Period, First Quarter 1965–Fourth Quarter 1971 1

article image

Error statistics under columns 1−8 are computed on the basis of simulations that are run from one to eight quarters ahead.

RMSE = Σi=0n(PA)2/N

SRMSE = (RMSE/Ā) ⋅ 100

Bias = p¯ − Ā

where:

A = Actuals; Ā = mean of A

P = Predicted; p¯ = mean of P

N = Number of observations

In per cent.

The root-mean-square error (RMSE) for the general price level that is one quarter ahead is about 0.06, which is about 6.6 per cent of the average price level during the sample period of the simulation. The average bias was negligible—0.008. The error declines, in the second quarter, but it starts increasing moderately thereafter until it declines dramatically in the eighth quarter. The error calculated from the simulation for the price level that is eight quarters ahead is considerably less than the error in that which is one quarter ahead. The reasons for this may perhaps be found in the model’s self-correcting properties, but they are by no means clear. Compared with the price level, the stock of money and net foreign assets is predicted with considerably larger errors in terms of the ratio of the RMSE to the average of actual value (SRMSE). This is due primarily to the errors in predicting the imports of intermediate goods as well as consumer goods; particularly, the error in the imports of consumer goods is very large. This is indicative of the extent to which the behavioral relationship of the imports of consumer goods has been subjected to artificial control measures on such imports.15 The RMSE for real income changes little as we move from a simulation that is one quarter ahead to one that is eight quarters ahead; this tends to suggest that errors in predicting some endogenous variables are partly offset by those in other variables in such a way that the prediction of real income is fairly good. The money multiplier is predicted with an error of 0.05 (the bias being less than 2 per cent) for the simulation that is one quarter ahead, but it increases gradually over time; in no case is the error doubled.

Table 5 contains the correlation coefficients between predicted errors of the jth variable and those of the kth variable. Except for the high correlation coefficients that are due to the definitional nature of direct and indirect linkage,16 the cross-correlation coefficient of errors in the predicted value of endogenous variables is in general quite small. They are all less than 0.5 and indicate that no significant variables are omitted in the model. These simulation exercises suggest that the model is generally valid, particularly for the sample period, and to a lesser extent for the quarters beyond the sample period.

Table 5.

Korea: Cross-Correlation Coefficients of Errors in Simulation That Is Run for One Quarter Ahead, First Quarter 1965–Fourth Quarter 1973 1,2

article image

See Table 1 for definitions of variables.

The cross-correlation coefficient between the jth and kth variables (rjk) is calculated by the following formula:

r j k = Σ ( e j e ¯ j ) ( e k e ¯ k ) [ Σ ( e j e ¯ j ) 2 ( e k e ¯ k ) 2 ] 1 / 2

where ej = PjAj as shown in Table 3.

properties of symmetry

It is important from an economic point of view to examine asymmetric responses of the model—see Zellner and Peck (1973). For example, does a 10 per cent increase in the interest rate on savings and time deposits have a deflationary impact to the same extent that a 10 per cent reduction has an inflationary effect? If not, to what extent does the model show its asymmetry? A similar question may be asked concerning changes in the central bank’s credit to the Government and other policy instruments.

Our experimental designs are as follows. The estimated model was subjected to fully dynamic simulation exercises for the period from the first quarter of 1970 through the fourth quarter of 1973—16 quarters—using historically observed values. The outputs of this base run are compared with outputs of other runs in which certain exogenous policy variables are changed from their actually observed values. In our case, we introduced the following changes in the policy instruments: the interest rate on savings and time deposits was increased or reduced by 10 per cent (not 10 percentage points), and the central bank’s credit to the Government was increased or decreased by 10 per cent. To give the reader an indication of the sizes of the foregoing changes, we may note that the observed interest rate declined from 22.8 percentage points in the first quarter of 1970 to 12.6 percentage points in the fourth quarter of 1973; the central bank’s credit to the Government changed from about W 66 billion to W 170 billion over the same period.

The results of the experiments are shown in Chart 5. From an inspection of Chart 5, we can see that the model’s response to the changes in the credit to the Government is remarkably symmetric. For example, upward and downward deviations of the price level from the base run values are almost identical in magnitude, and so is that of the stock of money.17 The responses of other endogenous variables, such as real income, imports of intermediate goods, net foreign assets, and the money multiplier are also shown to be symmetric to a similar degree. However, the model’s response is not so symmetric with respect to the upward and downward changes in the interest rate on savings and time deposits. All the endogenous variables, except the stock of money and the money multiplier, tend to respond more sensitively with respect to the decline in the interest rate than to the increase in the rate.

Chart 5.

Korea: Symmetrical Properties of the Model (Selected Variables)1

A06crt05

1 Ṙ = BĠ = 10 per cent; that is, + (+ BG˙) means a 10 per cent increase in R(BG), and means a 10 per cent decline in Ṙ(B).

Table 6 contains a variety of measures that substantiate the visual inspection of Chart 5. For each endogenous variable, the value of the average deviations from the base run values (AD), given a 10 per cent increase in credit to the Government, is almost identical, but with an opposite sign to that for a 10 per cent decline in credit to the Government. However, price level, real income, net foreign assets, and imports of intermediate goods (particularly, net foreign assets) respond more elastically to the decline in the interest rate than to the increase. For example, net foreign assets tended to increase, on average for the 16-quarter period, by 0.32 per cent when the rate of interest was decreased by 10 per cent, but they barely changed when the rate was increased by 10 per cent. On average, the price tends to increase by 2.5 per cent in response to a cut of 10 per cent in the interest rate, while it is estimated to decline by only 2.1 per cent in response to an increase of 10 per cent. This kind of asymmetric response of the model to the change in the interest rate can be traced to the fact that some of the endogenous variables are affected not only directly by the interest rate but also indirectly by other endogenous variables. On the other hand, the changes in the central bank’s credit to the Government (BG) will affect the endogenous variable only indirectly (except for the base money (B), which has BG as its source), and thus there is no direct effect of BG. However, it is not clear why the model’s response is symmetric when the endogenous variables are affected indirectly by the changes in BG.

Table 6.

Korea: Calculated Values of Average Deviation, Symmetry, and Distance Measures for Selected Endogenous Variables 1

article image

AD = Average proportional deviation of the endogenous variable from the base run value. The upper figure represents this deviation in response to (or ) = − 10 per cent, while the lower, to (or ) = + 10 per cent.

SYM = Measure of symmetry. A value of SYM close to zero denotes a high degree of symmetry, while a large value indicates a lack of symmetry.

DIST = Mean absolute distance from the base run value, in terms of proportional deviation.

See Zellner and Peck (1973) on the deviation from the base run value, on which SYM and DIST are calculated. In their study, the deviations are measured in the additive form, while in this paper they are measured in the multiplicative form.

multiplier analysis of fiscal, monetary, and exchange rate policies

In this section, we try to analyze multiplier effects of fiscal, monetary, and exchange rate policies and to indicate certain characteristics of the model’s response; only passing remarks are made with respect to the relative impact on major macroeconomic variables of alternative policy instruments.18 For this purpose, four types of simulation exercise are performed for the period from the first quarter of 1970 through the fourth quarter of 1973. The policy configurations are as follows:

Case A. The central bank’s credit to the Government is reduced by 10 per cent from the historically observed values.

Case B. The exchange rate is increased by 10 per cent from the actual rate.19

Case C. The average reserve requirements on deposits are increased by 10 per cent from the historical data.

Case D. The interest rates on savings and time deposits are raised by 10 per cent (not by 10 percentage points).

Case A may be regarded as the fiscal policy, and Cases C and D as the monetary policies. Case B represents the exchange rate policy.

The qualitative effects of these policy changes on major endogenous variables are tabulated in Table 7.20

Table 7.

Korea: Qualitative Effects of Changes in Exogenous Variables

article image

A reduction in the central bank’s credit to the Government has a depressing effect on the price level, stock of money, real income, imports of intermediate goods, and the money multiplier, but it raises the net foreign assets position of the central bank. An increase in the exchange rate of the won per U. S. dollar (i.e., depreciation) tends to increase the general price level but to reduce real income, the volume of imports of intermediate goods, and the money multiplier, as expected. The depreciation of the currency initially increases the local currency value of the existing net foreign assets of the central bank. Because of the price-inelastic import function for intermediate goods, however, the local currency value of imports tends to increase, offsetting the initial increase in the net foreign assets position of the central bank. However, as the income and the general price level continue to fall, so does the volume of imports, which eventually leads to a reduction in import payments. As a result, the net foreign assets position begins to rise again. The peak is reached in about three years. The stock of money tends to decline during most of the period, except at the very beginning and end, primarily because of a decline in the money multiplier associated with the fall in real incomes.

An increase in the average reserve requirements on deposits has an unequivocally negative effect on price level, real output, imports of intermediate goods, stock of money, and money multiplier, but it exerts a positive effect on the net foreign assets position.21

An increase in the interest rate on savings and time deposits unequivocally raises the stock of money and the money multiplier, but it decreases the level of prices. However, imports of intermediate goods and real income tend to increase initially but start to decline thereafter. Initially, the high interest rate tends to decrease the currency/money ratio, which in turn raises the money multiplier. As a result, the stock of money increases. It will also cause a deflationary pressure on prices, which results in an increase in real cash balances, for which a higher demand is generated by the higher level of the interest rate. At the same time, the nominal value of exports declines and import payments increase in the initial period, contributing to a decline in the net foreign assets position. The higher level of imports of intermediate goods raises the real output in the nonprimary sector and, consequently, the real income. Such a development, however, ceases in a few quarters. The declining prices tend to cut down the output, which in turn reduces the demand for imports of intermediate goods. As a result, the balance of payments situation tends to improve and the net foreign assets posisame time, the nominal value of exports declines and import payments tion will start to rise.

The relative effectiveness of the policy instruments is influenced not only by the magnitude of policy changes but also by the initial conditions and the time lags. Therefore, any empirical results must be interpreted with these elements in mind. Rather than focusing on the issue of the most effective policy instruments, it might be more useful to examine factors that affect the relative impact of the policy instruments. Table 8 indicates the various elasticities of selected macroeconomic variables with respect to the policy instruments. Such elasticities are listed with different time lags. It can be interpreted that the higher the elasticity, the more effective the policy variable would be with respect to that particular endogenous variable. In this limited sense, the exchange rate policy and the interest rate policy seem to have been equally effective in influencing the price level for the simulation period, compared with the changes in the central bank’s credit to the Government or the average reserve requirements.

Table 8.

Korea: Relative Effectiveness of Alternative Policy Instruments

article image

Real income seems to have been most sensitive to the changes in the exchange rate, compared with other policy instruments, throughout the period;22 however, its sensitivity to the other policy variables has changed considerably, depending on the lags. The net foreign assets position of the central bank is most influenced by changes in the reserve requirements for the first year, but thereafter the exchange rate policy tends to have more influence than the rest of the policy instruments.

The fiscal policy (the central bank’s deficit financing of the Government) seems to have the least effect among the policy instruments under consideration on these major macrovariables. This is so mainly because the central bank’s credit to the Government provided a relatively small source of the base money; however, when the share increases, its effectiveness tends to increase.

The relatively large influence of the exchange rate policy on the economy reflects the openness of the Korean economy, and, since the share of trade continues to increase, this policy will be even more important in influencing the economy.23

Another interesting result that has emerged from the simulation exercise is an answer to the question, at least in relation to the Korean economy, of how the monetary impulse is divided between the changes in the price level and in real output.

A sustained decrease of 10 per cent in the central bank’s credit to the Government led to a decline of 1.6 per cent in the stock of money in the fourth quarter, 1.9 per cent in the eighth quarter, 4.4 per cent in the twelfth quarter, and 6.8 per cent in the sixteenth quarter. The decline in the stock of money tends to have a depressing effect on the price level and real income (output) whose magnitude is not significantly different for the first 16 quarters, that is, the price will be affected with an elasticity of 0.12 to 0.17, and the real income (output) with an elasticity of 0.09 to 0.12, with respect to the stock of money for that period.24 However, the net foreign assets position of the central bank is estimated to be affected by a change in credit to the Government, for the first 12-quarter period, with a much higher elasticity than the price or real income. The elasticity of net foreign assets with respect to the stock of money for this period is estimated to be about − 0.7, but its sensitivity dwindles considerably by the sixteenth quarter (the elasticity being − 0.12).

These findings suggest that the impact of a sustained decrease in the central bank’s credit to the government sector on the price level, real income, and net foreign assets tends to be unique on each endogenous variable.

Using Table 8, it is quite easy to make a similar analysis of the responses of price, real income, etc., to changes in the stock of money, which are induced by other policy instruments. It is important, however, to point out that these responses are quite different from those analyzed earlier. In other words, policy-induced changes in the stock of money may have quite different impacts on price, real income, etc., depending upon policy instruments. This is so because price, real income, etc., are influenced by both the policy instruments and the stock of money, among other endogenous variables, but not by the money alone. Therefore, it is quite meaningless to discuss the effects of money on prices or income without knowing what the policy instrument is.

These analyses suggest that a combination of monetary, fiscal, and exchange rate policies with varying degree of intensities can achieve a set of major economic objectives, such as price stability, increase in output, and improvement in the net foreign assets position. The exact combination of policy packages for a given set of economic objectives must be found by iterative procedures.25

III. Concluding Remarks

The purpose of this paper is threefold. First, to construct a monetary model that incorporates interactions among the monetary sector, aggegate demand and supply, and the foreign sector by remedying various defects of a single-equation approach adopted by Harberger (1963), Vogel (1974), and Andersen and Jordan (1968). Second, to estimate the model using quarterly data from 1962 to 1973 in Korea, and to test the validity of the model. Third, to perform a number of simulation exercises in order to study the model’s properties, including its stability, predictive ability, and the role of money. From these exercises, it is hoped that some implications for stabilization policies can be drawn.

In Section I, we demonstrated that it is possible to construct such a monetary model that is consistent with the neoclassical profit-maximizing behavior of economic agents and still reflects the interaction between the monetary and real sectors in an open economy. As for the estimated model, all the estimated coefficients of the behavioral equations are shown to have the right signs from a theoretical point of view.

Simulation exercises presented in Section II indicate the following properties: (1) The model is stable, at least locally, to the extent that it returns to the original equilibrium condition, albeit at different speeds of adjustment, after once and for all exogenous shocks are administered to the system. (2) The model is shown to be reasonably valid, particularly for the sample period, but does not perform well beyond the sample period. (3) It possesses a remarkably symmetric response to the exogenous shocks given to the fiscal and monetary variables. (4) Policy simulations have shown that the model will, at least from a qualitative point of view, produce expected changes of endogenous variables in response to changes in the policy variable.

These findings suggest the following observations. First, the model seems useful for stabilization policies; a variety of simulation can be performed by changing policy variables to achieve a set of economic objectives, such as relative price stability, moderate economic growth, and balance of payments equilibrium. Second, the model’s validity can be maintained in the future if we re-estimate the model, using more recent data as they become available.

APPENDIX: Data Sources and Derivations

data sources

The data for the variables listed in Table 1 were obtained from the following sources.

  • A Bank of Korea, Economic Statistics Yearbook

  • B International Monetary Fund, International Financial Statistics

  • C Bank of Korea, Monthly Economic Statistics

derivations

Most of the data for the variables were taken directly from sources A and C; however, others were derived from data in A, B, and C, and are detailed as follows:

IMe = IM*c · ER*/PMc

where IM*c = the current value of imports of consumer goods in terms of U.S. dollars (A and C) and includes food, beverages, tobacco, and miscellaneous manufactured articles.

ER* = trade conversion factor (B)

IMi = IM*i · ER*/PMi

where IM*i = the current value of imports of intermediate goods in terms of U.S. dollars (A and C), and includes crude materials, mineral fuels, lubricants, animal and vegetable oils and fats, chemicals, and manufactured goods classified as materials, machinery, and transport equipment.

Yms = Yn + PMi · IMi/P

where Yn is real output of the nonprimary sector (in billions of won at 1970 constant prices). (A and C)

m = TM/B (A and C)

PMi = wholesale price index of imported intermediate goods. Items covered in this index include metals and metal products, machinery and machine parts, industrial chemicals, rubber, and lumber. The weight attached to this category in calculating the general import price index (PM) is 0.848. (A and C)

PMc = (PM − 0.848 · PMi)/0.152, derived from the identity PM = 0.848 PMi + 0.152 PMc

k = R/(DD + TD)

where R = the total reserve, DD = demand deposits, and TD = time deposits. (A and C)

FK = derived as a residual from the identity equation (viii) in Table 2. (A and C)

W The derivation of the wage index in the nonprimary sectors is as follows:

(1) Add “compensation of employees” and “income from unincorporated enterprises” (excluding the primary sector’s income), on annual basis (A).

(2) Using quarterly data on the nominal wage index for nonprimary sector (A and C), the quarterly distribution of the annual income is determined. For example,

Let WYt be the total wage paid in (1) for the year t.

Let Iti be the wage index for the ith quarter in the year t (i = 1, 2, 3, 4)26

The quarterly wage payment, WQti, for the ith quarter in the year t is determined by the following formulas:

W Q t i = ( W Y t ) l t i ( Σ i = 1 4 I t i )

Monthly average data, WMti, would be WQti/3.

(3) By using the average monthly data on the number of employees in the nonprimary sector, LNAti, for the ith quarter in the year t, we can derive the average monthly per capita wage earnings (Wcti), that is,

W c t i = W M t i L N A t i

(4) Dividing Wcti by the average days worked in the month for the ith quarter in the year t, we can derive the average daily per capita wage earnings (Wcdti), that is,

W c d t i = W c t i D t i

where Dti is the average number of days worked in one month for the ith quarter in the year t.

(5) Wcdti was converted into index with 1970 = 1.00, as represented by W.

REFERENCES

  • Amemiya, Takeshi, “The Nonlinear Two-Stage Least-Squares Estimator,” Journal of Econometrics, Vol. 2 (July 1974), pp. 10510.

  • Andersen, Leonall C., and Jerry L. Jordan, “Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization,” Federal Reserve Bank of St. Louis, Review, Vol. 50 (November 1968), pp. 1123.

    • Search Google Scholar
    • Export Citation
  • Bardhan, Pranab K., Economic Growth, Development, and Foreign Trade—A Study in Pure Theory (New York, 1970).

  • Brown, Gilbert T., Korean Pricing Policies and Economic Development in the 1960s (Johns Hopkins University Press, 1973).

  • Dornbusch, Rudiger, “Devaluation, Money, and Nontraded Goods,” American Economic Review, Vol. 63 (December 1973), pp. 87180.

  • Edgerton, David L., “Some Properties of Two Stage Least Squares as Applied to Nonlinear Models,” International Economic Review, Vol. 13 (February 1972), pp. 2632.

    • Search Google Scholar
    • Export Citation
  • Evans, Michael K., Macroeconomic Activity: Theory, Forecasting, and Control—An Econometric Approach (New York, 1969).

  • Fisher, G.R., and D.K. Sheppard, “Interrelationships Between Real and Monetary Variables: Some Evidence from Recent U.S. Empirical Studies,” in Issues in Monetary Economics: Proceedings of the 1972 Money Study Group Conference, ed. by H. G. Johnson and A. R. Nobay (Oxford University Press, 1974), pp. 179259.

    • Search Google Scholar
    • Export Citation
  • Harberger, Arnold C., “The Dynamics of Inflation in Chile,” in Measurement in Economics: Studies in Mathematical Economics and Econometrics in Memory of Yehuda Grunfeld (Stanford University Press, 1963), pp. 21950.

    • Search Google Scholar
    • Export Citation
  • Howrey, Philip, and H.H. Kelejian, “Simulation Versus Analytical Solutions,” in The Design of Computer Simulation Experiments, ed. by Thomas H. Naylor (Duke University Press, 1969), pp. 20731.

    • Search Google Scholar
    • Export Citation
  • Johnson, Harry G., “The Monetary Approach to Balance-of-Payments Theory,” Journal of Financial and Quantitative Analysis, Papers and Proceedings, Vol. 4 (March 1972), pp. 155572.

    • Search Google Scholar
    • Export Citation
  • Kelejian, Harry H., “Two-Stage Least Squares and Econometric Systems Linear in Parameters but Nonlinear in the Endogenous Variables,” Journal of the American Statistical Association, Vol. 66 (June 1971), pp. 37374.

    • Search Google Scholar
    • Export Citation
  • Naylor, Thomas H. (ed.), Computer Simulation Experiments with Models of Economic Systems (New York, 1971).

  • Naylor, Thomas H., and J.M. Finger, “Validation,” in Computer Simulation Experiments with Models of Economic Systems, ed. by Thomas H. Naylor (New York, 1971), pp. 15364.

    • Search Google Scholar
    • Export Citation
  • Otani, Ichiro, “Inflation in an Open Economy: A Case Study of the Philippines,” Staff Papers, Vol. 22 (November 1975), pp. 75074.

    • Search Google Scholar
    • Export Citation
  • Polak, J.J., “Monetary Analysis of Income Formation and Payments Problems,” Staff Papers, Vol. 6 (November 1957), pp. 150.

  • Silveira, Antonio M., “The Demand for Money: The Evidence from The Brazilian Economy,” Journal of Money, Credit and Banking, Vol. 5 (February 1973, Part 1), pp. 11340.

    • Search Google Scholar
    • Export Citation
  • Sowey, Eric R., “Stochastic Simulation of Macroeconometric Models: Methodology and Interpretation,” in Econometric Studies of Macro and Monetary Relations (Papers presented at the second Australasian Conference of Econometricians held at Monash University, August 9-13, 1971), ed. by Alan A. Powell and Ross A. Williams (Amsterdam, 1973), pp. 195230.

    • Search Google Scholar
    • Export Citation
  • Swoboda, Alexander K., “Monetary Policy Under Fixed Exchange Rates: Effectiveness, the Speed of Adjustment, and Proper Use,” in Issues in Monetary Economics: Proceedings of the 1972 Money Study Group Conference, ed. by H.G. Johnson and A.R. Nobay (Oxford University Press, 1974), pp. 5274.

    • Search Google Scholar
    • Export Citation
  • Toyoda, Toshihisa, “Price Expectations and the Short-Run and Long-Run Phillips Curves in Japan, 1956-1968,” Review of Economics and Statistics, Vol. 54 (August 1972), pp. 26774.

    • Search Google Scholar
    • Export Citation
  • Vogel, Robert C., “The Dynamics of Inflation in Latin America, 1950-1969,” American Economic Review, Vol. 64 (March 1974), pp. 10214.

    • Search Google Scholar
    • Export Citation
  • Zellner, Arnold, and Stephen C. Peck, “Simulation Experiments with a Quarterly Macroeconometric Model of the U.S. Economy,” in Econometric Studies of Macro and Monetary Relations (Papers presented at the second Australasian Conference of Econometricians held at Monash University, August 9-13, 1971), ed. by Alan A. Powell and Ross A. Williams (Amsterdam, 1973), pp. 15068.

    • Search Google Scholar
    • Export Citation
*

Mr. Otani, economist in the Asian Department, is a graduate of the University of California at Berkeley and the University of Minnesota.

Mr. Park, a graduate of Seoul National University and the University of Minnesota, was an economist in the Financial Studies Division of the Fund’s Research Department when this paper was prepared. He now teaches at Korea University, Seoul, Korea.

An earlier version of the paper was presented at the winter meetings of the Econometric Society, held in Dallas, Texas, in December 1975.

1

The expression “reduced form approach,” which is often used in studies of monetary economics, is misleading, since it does not usually refer to the approach of estimating a behavioral equation in terms of all the exogenous variables that appear in the system of equations. Rather, it refers to the estimation of a single equation in terms of a subset of these exogenous variables. Our criticism refers to this kind of a single-equation approach.

2

This model may meet the challenge posed by Fisher and Sheppard (1974, pp. 236 and 223), who argued very convincingly that “the FRB-St.L. approach as typified in the work of Andersen and Jordan [1968], is based on two econometric heresies: disregard for structure and the spurning of disaggregation” and that “large econometric models … typically … violate almost every tenet of econometric theory.” They also criticize large models for not being consistently formulated among different blocks of equations.

3

On a purely theoretical level, a2 is negative or positive depending upon the elasticity of the demand for real narrow money, and of the demand for real savings and time deposits, with respect to the interest rate (R). Therefore, the sign of a2 is an empirical question.

4

That is, the adjustment coefficient of the adaptive expectation hypothesis is unity. The validity of this assumption and empirical support can be found in Toyoda (1972), Silveira (1973), and Otani (1975), where the rate of inflation is calculated at an annual rate.

5

a1*=1, a4*=λ, and a0*=1.

6

This assumption implies that the price elasticity of the foreign demand for the Korean exports is zero and that there are no close substitutes for them in the world market. Such assumptions may be a bit strong, but we reluctantly made these assumptions to avoid a considerable degree of complications in the model that may arise from alternative specification to allow the existence of close substitutes for the Korean exports in the world market.

7

Since the manufacturing sector of the Korean economy engages in the production of items that contain substantial amounts of imported raw materials and other intermediate goods, it is essential to treat these imports as a factor of production. See Bardhan (1970), ch. 4, for theoretical treatment of this issue.

8

Solving equations (14) and (15) for N and IMi, we attain

N = (CTCF)/W ⋅ (1 + γ/β)

IMi = (CTCF)/PMi ⋅ (1 + β/γ)

Substitution of these values in the production function—that is, equation (11)—gives rise to

Y m s = A K 0 e α g t [ ( C T C F ) W ( 1 + γ / β ) ] β [ ( C T C F ) P M i ( 1 + β / γ ) ] γ

Solving this expression for CT in terms of Yms and the parameters, the total cost function is obtained. Differentiate CT with respect to Yms and set it equal to the price of output, P. The resultant expression becomes a supply function for Yms, which has the form

log Yms = f(log P, log W, log PMi, t)

where f1 > 0, f2 < 0, f3 < 0, and f4 > 0;

fi being the partial derivative with respect to the zth argument. Since the foregoing formulation gives rise to the problem of multicollinearity, the final expression contains real wages and import prices relative to the domestic prices as shown in equation (16).

9

See also Evans (1969), ch. 10. His arguments can be summarized as follows. A neoclassical type of supply function is not suitable for explaining the short-term fluctuations in output. This happens primarily because firms do not immediately adjust their short-run position to the one suggested by static microeconomic theory. Equation (17) can be obtained by modifying the production function—that is, equation (11)—in the following form:

Yms = A(K ⋅ CP)α ⋅ Nβ ⋅ (IMi)γ

10

It can be shown that the coefficient of log (CP)—that is, g4—is expected to be positive.

11

An empirical counterpart of this example is provided later.

12

See Edgerton (1972) for this condition in a nonlinear system.

13

Howrey and Kelejian (1969) argue that when a model contains analytical solutions concerning its properties, simulation exercises would not provide additional information, and that simulation exercises should be used only when analytical techniques are not available for obtaining solutions. But Naylor (1971, p. 299) argues that there may be many cases in which “an economist may conclude that the model may not have a known analytical solution. However, it may very well be that if the economist made a thorough search of the literature in mathematics or consulted with a mathematician, it might be possible to find an analytical solution to the model.” Naylor contends that such a search would not be worth the effort.

14

“To validate any kind of model (for example, economic models) means to prove the model to be true. But to prove that a model is ‘true’ implies (1) that we have established a set of criteria for differentiating between the models that are ‘true’ and the models that are ‘not true,’ and (2) that we have the ability to readily apply these criteria to any given model” (Naylor and Finger, 1971, pp. 153-54). “By validation … is understood simply determining whether the model fulfils well the demands made of it. It is not a question whether the model embodies strictly causal mechanisms, but rather whether the estimated model, with all its inherent imperfections, does an adequate job of prediction, both within and beyond the estimation period” (Sowey, 1973, pp. 195-96). These authors define “validation” quite differently; in fact, Naylor and Finger emphasize much stricter conditions than does Sowey. They argue that the model must have a high degree of “goodness of fit” and that the assumptions upon which it rests must be valid. This section considers the first point of their argument; Section I has already considered the second point.

15

Some of these controls were adjusted by the introduction of a proxy for the foreign exchange limitation or dummy variable for imports of intermediate goods, but they are evidently not sufficient to capture all the effects of controls.

16

That is, the direct linkage between NF and B; the indirect linkage between TM and NF(B); IMi and NF(B); IMc and NF(B); Y and Yms; C/TM and m.

17

Deviation from the output of the base run is measured in percentage terms.

18

We refrain from making any judgments concerning which policy in a global sense would be most effective, since effectiveness would vary according to the magnitude of the policy package, the range in which the magnitude can change, and the initial condition.

19

This is done by raising the actual values of import prices by 10 per cent for the period of simulation. In addition, the net foreign assets position of the central bank at the beginning of the simulation period is also increased by 10 per cent to reflect the increased value in terms of the local currency.

20

In the remainder of this section, an increase (+) or a decrease (−) refers to the position of the simulation results relative to the output of the “base run.”

21

The increase in the net foreign assets position results in an increase in the base money; however, because of a decline in the money multiplier, the stock of money decreases.

22

Some explanation for the negative elasticity of real income with respect to the exchange rate changes is in order. In this model, there is no automatic mechanism by which the depreciation of the currency affects the exports. The exports are determined by the demand condition of the rest of the world. Only to the extent that the depreciation induces changes in the general price level does it give rise to an increase in export receipts in local currency. Therefore, the effects of the depreciation on the economy are felt only through changes in imports and net foreign assets.

23

It can also be said that the Korean economy tends to be influenced even more by changes in import prices.

24

Elasticity of a variable with respect to the stock of money can be obtained by dividing the elasticities of the variable with respect to policy instruments by the elasticities of the stock of money with respect to the corresponding policy instruments.

25

There is no guarantee that the four policy instruments are sufficient to reach the four, or even three, policy targets in this nonlinear system. If we find that the policy goals are not achieved by these instruments for one reason or another (say, political constraint), additional policy instruments must be included. For example, wage rate and (net) foreign capital inflow would be good candidates for additional instruments in our model.

26

This index alone is not a good indicator over a long period of time, but it seems sufficient for the purpose of determining general fluctuations within one year.

  • Collapse
  • Expand