Most developing countries have at one time or another faced the twin problems of a high domestic rate of inflation and a deficit in the balance of payments. The cause of these problems can often be traced to a situation of government fiscal deficits that result in excessive monetary expansion and feed domestic demand. Stabilization programs are typically put into effect to reduce these pressures. Policymakers have long recognized that the implementation of a stabilization program will have simultaneous effects on output, inflation, and the balance of payments. While practitioners generally attempt to make allowances for these effects in qualitative terms, relatively little is known about the precise quantitative nature of the relationships among these major economic aggregates in the context of developing countries. This lack of knowledge, of course, creates considerable problems when one wishes to assess the effects of a particular policy initiative—say, for example, a change in monetary policy—on important macroeconomic variables, and, conversely, to derive the appropriate policy to achieve specific stabilization objectives.
Most economists would probably accept the general proposition that monetary expansion will not only create inflationary pressures and cause the balance of payments to deteriorate but also, particularly when it is unanticipated, increase real income. However, the actual operation of the transmission mechanism in developing countries, and the relative size and timing of the effects of a change in policy, are a matter of considerable doubt. There is now an extensive literature on some of the more important individual relationships. For example, substantial work has been done on the links between monetary growth and inflation,1 and between money and the balance of payments.2 A relatively smaller number of empirical studies have also been made of the relationship between money and growth.3 Systematic analysis of these various relationships across developing countries, however, is very limited. Models have been specified and estimated for some individual countries—for example, by Aghevli (1977) for Indonesia, Khan (1976) for Venezuela, and Otani and Park (1976) for Korea—but it is difficult to generalize these results to other countries because of differences in model specification, estimation methods, and periods of estimation.4
Furthermore, there has been only limited formal analysis of the effects that a stabilization program may have on a developing economy. This, of course, has considerable implications for the Fund’s lending operations, since the relationships just mentioned figure prominently in the stabilization programs that are established by member countries in connection with stand-by arrangements. Broadly defined, a stabilization (or financial) program is a package of policies designed to eliminate disequilibrium between aggregate demand and supply in the economy, which typically manifests itself in balance of payments deficits and rising prices. Thus, the fundamental objective of a financial program is to find “a suitable relationship between resource availabilities and needs that causes minimum strain on the internal price level and produces a desired balance of payments result.”5
While no single theoretical model underlies all financial programs, a broad framework within which most of them are formulated has evolved in the IMF over the years.6 In this framework there is a fairly well-defined relationship between money, the balance of payments, and domestic prices, in which the supply of and demand for money play a central linking role. The effects of policies on the real sector are treated less explicitly. When feedbacks from real output are taken into consideration, the analysis is made on a more informal case-by-case basis rather than in the context of an explicit and consistent general methodology. Because of this, it is difficult to say a priori whether a given financial program will have undesirable consequences for growth and employment, something that has worried policymakers and academic economists alike.7 For example, as Robichek (1967, p. 4) has observed, in implementing a financial program one has to be aware of “the need to frame programs that are compatible with aspirations for rapid economic growth.”
Generally, treating output as independent of monetary or credit factors has been rationalized on the grounds that the typical program is essentially short term and that in the short run one can assume that the domestic supply of resources is effectively fixed. This assumption has apparently been supported by some empirical evidence. For example, Reichmann and Stillson (1978), after examining a number of Fund stand-by arrangements over the period 1963-72, conclude (p. 304) that “one cannot observe a systematic relationship between the introduction or implementation of programs and rates of growth in the short run.” This conclusion, however, does not distinguish between the rate of growth of capacity output and the rate of capacity utilization, and the latter could be affected even in the short run. Furthermore, in view of the extension of the time span of stand-by arrangements that has occurred in more recent times, this rationale can no longer be relied upon, and anyone engaged in setting up a financial program has to face the growth problem directly.
While the principal objectives of a stabilization program may be to reduce the inflation rate and improve the country’s external payments position, the policies followed will have repercussions on the other variables in the economy, some of which may be desirable and some not. Yet these other effects are clearly of interest to the policymaker. For example, a restrictive monetary policy may involve an unwanted temporary loss of output, even though eventually such a policy would reduce inflation and improve the balance of payments. The size of this loss of output, and the period over which it occurs, is often of overriding importance to the authorities in developing countries from both economic and social perspectives. One can easily conjure up a scenario in which a deflationary policy initially has little effect on prices, perhaps because they are “sticky” in the short run,8 but causes a substantial slowdown in the growth of real income and employment. Such a scenario would depend on the relative sizes of the various parameters in the system and the time lags involved in the response of variables to policy changes. The question that arises is whether there are combinations of private sector behavior and government policy that could potentially give rise to this result, and how likely it is to occur.
Handling this type of problem would naturally require a dynamic model that could simultaneously capture the major relations between prices, the balance of payments, and output. The purpose of this paper is to propose a formal framework for examining these interrelations, and, more important, to use this framework to analyze the effects of policy changes on all these variables. In a sense, the model can be viewed as providing one particular interpretation of the basic theoretical paradigm underlying the financial programming exercises of the Fund. Within the context of the simple dynamic model developed here, questions regarding the effect of stabilization programs on resource utilization in both the short run and the long run can be handled. Such an analysis is important, since the time path that the variables follow during the course of a stabilization program is often just as crucial as the positions they ultimately reach.
The model proposed here, although highly aggregated and simple in structure, is nevertheless able to meet this main requirement. It stresses the crucial role played by the demand for money and monetary disequilibrium in the behavior of such major macroeconomic variables as prices, output, and the balance of payments. Thus, the analysis can be considered a generalization of the models developed in the context of the monetary approach to the balance of payments,9 and is also consistent with the philosophy underlying the financial programming approaches followed in the Fund.10 Nevertheless, while monetary factors are assigned a dominant role, it is explicitly recognized that the money supply is not necessarily under the close control of the authorities. The domestic money stock definitionally equals the net foreign assets of the consolidated banking system plus bank credit to the government and the private sector, and it is by now well established that in a small open economy operating under a system of fixed exchange rates (a description that would fit most developing countries), changes in the money supply can be brought about through the balance of payments. Furthermore, in countries that lack a developed capital market, the growth of domestic credit may be closely linked to the government’s borrowing requirements and hence to its fiscal policy. In this model, monetary (cum fiscal) policy is the relevant means by which the authorities seek to achieve their objectives, and it is the domestic component of the money stock that is the instrument to be used to this end.
This stress on monetary factors and short-run dynamics obviously gives an unmistakably demand-oriented flavor to the model. No account is taken of wealth and capital accumulation and their role in affecting the long-run growth path of the economy. Essentially, the model attempts to capture only short-term deviations of real income from its long-run trend, which is here assumed to be exogenously determined. However, the analysis could be extended by specifying the determinants of capacity output.
As the focus of the exercise is to derive meaningful policy conclusions, the model is estimated using a systems method in order to provide parameters that can be used to quantify the various relationships. Through the use of pooled time-series cross-sectional data covering 29 developing countries, an attempt is made to achieve sufficient generalization that the estimated parameters can be taken as representative of a broad class of developing countries. Finally, and this is perhaps the most important part of the paper, a set of simulation experiments is conducted using the estimated version of the model. In the first simulations, the effects of certain shocks are traced, including a change in monetary policy and an exogenous increase in capacity output. The results of these simulations are compared with those obtained from a model that does not allow for simultaneous interaction between output, inflation, and the balance of payments to highlight the differences that emerge when such interaction is taken into account. In the second set of simulations, an attempt is made to derive the appropriate financial policy to achieve a specified improvement in the balance of payments. This type of simulation is, of course, similar to the typical financial programming exercise. Indeed, one of the objectives of this part of the paper is to compare the price and output consequences of two types of stabilization program: a standard short-term program, and a longer-run or “extended” program.
Section I describes the structure of the model. The results from estimating the model, and certain of its overall empirical characteristics, are discussed in Section II. The various simulation experiments are considered in Section III, and a summary of the findings of the study is presented in Section IV.
I. Specification of the Model
The model contains nine equations, of which six are behavioral. Such simplicity was dictated mainly by a desire to focus on the general aspects of the issues considered here and to develop an analysis that is applicable to a variety of developing countries. Any greater detail would necessitate consideration of the institutional characteristics of individual countries, and thus make it difficult to generalize from the results.11 Furthermore, any attempt to construct a more disaggregated model for developing countries immediately runs into the constraints of the limited availability of data. Essentially, this model describes an economy that is small relative to the rest of the world. It is open to international trade and financial flows, and maintains a pegged exchange rate.12 It is assumed that the domestic financial sector is relatively underdeveloped. This specifically implies that the number of financial assets that could substitute for money holdings is very limited, and/or that the authorities control the interest rates on those assets that are available.13 Naturally, not all developing countries would fit exactly into the framework given by these assumptions, but it is argued that these features seem to be characteristic of most developing countries.
The stochastic equations of the model explain inflation, the overall balance of payments, the fiscal budget (that is, government expenditure and revenues), and output. Identities for the supply of money and its domestic component (domestic credit) are used to close the system.
Inflation
The specification for price changes is an extension of the monetary disequilibrium model of Goldman (1972) to an open economy. The domestic rate of inflation, relative to the foreign rate, is assumed to be positively related to the excess supply of real money balances, and a negative function of the deviation of domestic prices from their equilibrium (purchasing power parity) level. Formally, the specification is written as follows:
where
-
P = domestic price level
-
ϵ = exchange rate, in units of domestic currency per unit of foreign currency
-
Pf= foreign price level
-
m = stock of real money balances, that is, the nominal stock of money, M, deflated by the domestic price level
-
λ1 = constant (reflecting the steady-state properties of the system).14
The superscript d denotes demand, and Δ is a difference operator, so that Δlog P is the domestic rate of inflation, and Δlog ϵ and ΔlogPf are the proportionate rates of change in the exchange rate and the foreign price level, respectively.
On the simplifying assumption that a country’s equilibrium real exchange rate is not changing secularly, β0 may be taken as a parameter rather than varying over time.15 If there is no excess demand for real money balances and domestic prices are equal to their equilibrium level μ0, then, with the exchange rate fixed, domestic inflation will be equal to the rate of inflation prevailing in the rest of the world.16 In other words, domestic price setters always attempt to keep their prices in line with those charged in foreign markets. Divergences from this equilibrium relationship can arise from two sources. First, any expansion of the money stock that results in an excess supply of real money balances will (in the next period) create inflationary pressures that tend to eliminate the disequilibrium in the money market. Second, if domestic prices are pushed away from their equilibrium level, for whatever reason, they will move in the direction that restores the relationship. In a sense, this second term in equation (1) represents a type of “catch-up” effect to any erosion that may occur in a country’s international competitiveness.17 While the preceding interpretation of the inflation equation is somewhat heuristic, it is possible to derive it from a theoretical model involving traded and nontraded goods. This derivation is shown in Appendix II.
Feeding into equation (1) is the stock demand for real money balances.
Here we follow the standard literature in relating money demand to real income (y) and to the expected rate of inflation (Π).
where both parameters are expected to be positive.
This formulation, which is typically used for developing countries,18 differs from theoretical models in excluding the rates of interest on other financial assets from affecting money demand.19 This follows directly from the assumption above regarding the paucity of financial alternatives to money in developing countries.20 The relevant substitution in such countries is therefore between money and goods, or real assets, with the opportunity cost being the expected rate of inflation.
Balance of Payments
The overall balance of payments, as represented by the proportionate change in the stock of international reserves (in terms of domestic currency), is specified as a positive function of the excess demand for nominal money balances and a negative function of the deviation of the domestic price level from its purchasing power parity equilibrium
where
-
R = net stock of international reserves, and
-
M = nominal stock of money.
The other variables are defined as before.
In equation (4), variations in the domestic currency value of foreign exchange reserves that are due solely to exchange rate movements are eliminated by subtracting the percentage change in the exchange rate from the left-hand side of the equation.21 Equation (4) is a dynamic version of models in the tradition of the monetary approach to the balance of payments and, following that literature, it does not distinguish between the current and capital accounts of the balance of payments. It makes no prediction as to whether domestic residents rid themselves of excess money balances by increasing expenditure (that is, absorption) relative to output, or by purchasing financial assets abroad. Laidler and O’Shea (1980) note that even the second term, which says that the balance of payments will deteriorate when domestic prices rise relative to foreign prices, does not reflect current account factors alone, since such a decline in a country’s competitive position may induce domestic asset holders to export capital on the expectation that the probability of a future devaluation of the (fixed) exchange rate has increased. Thus, the present treatment of the overall balance of payments in a single equation is consistent with our neglect of domestic financial markets. Furthermore, since many developing countries impose various restrictions on current and capital transactions, it would be difficult to deal empirically with this distinction in a cross-country framework without taking a large number of country-specific factors into account. The balance of payments adjustment equation, therefore, has the virtue of simplicity as well as generality.
Most empirical applications of the monetary approach to the balance of payments assume that the change in a country’s international reserves is exactly equal to the difference between the flow demand for money and the flow supply of domestically created money. This standard assumption does not seem very realistic in the context of developing countries, where the degree of international mobility of goods and assets may not be sufficient to allow an excess supply of money to be offset fully and instantaneously by balance of payments leakages. The equation that is specified here for international reserves is consistent with the broad framework of the monetary approach, but it includes a degree of dynamic adjustment as measured by the parameter γ6. Thus, it allows for inertia in the response of reserve flows to monetary disequilibrium in the short run, while still retaining the feature that the effect of an expansion in domestic credit on the money stock is completely offset in the long run.
Substituting for the nominal demand for money gives22
Government Sector
Fiscal policy and the government’s budgetary position are modeled explicitly because of the crucial role that they play in the money supply process and in overall economic activity in developing countries. In most cases, excess demand in the economy can be traced back to the deficits of the public sector, and consequently stabilization programs often contain requirements to reduce or eliminate fiscal deficits. The causes of these deficits, and their impact on the economy, are therefore important questions that need to be handled in any analysis where one must make recommendations about desirable changes in domestic credit policy.
The model of the government sector that we utilize is taken from Aghevli and Khan (1978),23 where it is argued that nominal government expenditure adjusts proportionally to the difference between the authorities’ target spending and the actual level of expenditure in the previous period
where G and G*are the actual and desired levels of nominal government expenditure, respectively, and γ8 is the coefficient of adjustment,0≤γ8≤1. The desired level of expenditure is simply related to the level of nominal income24
It is probably reasonable to assume that in the long run the government would wish to increase its expenditure in line with the growth of nominal income, and therefore one would expect a priori that the income elasticity, γ9, would be equal or close to unity. Such a restriction would normally also be required to ensure that the overall model has a steady-state solution when capacity income and foreign prices, or the exchange rate, are allowed to change over time. This constraint is not imposed on the model during estimation, since there is no reason to suppose that it has held during the sample period across our group of countries.
Substituting equation (7) in equation (6) and solving for the (logarithmic) level of government expenditure, one obtains
As with expenditure, nominal government revenues (T) adjust to the difference between planned revenues (T*) and the actual revenues obtained in the previous period
Desired nominal revenues are specified as a function of nominal income
Substituting from this equation for T* in equation (9) gives25
Real Income
Reflecting the short-term perspective of a stabilization program, this model focuses on determining the deviations of actual output from its full capacity level, rather than on capacity output itself. Since capacity output is treated as exogenous to the model, such factors as capital accumulation, population growth, and technical progress are not considered here. However, because this model distinguishes clearly between capacity output and current output, it would not be difficult to extend it to allow for endogenous capacity growth if a more detailed analysis of the supply side of the economy were desired, for example, in the context of programs designed for purposes of structural adjustment.26
Basically, it is argued that the rate of growth of output is positively related to the excess stock of real money balances, and to the so-called output gap, represented here by the difference between normal capacity output and actual output of the previous period27
where Δlog y is the growth of real income, and y* is the normal (orcyclically adjusted) level of output.28 This latter variable is simply proxied by the trend level of real income, that is,
where y*0 represents the base level, and g the trend growth rate, of real income.
This formulation, which is very close to that outlined by Laidler and O’Shea (1980), states that any disequilibrium in the money market will result in a temporary expansion of real income, and, of course conversely, any tightening of monetary policy that results in a fall in real money balances will have output consequences through hoarding effects on the level of real expenditure.29 The degree to which this occurs is measured by the parameter γ12. While there are no strong theoretical priors on the size of this parameter, conventional wisdom would probably tend to argue that it would be small. However, this is clearly an empirical question, and one would wish to remain agnostic about the extent to which monetary changes affect real income pending the estimation results.
Equation (12) also hypothesizes that when the actual level of real income is below its normal capacity level, current output will tend to expand. If it were argued that there is a one-for-one relationship between growth and this gap, that is, γ13=1, then equation (12) would simply say that current real income would deviate from capacity only when there was monetary disequilibrium. In other words, equation (12) could then be written as
This constraint was not imposed on the structure, however, and the parameter, γ13,was left to be freely determined. Substituting for md in equation (12) gives
or, in terms of the level of real income
Expected Inflation
Expectations of inflation, II, are assumed to be generated by the adaptive-expectations model of Cagan (1956), in which these expectations are revised proportionally to the difference between the actual rate of inflation in the previous period and the rate that was expected to prevail in that period
where γ14 measures the extent to which the revision of expectations responds to the error, and 0 ≤ γ14≤ 1.
This expectations mechanism has certain theoretical problems,30 and does not fit easily into the currently popular rational-expectations framework developed by Sargent and Wallace (1973) and Barro (1977; 1978), among others; but it is still the most commonly used because of its inherent simplicity, a property that is of considerable importance given the limited availability of data for developing countries. For this reason, we have also used this approach to modeling expectations.
Domestic Credit and Money Supply
Generally speaking, in an open economy the domestic component of the money stock—namely, the net level of domestic credit extended by the banking system—is taken to be the basic monetary tool. However, any model for a developing country must recognize the linkage that exists between government fiscal operations and the supply of money. For this reason, domestic credit is allowed to be determined endogenously in the following manner. Changes in domestic credit (ΔDC) can take place through changes in the banking system’s claims on the government (ΔCG) and on the private sector (ΔCP), that is,
or,
If all changes in claims on the government are a reflection of the fiscal deficit of the government, then equation (19) can be written as
In this formulation, any expansion of the fiscal deficit results in an equivalent increase in the stock of domestic credit. This implicitly assumes that the government finances its deficit by borrowing from the banking system, using its cash balances held with banks, or by borrowing abroad and converting the proceeds into domestic currency. Only if the government were able to borrow domestically from the nonbank sector—say, by selling bonds or bills—would this identity break down. It is obvious that here the assumption of the lack of a sufficiently developed domestic market for securities, government or otherwise, becomes crucial. Despite recent progress in the development of these markets, the scope for such borrowing is fairly limited in developing countries, thereby confirming the appropriateness of definition (20).
The supply of money—broadly defined to include currency, demand deposits, and time and savings deposits—is identically equal to the net stock of international reserves (in domestic currency terms) and the level of net domestic credit extended by the banking system
Complete Model
The full structural model, along with definitions of the relevant variables, is shown in Table 1. Generally speaking, in this model it is expected that a once-and-for-all expansion in domestic credit will, through increasing the nominal supply of money, simultaneously raise the rate of inflation and real income and worsen the balance of payments. Both the increase in domestic prices and the leakage through the balance of payments will tend to lower the real stock of money, thereby reversing the process. Because of the rise in real income, real demand for money will also rise and thus support the movement of the system toward equilibrium.31 Eventually, if the system is stable, long-run domestic inflation will be equal to foreign inflation, and the level of real income will be determined by capacity output. The relative sizes of the various effects, and the lags in adjustment involved, obviously depend on the particular values of the parameters in the system.
Specification of the Model
Specification of the Model
Inflation | |
Balance of Payments | |
Government Sector | |
Expenditure | |
Revenues | |
Real Income | |
Expected Inflation | |
Domestic Credit | |
Money Supply | |
Real Money Balances | |
Definition of Variables | |
Endogenous | |
Δlog P=rate of inflation | |
Δlog R=growth of international reserves | |
G=nominal government expenditure | |
T=nominal government revenues | |
y=real income | |
∏=expected rate of inflation | |
DC=domestic credit of the consolidated banking system | |
M=nominal stock of money | |
m= real money balances | |
Exogenous | |
ϵ= exchange rate, index of units of domestic currency per unit of foreign currency | |
Pf=foreign price index | |
y*= trend value of real income | |
ΔCP=change in net claims of the banking system on the domestic private sector, and other items (net) in the banks’ consolidated balance sheet |
Specification of the Model
Inflation | |
Balance of Payments | |
Government Sector | |
Expenditure | |
Revenues | |
Real Income | |
Expected Inflation | |
Domestic Credit | |
Money Supply | |
Real Money Balances | |
Definition of Variables | |
Endogenous | |
Δlog P=rate of inflation | |
Δlog R=growth of international reserves | |
G=nominal government expenditure | |
T=nominal government revenues | |
y=real income | |
∏=expected rate of inflation | |
DC=domestic credit of the consolidated banking system | |
M=nominal stock of money | |
m= real money balances | |
Exogenous | |
ϵ= exchange rate, index of units of domestic currency per unit of foreign currency | |
Pf=foreign price index | |
y*= trend value of real income | |
ΔCP=change in net claims of the banking system on the domestic private sector, and other items (net) in the banks’ consolidated balance sheet |
II. Estimation Results
The model described in the previous section was estimated using a pooled sample of time-series cross-sectional data for 29 developing countries.32 For each country, eight annual observations were obtained on the relevant variables. While allowance was made for country differences arising from, say, size, etc., through the use of country dummies, the estimation procedure basically assumes that behavioral parameters are the same across countries.33 Since the same money demand parameters appear in three different equations, the criterion of efficient estimation requires the imposition of appropriate across-equation restrictions. This is accomplished by employing a full-information maximum-likelihood (FIML) estimator that allows nonlinear constraints to be placed on parameters both within and across equations.34
The FIML estimation method that is used does, however, require that the model be linear in (the logarithms of) the variables, while the model as specified is nonlinear, owing to the identities defining domestic credit and the nominal supply of money. For estimation purposes, and for analyzing questions of the dynamic stability of the model, these two identities have been approximated by relationships that are linear in the logarithms of the variables. For domestic credit, the equation used is
and for the money supply35
The method of linearization that yields equations (22) and (23), and the calculated values for the constants of linearization—namely, γ15to γ23 are shown in Appendix III.
After introducing equations (22) and (23) into the model and adding the individual country dummies in each of the equations,36 the complete model was estimated by the FIML method. The data used are described in Appendix I., and the results are presented in Table 2.37 This table shows the point estimates of the individual behavioral parameters (excluding, for convenience, the estimated coefficients of the dummy variables) and the ratios of the coefficients to their respective standard errors.38 The constant terms in each equation were not constrained, and are thus reported in their composite forms. To give some general idea of the goodness of fit of each of the equations, the respective mean-square errors (MSE) and the corresponding variances of the dependent variables (σ2γ) are presented. Such a heuristic comparison turns out to be necessary, since it is well known that more standard measures of goodness of fit of individual equations, such as the coefficient of determination and the standard error of estimate, do not have any formal meaning in the framework of systems estimation methods.39
Parameter Estimates1
In this table, σ2γis the variance of the level of the dependent variable, and MSE is the mean-square error from the estimated equation.
Value imposed.
These parameters are imposed as required by the linearization of the two nonlinear identities in the model. See Appendix III.
Parameter Estimates1
Parameter | Point Estimate | T-Value | ||||||
---|---|---|---|---|---|---|---|---|
Inflation | ||||||||
Adjustment | γ1 | 0.331 | 6.89 | |||||
γ2 | 0.273 | 6.76 | ||||||
γ3 | 1.02 | — | ||||||
Money demand | ||||||||
Income | γ4 | 1.213 | 8.99 | |||||
Expected inflation | γ5 | 1.206 | 5.32 | |||||
Constant | -(γ1β1-γ2β0) | 0.644 | 3.42 | |||||
σ2γ= 0.841 | ||||||||
MSE= 0.009 | ||||||||
Balance of Payments | ||||||||
Adjustment | γ6 | 0.494 | 2.74 | |||||
γ7 | 0.245 | 2.00 | ||||||
Constant | γ6β1+γ7β0 | -1.060 | 2.29 | |||||
σ2γ = 5.948 | ||||||||
MSE = 0.086 | ||||||||
Government Sector | ||||||||
Expenditure | ||||||||
Adjustment | γ | 0.744 | 14.04 | |||||
Income elasticity | γ9 | 1.041 | 55.56 | |||||
Constant | γ9β2 | -1.043 | 9.55 | |||||
σ2γ = 5.289 | ||||||||
MSE = 0.025 | ||||||||
Revenues | ||||||||
Adjustment | γ10 | 0.752 | 16.52 | |||||
Income elasticity | γ11 | 1.112 | 58.76 | |||||
Constant | γ10β3 | -1.543 | 17.80 | |||||
σ2γ=5.535 | ||||||||
MSE = 0.014 | ||||||||
Output | ||||||||
Adjustment | γ12 | 0.043 | 2.98 | |||||
γ13 | 0.819 | 13.46 | ||||||
Constant | γ12β1 | 0.096 | 2.48 | |||||
σ2γ= 5.757 | ||||||||
MSE = 0.002 | ||||||||
Expected Inflation | ||||||||
Adjustment | γ14 | 1.02 | — | |||||
Domestic Credit | ||||||||
γ15 | 1.2363 | — | ||||||
γ16 | 1.0363 | — | ||||||
γ17 | 0.1623 | — | ||||||
γ18 | 0.8053 | — | ||||||
Constant | β4 | -3.162 | 7.03 | |||||
σ2γ = 9.323 | ||||||||
MSE= 1.724 | ||||||||
Money Supply | ||||||||
γ19 | 1.1053 | — | ||||||
γ20 | 0.2503 | — | ||||||
γ21 | 0.2473 | — | ||||||
γ22 | 0.5683 | — | ||||||
γ23 | 0.4673 | — | ||||||
Constant | β5 | 0.925 | 3.58 | |||||
σ2γ=6.119 | ||||||||
MSE = 0.015 |
In this table, σ2γis the variance of the level of the dependent variable, and MSE is the mean-square error from the estimated equation.
Value imposed.
These parameters are imposed as required by the linearization of the two nonlinear identities in the model. See Appendix III.
Parameter Estimates1
Parameter | Point Estimate | T-Value | ||||||
---|---|---|---|---|---|---|---|---|
Inflation | ||||||||
Adjustment | γ1 | 0.331 | 6.89 | |||||
γ2 | 0.273 | 6.76 | ||||||
γ3 | 1.02 | — | ||||||
Money demand | ||||||||
Income | γ4 | 1.213 | 8.99 | |||||
Expected inflation | γ5 | 1.206 | 5.32 | |||||
Constant | -(γ1β1-γ2β0) | 0.644 | 3.42 | |||||
σ2γ= 0.841 | ||||||||
MSE= 0.009 | ||||||||
Balance of Payments | ||||||||
Adjustment | γ6 | 0.494 | 2.74 | |||||
γ7 | 0.245 | 2.00 | ||||||
Constant | γ6β1+γ7β0 | -1.060 | 2.29 | |||||
σ2γ = 5.948 | ||||||||
MSE = 0.086 | ||||||||
Government Sector | ||||||||
Expenditure | ||||||||
Adjustment | γ | 0.744 | 14.04 | |||||
Income elasticity | γ9 | 1.041 | 55.56 | |||||
Constant | γ9β2 | -1.043 | 9.55 | |||||
σ2γ = 5.289 | ||||||||
MSE = 0.025 | ||||||||
Revenues | ||||||||
Adjustment | γ10 | 0.752 | 16.52 | |||||
Income elasticity | γ11 | 1.112 | 58.76 | |||||
Constant | γ10β3 | -1.543 | 17.80 | |||||
σ2γ=5.535 | ||||||||
MSE = 0.014 | ||||||||
Output | ||||||||
Adjustment | γ12 | 0.043 | 2.98 | |||||
γ13 | 0.819 | 13.46 | ||||||
Constant | γ12β1 | 0.096 | 2.48 | |||||
σ2γ= 5.757 | ||||||||
MSE = 0.002 | ||||||||
Expected Inflation | ||||||||
Adjustment | γ14 | 1.02 | — | |||||
Domestic Credit | ||||||||
γ15 | 1.2363 | — | ||||||
γ16 | 1.0363 | — | ||||||
γ17 | 0.1623 | — | ||||||
γ18 | 0.8053 | — | ||||||
Constant | β4 | -3.162 | 7.03 | |||||
σ2γ = 9.323 | ||||||||
MSE= 1.724 | ||||||||
Money Supply | ||||||||
γ19 | 1.1053 | — | ||||||
γ20 | 0.2503 | — | ||||||
γ21 | 0.2473 | — | ||||||
γ22 | 0.5683 | — | ||||||
γ23 | 0.4673 | — | ||||||
Constant | β5 | 0.925 | 3.58 | |||||
σ2γ=6.119 | ||||||||
MSE = 0.015 |
In this table, σ2γis the variance of the level of the dependent variable, and MSE is the mean-square error from the estimated equation.
Value imposed.
These parameters are imposed as required by the linearization of the two nonlinear identities in the model. See Appendix III.
As theoretically expected, the results indicate that an excess supply of real money balances results in an increase in the rate of inflation. The parameter measuring this effect has a positive sign and is significantly different from zero at the 1 percent level. The money demand function embedded in this equation, which is restricted to have the same parameter values in the balance of payments and output equations, appears to be reasonably well determined. The equilibrium income elasticity, is significantly greater than unity and close to the value that is generally obtained for developing countries.40 In financially developed economies, one might expect to observe a roughly proportional relationship between real money balances and real income,41 but in developing countries the demand for cash balances often rises more than proportionately to the growth in income, owing to the secular process of monetization and the absence of alternative liquid financial assets in which private savings may be held.
The semielasticity of real money balances with respect to the expected rate of inflation has the correct sign and is significantly different from zero at the 1 percent level. The size of this parameter is also well within the range that has been observed in estimates for developing countries.42 In generating the series on expected inflation, II, the coefficient of expectation, γ14, was arbitrarily constrained to be unity. This was done mainly on grounds of simplicity, since allowing it to vary would have greatly complicated the estimation model.43 Furthermore, ordinary least-squares tests on the pooled sample using the money demand equation in Appendix V. revealed that the value of γ14that maximized the log-likelihood function was indeed unity. In the light of this result, our restriction on γ14 (which implies that full adjustment occurs within a year) would certainly seem to be reasonable.
If the domestic price level is pushed above its equilibrium relationship to foreign prices, pressures are built up that force the domestic rate of inflation down. This is captured by the parameter γ2, which turns out to be statistically significant at the 1 percent level. Since our theoretical analysis suggested that the parameter γ3 should take on a value of unity, this restriction has been imposed for efficiency of estimation. It should be remembered then that this model explains the deviations of domestic inflation from the foreign rate of inflation.
Unlike the equilibrium model emerging from the analysis associated with the monetary approach to the balance of payments, which assumes that international reserve flows will immediately offset any excess supply of money, this dynamic model assumes that some inertia exists in this relationship. The empirical results indicate that only about one half of the excess supply is reflected in reserve variations. As mentioned in Section I, the parameter γ6 is a mixture of the impacts of monetary changes on both the current and capital accounts, and its observed sign is consistent with either of these two channels. From the results, it appears that if the domestic price level deviates from the foreign price level, international reserve flows will be generated and this elasticity is significant at the 5 percent level.
The nominal income elasticities of both government expenditure and revenues are close to unity and are highly significant. The specific values imply that in the steady state one would expect that government expenditure would move proportionally with inflation. On the other hand, it appears that revenues would tend to rise somewhat faster.44
While the major determinant of changes in real income is the difference between capacity real income and the actual level, monetary disequilibrium is a factor as well. The elasticity measuring the impact of the excess demand for real money balances on the growth rate, γ12, is significantly different from zero at the 1 percent level. This is one of the more important results to emerge from this exercise, both because it has not been investigated extensively in the context of developing countries and because of its important implications for stabilization programs.45
It also turns out that real income will consistently tend toward its capacity level, with any discrepancy between capacity real income and actual real income (of the previous period) being eliminated fairly rapidly.46 The coefficient of adjustment, γ13, is fairly large, although not quite unity,47 implying that deviations from capacity real income could apparently persist for only a short period if there were no excess demand for money.
The output effect of changes in the money supply allows real money balances to rise initially, or the income velocity to fall, when there is an increase in the rate of monetary growth, since this would raise money demand itself. This situation makes the results of the present model consistent with those emerging from the model proposed by Khan (1980), where this phenomenon is investigated at length. Finally, it should be noted that certain alternative specifications for real income, such as the introduction of relative prices into the equation, were fruitless.48 The coefficient of the ratio of domestic to foreign prices was not significant even in a sample as large as this one.49 The simple model utilized here does reasonably well in explaining movements in real income.
Methods of testing the overall goodness of fit of structural models in any rigorous way are still in their infancy. One statistic that is customarily used, namely, the Carter-Nagar R2, was 0.998, indicating the reasonableness of the overall specification. For the individual equations, the task of assessing goodness of fit is equally difficult, although a rough idea can be obtained by examining the mean-square errors relative to the variance of the dependent variable.50 All seven of the stochastic equations seem to be well specified on this criterion.51 This result was achieved without any recourse to special dummy variables other than the country dummies required by the pooled time-series cross-sectional estimation procedure.
Given that the model is dynamic and involves several feedbacks, it is also important to determine if the estimated parameters combine to yield a stable model.52 For this reason, the eigensystem of the estimated model was calculated. All the moduli, which are shown in Appendix IV, are less than unity, and therefore the estimated model can be considered dynamically stable. A related question is that of the sensitivity of the overall stability of the model to the particular values of the estimated parameters. This is a consideration in any theoretical analysis because it helps to identify the key parameters in the system. In the context of this model, the analysis can be made by evaluating the derivatives of the calculated moduli with respect to the relevant parameters.53 Three parameters are particularly crucial for the stability properties of the model. First, a small increase in the impact of excess real money balances on the rate of inflation would raise one of the moduli above unity. Second, the stability of the system is apparently sensitive to both parameters in the output equation. This finding supports the earlier claim that the feedback running from real money balances to growth is very important to the dynamic behavior of the system.
III. Policy Implications of the Model
The estimated model of the previous section is now used to illustrate some policy issues that arise in connection with the implementation of a stabilization program designed to improve a country’s external payments position. For this purpose, it is assumed that the estimated model can be taken to reflect the dynamic behavior of a “representative” developing country. This is, of course, a rather strong assumption, but it can be justified on two counts. First, since the parameters of the model were estimated from a pooled sample of observations drawn from 29 developing countries, the estimates represent in a sense the average parameter values for this sample as a whole. Second, the purpose is not to describe the specific behavior of a particular country, but rather to investigate the general path that might be followed by major economic aggregates in different circumstances. To illustrate the dynamic behavior, it is first shown how the model responds to two types of shock: a one-period increase in domestic credit, and a permanent increase in the level of capacity output.54 In each case, domestic prices are initially in equilibrium at given levels for the exchange rate and foreign prices. Output and employment begin at the capacity level; the government budget is in balance; and domestic credit, international reserves, and the nominal money stock are all constant. Then the focus shifts to the paths taken by these variables in response to each exogenous change. After the effects of these shocks are examined, the model is used to analyze the implications of two different types of financial program for a country’s inflation rate, output and employment, and domestic credit. For purposes of the simulations, the model is no longer required to be linear in variables, and therefore its original nonlinear form is restored in order to conduct the various experiments reported in this section. This way the results are more useful, in the sense that they are taken directly from the type of (nonlinear) model that policymakers would presumably be considering. Also, in this case one can be certain that none of the simulation results are attributable to the linearization procedure employed in estimation.
First, consider the effects of a one-period exogenous shock to domestic credit that causes a rise of 10 percent in the nominal money stock in the first year. This is accomplished by adding a stochastic component, Z, to the identity relating the rate of domestic credit expansion to the government’s budgetary deficit.55 The Z variable shocks the system for one period, after which the government spending and tax functions again determine the evolution of the budgetary deficit or surplus. In the sample used in this paper, domestic credit makes up, on average, about 75 percent of the money stock, while the remainder is backed by international reserves. Thus, Z must create an increase of about 13.3 percent in domestic credit to raise the money stock by 10 percent at the end of the first period.56
The lines in Charts 1 and 2 trace the effects of the one-period monetary shock on the endogenous variables of the model. The path of prices is also compared with that of a simple closed-economy monetary model57 in which real income is treated as exogenous, and thus the price level is the only endogenous variable. Despite the fact that both models are estimated using the same sample of data, Chart 1 shows that the dynamic path of prices in the open-economy system is quite different from that generated by the naive model. In the closed-economy model, the price level will eventually rise by 10 percent in response to a once-for-all increase in the nominal money stock. In this case, prices rise by almost the same percentage as the monetary expansion in the first period, but then they overshoot the new equilibrium level as the increased opportunity cost of holding money reduces the demand for real balances at constant income. After rising by nearly 12 percent, the price level gradually tracks back to 10 percent above its initial level, with most of the adjustment completed in five years. By contrast, in the expanded approach, the monetary expansion initially stimulates both prices and real output and employment. As in the previous case, the price level at first rises quite strongly in response to the excess supply of money. However, the fact that domestic prices are increasing relative to foreign prices soon induces price setters to moderate the domestic inflation rate, and at the same time the induced outflow of international reserves causes the money supply gradually to fall back toward its preshock level. Four years after the monetary expansion, domestic prices reach a maximum that is about 5.5 percent above their long-run equilibrium value and then begin to track back—at first quickly, but then more and more slowly—to equilibrium, as the cumulative effects of continuing payments deficits bring the money stock back toward its initial level. Nevertheless, in the model the inflation of domestic prices induced by the monetary expansion keeps the home country’s exchange rate overvalued for quite a long time. Eight years after the expansion, the price level is still about 2.5 percent above its final equilibrium.
Effects of a Monetary Expansion of 10 Percent on Money, Prices, and Output
(As percentage of initial value)
Effects of a Monetary Expansion of 10 Percent on International Reserves and Domestic Credit
(In percent)
After the initial expansion, the government’s budgetary position plays only a minor role in the evolution of domestic credit and the money stock. A small fiscal surplus is created as nominal income rises, because the response of tax revenue is slightly larger than that of nominal government spending; this surplus disappears as prices and output fall back to their initial levels. These effects are small because the parameters (γ8γ9) that govern the short-run response of nominal government spending are estimated from the sample to be only slightly smaller than those that govern short-run changes in tax revenue (γ10γ11). 58 However, the dynamic path of prices is quite sensitive to the values of these parameters. Since the government has the power to alter taxing and spending policies at will, it would undoubtedly be interesting to analyze the consequences of different types of budgetary policy in the model in more detail, although this is not done here.
While the simple model explicitly assumes that a monetary expansion leaves real output unchanged, the expanded model allows it to stimulate output in the short run. In percentage terms, the output effect of expansionary monetary policy is much smaller than its price effect. Output rises by about ½ of 1 percent in the second year after the monetary expansion, before slowly declining to its equilibrium level. Nevertheless, the stimulative effect of money on real output may not seem so small when it is recalled that this represents over employment—that is, output is sustained for several periods at nearly ½ of 1 percent above its normal capacity level. If one is prepared to assume that the ratio of labor input to total output is reasonably stable, then the expanded model implies that a monetary expansion will temporarily increase employment of the labor force, and this employment effect will persist for several years. Conversely, if real money balances were lowered by 10 percent, this decline would have undesirable consequences for employment. While it certainly could be argued that the rise in unemployment would be fairly small, it should be kept in mind that in the typical developing country the impact falls disproportionately on certain sectors of the economy. The agricultural and rural sectors, since they are less reliant on credit, may not suffer as much from a restrictionary monetary policy, and often it is the fledgling manufacturing, services, and export sectors that bear the brunt. Thus, a small rise in overall unemployment may well mean an undesirable increase in urban unemployment, with all its attendant social and political ramifications.
Chart 2 traces the effect of the shock on international reserves and domestic credit. After the shock, reserves decline continuously until they approach a new equilibrium at about 68 percent of their initial level. As a result, domestic credit increases from just under 75 percent of the money stock to about 83 percent. In the new equilibrium, the expanded model yields the standard conclusion of the literature on the monetary approach to the balance of payments, namely, that an increase in domestic credit will ultimately lead to an equal reduction in international reserves and leave output, prices, and the nominal money stock unchanged. However, the adjustment path of this model is richer than the simple model in the sense that it allows a monetary disturbance to affect output, employment, and prices during the adjustment process. Furthermore, the simulation experiment using the parameters actually estimated for the sample of 29 developing countries suggests that even a one-period shock expansion of domestic credit can induce disequilibrium price effects that cause the country’s currency to be overvalued for a substantial period, a result that seems to correspond to the experience of a number of developing countries.
Next, the effect of a once-for-all increase of 10 percent in the level of capacity output is considered. This increase might be due to such factors as technical innovation or growth in the labor force. Charts 3 and 4 present the effects of such a change. Again, the price effect in the closed-economy model is compared with the price path traced by the open-economy model with endogenous output and money stock. To ensure that the government budget will be in balance at the new higher level of real income, the income elasticity of tax revenue is arbitrarily set equal to that of government spending and the constant term in the tax function is adjusted so that this property holds. In the simple model, an increase of 10 percent in output would cause prices to decline steadily until they were about 11 percent below their starting level, a direct consequence of the fact that the income elasticity of money demand is estimated to be about 1.18 for the sample. In the expanded model, current real output and employment begin to rise quite rapidly after the jump in capacity output (output has risen by nearly 9 percent by the end of the second year). The domestic price level at first falls sharply relative to prices abroad, declining by about 5½ percent (compared with its initial value) in the first two years. The excess demand for money, however, creates a balance of payments surplus that increases the money stock, so that gradually the price level turns around and there is a period of both rising prices and rising output that continues until current output has increased by 10 percent and domestic prices have risen to their initial level. In the final equilibrium, international reserves have risen by approximately 60 percent above their initial value (Chart 4), so that the proportion of the money stock that is backed by domestic credit declines from 75 percent to 62 percent.
Effects of an Increase of 10 Percent in Capacity Output
(As percentage of initial value)
Effects of an Increase of 10 Percent in Capacity Output on International Reserves and Domestic Credit
(In percent)
Government financing plays no more than a negligible role in the subsequent movements of domestic credit, owing to the assumption that γ9=γ11. Again, the final equilibrium is consistent with the conclusions of the monetary approach to the balance of payments. But here the adjustment process involves a rapid increase in actual output toward the capacity level, steady accumulation of international reserves, and a price level that first declines and then gradually returns to its initial equilibrium.
The results of a related simulation may also be summarized briefly. Suppose that there is a once-for-all rise in the rate of growth of capacity output, g, to 1 percent per period. The simulated time paths of most endogenous variables are broadly similar to those in the previous experiment. The only major difference is that because λ2 is nonzero after the increase in g,59 actual output grows immediately at the same rate as capacity output without the catch-up that was apparent in the earlier simulation. Again, prices at first fall below their initial level as rising output pushes up the demand for real money balances. Output growth also induces a small budget surplus because, given the estimated parameters of the model, growth causes government revenue to rise slightly faster than expenditure. But this mild contractionary influence on domestic credit is more than offset by an overall payments surplus that induces an inflow of reserves at a rate that rises to a maximum of about 5 percent a year before gradually declining again. In the present simulation this adjustment occurs much more gradually than it did for the one-shot increase in capacity output illustrated in Charts 3 and 4. Thus, after falling by about 7 percent by the twenty-fourth period after growth begins, prices start to rise toward their initial level, as international reserve accruals increase the rate of monetary expansion.
The simulations involving shocks to the level of capacity output and its rate of growth are both consistent with another frequently observed phenomenon that most models cannot explain: namely, that fast-growing developing countries frequently turn in balance of payments surpluses, even during years when domestic prices appear to be rising more rapidly than prices in the rest of the world. The results here are also consistent with an important converse proposition that has recently been emphasized by proponents of the “new” supply-side approach to macroeconomics. It is that government policies that adversely affect the rate of growth of capacity output may eventually give rise to a situation where the economy suffers from a protracted period of exchange rate overvaluation and balance of payments deficits. This proposition underscores the potential importance of policies that enhance the rate of growth of capacity supply in the design of economic stabilization programs for developing countries.
Having studied the effects of these exogenous shocks, we now make use of the extended model to analyze an important issue that frequently arises in the setting up of a stabilization program. Suppose that the authorities of a country wish to achieve some specific targeted improvement in the country’s external payments position—say, an increase of 50 percent in official holdings of international reserves.60 Suppose, too, that the authorities have the option of choosing between a “standard” program, in which the targeted increase is to be achieved in one year, and an “extended” program, in which the rise is spread over five years. Two questions now arise. First, given the estimated dynamic structure of our “representative” developing country, what are the annual domestic credit ceilings needed for success in achieving the balance of payments improvement that is targeted in the program and, given the revenue function, what is the size of the cut in government spending that would be required to attain these ceilings? This question is a particular instance of the “targets and instruments” approach to the theory of economic policy. In this case, the targeted improvement in the country’s external payments position is taken as given, and the structure of the model is allowed to determine both the setting of the policy instrument (domestic credit) required to hit the reserves target and the consequences for other important economic variables, such as output, employment, and prices.61 This leads to the second question: Which of the two programs—standard or extended—is likely to have fewer undesirable effects on domestic prices, output, and employment? In particular, given some appropriate social rate of discount, which program involves the smallest total output and employment losses over its lifetime?
In Chart 5, the solid line is the target path for reserves in the standard program while the dashed line is the target path of reserves in the extended program. The dotted line and the dash-and-dot line give the corresponding settings of domestic credit that are required to achieve the targeted result within the time frame of each program. Several important implications of the expanded model are immediately evident from Chart 5. First, one might have expected that a one-step increase in reserves could be achieved via a similar reduction in domestic credit, after which the credit ceiling could be lifted. But the simulation results in Chart 5 show that this intuitive conclusion is not correct. To achieve a steady improvement in the economy’s external position, it is necessary to have a fluctuating path of domestic credit in both cases, rather than steady restraint.62 While improvement in the external position requires domestic credit restriction in the first year of each program, the credit ceiling has to be raised again in the second year. This increase in domestic credit is quite small for the extended program but rather large for the standard program—a point that is discussed later. The main point, however, is that because the private sector’s observed behavior is characterized by lagged adjustment, steady improvement in a country’s balance of payments cannot, in general, be achieved through imposing a domestic credit ceiling that is fixed over time.63
Time Paths of Domestic Credit Needed to Achieve Sudden and Gradual Increases in International Reserves
(As percentage of initial value)
A second and more general conclusion from Chart 5 is that the one year program requires changes in the setting of domestic credit that are much larger than those in the five-year program, and these oscillations have to continue for several years after the reserve improvement has been obtained, in order to hold international reserves steady at their new level. Specifically, given the assumption about price expectations, the standard program requires first an extremely large reduction in domestic credit, then a subsequent expansion to 180 percent of the initial level, and then a second rather severe contraction. By contrast, the extended program involves much less severe credit restraint, but one that continues, at varying levels, for the full five years. Chart 6 gives the budget surplus (+) or deficit (−) as a proportion of full employment nominal income. Again, the overall fiscal position of the government is seen to fluctuate much more in the standard program than in the extended one.
Budget Surplus as a Percentage of Full Employment Nominal Income
(In percent)
Having considered the setting of the goverment’s policy instruments, the discussion now turns to the effects of each program on the other important variables in the domestic economy. Chart 7 gives the price effects of the two financial programs, while Chart 8 indicates their respective output effects. The standard program induces a severe price deflation in the first year, followed by a sharp inflation that gradually tails off after two or three years, as domestic prices return to their initial equilibrium level relative to prices in the rest of the world. By contrast, the five-year program displays a prolonged but much shallower dip in domestic prices. It is even more interesting to compare the output effects of the two programs given in Chart 8. If one again assumes that the labor-to-output ratio is relatively fixed in the short run, this chart can also be interpreted as a comparison of the employment effects of each program. In the one-year program, unemployment rises by 5 percentage points during the first year and falls abruptly in the second year. The next two years are characterized by relatively rapid inflation and slightly overfull employment, which dissipates in the fifth year. On the other hand, the extended program is initially characterized by a relatively mild increase of about 1 percent in the unemployment rate, and this gradually declines to zero over the next four years, rising briefly above full employment in the year after the targeted increase in international reserves has been completed.
Prices as a Percentage of Their Equilibrium Level
(In percent)
Output as a Percentage of Capacity Output
(In percent)
In comparing the employment costs of the two programs, it is relevant to note that the area between the employment line and the full employment line for the standard program is much larger than the corresponding area for the extended program. This implies that if one takes a social rate of discount of zero and treats both underemployment and over-employment as equally “bad,” the employment costs of the standard program are considerably larger than those of the extended program.64 If unemployment alone is taken as “bad” (the areas under the full employment line), the employment cost of the standard program is still larger than that of the extended program, and this discrepancy is larger the higher is the authorities’ rate of time preference.65 It may be concluded that, given the expanded model, the employment costs of a standard one-year program are unambiguously larger than those of an extended program, so that on this criterion the latter is to be preferred.
It is, of course, important to avoid drawing excessively general conclusions from the simulation analysis. In the first place, these results apply only to the “average” member of the sample of developing countries. The structural model might have to be modified in significant ways not only for countries excluded from this sample but also to take account of the special characteristics of each of the 29 included countries.
A second and more important factor that qualifies these results is the limited treatment of the behavior of expectations in the simulation model. While the adaptive mechanism that generates expectations of inflation seems to be empirically reasonable for the present sample of countries, many alternative assumptions are possible about the way in which market participants form their expectations about future price developments. To take just one example, it would be interesting to study the behavior of the model in conditions where expectations of inflation are based on current and past rates of monetary expansion.
Another aspect of the simulation analysis that needs qualification for real-world applications is the assumption that price expectations—and the mechanism that generates these expectations—are unaffected by the setting up of a stabilization program. Anyone who has been involved in the formulation of a financial program is aware that expectations of future price movements may be strongly affected by the announcement and implementation of the policy measures. For example, in a high-inflation country, price expectations may shift downward immediately when a financial program is announced. This sudden reduction in the anticipated inflation rate reduces the perceived opportunity cost of holding cash balances, increasing nominal money demand at the initial level of income and prices. Such an effect is, by its very nature, exceedingly difficult to quantify. Nevertheless, if the private sector becomes convinced that the program will succeed, the reduction in domestic credit required to achieve the targeted improvement in the country’s external position is smaller than would have been required if price expectations had been unaffected by the announcement of the program. This reflects the familiar observation that if a program is widely viewed as stringent enough to achieve its objective, its domestic credit ceilings can be less deflationary than they would have had to be if the program was not regarded as credible.
It is also important to note that unless a government is prepared to intervene in the domestic price-setting process or the labor market, the price and output consequences of its financial program will essentially be determined by the private sector through behavior relations like those specified in Section I. Thus, the government cannot ensure by fiat that the conditions of stagflation will never occur. If expectations were affected by a program in the way described earlier and policymakers failed to take account of this in setting domestic credit ceilings, the program would tend to be more restrictive than was necessary to obtain the desired objective, and its price and output effects might be considerably more deflationary than had been anticipated.
The simplifying assumption that the reduction in domestic credit is achieved entirely by a cut in government spending requires less qualification. Practical experience with the setting up of programs like those designed by the International Monetary Fund suggests that once the implications of the required domestic credit ceiling for government financing are being worked through, it is often found that both government taxing and spending policies have important components that cannot be easily altered, so that it is usually around these policies that the most difficult political decisions take place. However, this should not be allowed to obscure the basic fact that once the target path for international reserves has been decided, there is only one set of domestic credit ceilings that is consistent with the achievement of that path, and the only question is that of making government financing requirements consistent with those ceilings.
The treatment of the supply side of the economy in this paper also warrants comment. In financial programming exercises, the effect of domestic credit restriction on aggregate supply has often been treated as a secondary problem. In particular, some analyses do not distinguish clearly between the effect of a program on actual output and its effect on full capacity output. This distinction is important because some of the same factors may affect both the growth of potential output and the rate of capacity utilization. Full capacity supply in the domestic economy is treated as exogenous to our model, while actual current output is endogenous. Thus, the model incorporates the short-run determinants of supply, but not the factors that are responsible for capacity growth in the longer run. If fixed capital formation is related to real government spending or taxes, however, then a deflation of domestic credit that is achieved by either reducing public sector expenditure or raising tax revenue will affect the rate of growth of capacity output as well as actual output.66 Much, therefore, depends on the strength and direction of these structural relationships in a given developing economy. Many economists would probably argue that—whether the domestic credit ceiling is achieved by a reduction in real government spending or an increase in taxes—the effect will be a decline in aggregate net investment and a reduction in capacity growth. Alternatively, some economists and policymakers have recently argued, in effect, that a reduction in government spending in some developing countries might lead to a “crowding in” phenomenon, in which the decrease in government-induced capital formation is more than offset by a rise in private sector investment. If the first argument is relevant, so that capacity output is positively related to real public sector spending, then the conclusion of the preceding analysis that the extended financial program is superior to the standard one would be reinforced. Alternatively, if there is significant crowding in, then a standard program might be superior to an extended program in certain circumstances. Although the standard program would still have the same deflationary effect on current output in the short run, these adverse impact effects might be more than offset by faster growth of both actual and capacity output in subsequent years. All that can presently be said is that the relationship between government spending and growth of full capacity supply is an important empirical question that deserves further research in the context of developing countries.
IV. Conclusion
The basic purpose of this paper has been to formulate a model for developing countries that allows output, prices, international reserves, money, and government taxing and expenditure policies to be determined simultaneously. We believe that the model developed is a formal interpretation of the theory underlying the typical stabilization program implemented by the authorities to combat problems of inflation and an adverse balance of payments.
Since the focus of the exercise is on practical policy questions, empirical estimates of the relevant parameters are obviously required. However, because of the limited availability of data, it would be difficult to estimate the parameters of such a model from data for a single developing country; and even if it were possible, it is not clear that these estimates would be applicable to other countries. Thus, a pooled sample of annual data for 29 countries, comprising 232 observations in all, was used to test the structure. The estimates indicate both that the model is representative of the structural characteristics of these 29 developing countries and that monetary disequilibrium does indeed have a significant effect on the behavior of prices, output, and reserves. To examine the policy implications, two simple simulation experiments were performed. It was found that during a portion of the adjustment period following an exogenous shock, the model produces time paths for prices and output that are consistent with real-world experience but that are not generally explained by simple monetary models. The last and most important aspect of the work was the comparison of two alternative types of program designed to achieve a specified improvement in the external position of a developing country—a standard (one-year) program and an extended (five-year) program. Several general conclusions may be drawn from this comparison.
The most important conclusion relates to the complexity of the linkages between the targeted reserve increase and the domestic credit policies needed to achieve it. The initial simulation experiment indicated that a change in domestic credit would essentially cause reserves to approach their new equilibrium level asymptotically. Nevertheless, when a given increase in international reserves must be achieved within a specified period of time, this model yields quite a complicated path for domestic credit ceilings, although the fluctuations are much more pronounced for the truncated one-year program than they are for the longer program. The implication of these results is that even a modest extension of the financial programming framework yields a model in which the relationship between the targeted reserve increase and domestic credit is complicated and depends on the structure of the economy. Conversely, measures to hold domestic credit at some prearranged level67 will not result in a smooth path of accumulation of international reserves, a conclusion that holds in this model even though it neglects such complications as the effect of changing expectations on international capital flows. The practical implication is that policymakers cannot “fine tune” domestic credit ceilings from quarter to quarter or even year to year without having much more comprehesive information about the structure of the economy than they can reasonably be expected to possess. One possible way out of this difficulty might be to devise some simple feedback rule in which the domestic credit ceilings are altered in response to current and past deviations between targeted and actual reserves. But the derivation of such a “closed-loop” policy rule remains a question for further research.
A second general conclusion suggested by the findings in this paper is that programs designed to achieve quick results on the balance of payments via sharp deflation are likely to have significant and undesirable effects on output, employment, and factor incomes, particularly in the short run. In the expanded model, unemployment rises by 5 percentage points in the first year of the sharply deflationary program. A rise of this magnitude may well impose a heavy burden on a developing country both because incomes are already near the subsistence level and because the employment effect is likely to fall disproportionately on the nascent industrial sector. Furthermore, it is extremely difficult to take adequate account of the deflationary impact of expectations when devising the financial program, a problem that may tend to exacerbate the employment effects. These results present those who devise financial programs with a dilemma. On the one hand, the conclusion that the negative employment effects of an extended program are unambiguously smaller than those of a standard program provides a theoretical and empirical rationale for greater use of programs that have a longer time span. On the other hand, the longer the duration of a program, the more likely it is that the economy’s adjustment may be blown off course by events that could not have been foreseen when the program was conceived. Careful and continuous monitoring of the program would thus seem to be essential if policymakers choose the gradualist approach.
While the model here is more comprehensive, and is believed to be more realistic, than those that have been specified hitherto, it clearly represents only a step in the direction of formulating, for developing countries, models that are capable of answering the questions that continually face policymakers in the context of their stabilization programs. To provide a general framework, obviously some realism was sacrificed by ignoring the special characteristics of individual developing countries. The hope is that this model can serve as a foundation on which more detailed structures can be built. Furthermore, the parameter estimates provided here, having been based on a cross-country sample, may be useful in analyses for developing countries where such information, for whatever reason, is not readily available.
Finally, it is important to underline the fact that both the estimates of the expanded model and the simulation results are sensitive to the assumptions made in this paper, particularly regarding the process of forming expectations. Because of space limitations, other potentially interesting simulation experiments—in particular, simulations involving simultaneous changes in exchange rates and monetary policy—have not been conducted or reported. The model is, however, fully capable of handling such types of simulation. Only those simulation experiments that seemed to highlight the basic workings of the model have been considered here.
APPENDIX I Data Definitions and Sources
Data
All data used in this study are taken from International Monetary Fund, International Financial Statistics (IFS), and are annual, covering the period 1968-75 for each country. Lagged values of variables, therefore, cover the period 1967-74. All stock data are measured at the end of the period, and the price data are period averages. The precise definitions of the variables and the IFS line numbers are as follows:
-
P= consumer price index, 1975 = 100: line 64
-
R = net international reserves valued in domestic currency (line 1d multiplied by line ae)
-
G= government expenditure; line 82
-
T= government revenues; line 81
-
y= real income. This variable was generated by deflating nominal gross domestic product (GDP)—line 99b—by the consumer price index.
-
DC= net domestic credit of the consolidated banking system; line 32
-
M= money plus quasi-money; line 34 plus line 35
-
m= real money balances, that is, M/P
-
y*=trend level of real income. This series was calculated from the equation
where y*0 is the 1968 value of real income and g is its trend growth rate over the period 1968-75.
-
Pf= U.S. consumer price index, 1975 = 100; line 64
-
ϵ = index of the U.S. dollar exchange rate, 1975 = 100; line ae
-
ΔCPt, = residual item obtained from the identity for the change in net domestic credit
Countries
The 29 countries in the sample are Argentina, Brazil, Chile, Colombia, the Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti, Honduras. Nicaragua, Panama, Paraguay, Uruguay, Jamaica, Jordan, Sri Lanka, India, Korea, Malaysia, Nepal, the Philippines, Singapore, Thailand, Burundi, Ghana, Kenya, Malawi, and Zambia.
The 28-country dummies, which take on a value of unity for the eight observations corresponding to a particular country and zero elsewhere, are entered into the equations in the same order as in the preceding list.
APPENDIX II Derivation of the Price Equation
Consider a small open economy that has both a traded goods sector and a nontraded sector. The demand for nontraded goods (QdN) can be specified simply as a function of relative prices, the excess demand for real money balances, and real income68 as follows:69
where
where
PN= price of nontraded goods
ϵPf = domestic price of traded goods, equal to the exchange rate times the foreign price level. This is assumed to be exogenous. All other variables are as defined in the text.
Similarly, the supply of nontraded goods (Q5N) can be made a positive function of relative prices and real output 70
The change in the relative price of traded goods is determined by their excess demand
where k > 0.
Substituting in this equation for the demand for and supply of nontraded goods given by equations (26) and (27), and further assuming that the income elasticities, a3 and b2 are equal, one obtains
Generally, statistics on the price of nontraded goods are not available for developing countries, so that it would be difficult to work with an equation of the form of equation (29). As an alternative, however, one can utilize the overall price index, which can be represented by a log-linear index of the form
where ω is the share of nontraded goods in total domestic expenditure.
In terms of the growth rate of the price of nontraded goods, equation (30) becomes
substituting for ΔlogPNt in equation (29) gives
This is precisely the reduced form inflation equation described in the text.
APPENDIX III Log-Linear Approximation of Identities
For purposes of estimation and for the stability analysis, the structural model had to be made linear in the logarithms of the variables. This meant approximating the linear identities for the change in domestic credit and for the nominal money supply by a log-linear form, evaluated at the sample means of the relevant variables.
The domestic credit identity, which is specified as
is approximated as
where
with a bar over a variable signifying its sample mean value. The constant μ4 was estimated within the model and not imposed.
For the money supply identity, which is defined in stock terms as
we have chosen a first-order difference equation approximation that specifies that the valuation effects of exchange rate changes do not influence the level of the domestic money stock
where
In equation (36), R is the domestic currency value of foreign exchange reserves. It can easily be shown that the inclusion of current and lagged values of the exchange rate, together with the preceding constants, ensures that both the money supply and the parameter estimates of the model are free of the effects of in-sample exchange rate changes on the domestic currency value of international reserves.
APPENDIX IV Dynamic Stability of the Model
To determine whether the estimated system is stable, the eigenvalues of the model were calculated from the endogenous part of the estimated system. If a, is the jth eigenvalue of the matrix of coefficients of the endogenous variables, then since the model is a set of difference equations, the necessary and sufficient condition for stability is that the values of all moduli be less than unity, that is,71
In the model here, it turns out that there are four real eigenvalues and two complex ones, which are complex conjugates of each other and which gave rise to an endogenous cycle. These, along with the corresponding moduli and the damping period (in years), are shown in Table 3. The values of the moduli indicate that the model is locally stable.
Stability Conditions of the Model
Stability Conditions of the Model
Eigenvalue | Damping Period | |||
---|---|---|---|---|
Real part | Imaginary part | Modulus | (years) | |
1 | 0.991 | 0.991 | 1.009 | |
2 | 0.871 | 0.871 | 1.148 | |
3 | 0.816 | 0.816 | 1.226 | |
4 | 0.184 | 0.184 | 5.434 | |
5 | 0.379 | ±0.495 | 0.624 | 1.603 |
6 | 0.251 | ±0.001 | 0.251 | 3.990 |
Stability Conditions of the Model
Eigenvalue | Damping Period | |||
---|---|---|---|---|
Real part | Imaginary part | Modulus | (years) | |
1 | 0.991 | 0.991 | 1.009 | |
2 | 0.871 | 0.871 | 1.148 | |
3 | 0.816 | 0.816 | 1.226 | |
4 | 0.184 | 0.184 | 5.434 | |
5 | 0.379 | ±0.495 | 0.624 | 1.603 |
6 | 0.251 | ±0.001 | 0.251 | 3.990 |
The sensitivity of the stability of the model to changes in particular parameters was examined by taking the numerical derivatives of each of the moduli with respect to the parameters in the system.72 If γi, is the ith parameter in the model, then we calculate
If this partial derivative is large, then a small change in a parameter could conceivably make the modulus greater than unity, and thus destabilize the system. Excluding the imposed constants, there are 14 behavioral parameters in the endogenous part of the model and six moduli. The results of the sensitivity analysis are given in Table 4.73
Sensitivity Matrix of Moduli with Respect to Estimated Parameters
Sensitivity Matrix of Moduli with Respect to Estimated Parameters
Moduli | |||||||
---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | ||
Parameters | 0.991 | 0.871 | 0.816 | 0.184 | 0.624 | 0.251 | |
γ1 | 0.331 | -0.002 | 0.057 | -0.001 | -0.014 | 0.944 | 0 |
γ2 | 0.273 | 0.002 | 0.130 | 0.003 | -0.008 | 0.058 | 0 |
γ3 | 1.0 | 0 | 0 | 0 | 0 | 0 | 0 |
γ4 | 1.213 | 0 | 0.001 | 0 | 0.003 | 0.018 | 0 |
γ5 | 1.206 | 0 | 0.004 | -0.002 | -0.003 | 0.261 | 0.001 |
γ6 | 0.494 | -0.005 | -0.081 | 0.011 | 0 | 0.027 | 0 |
γ7 | 0.245 | -0.006 | -0.144 | 0.024 | 0 | 0.044 | 0 |
γ8 | 0.745 | 0 | -0.094 | -0.020 | -0.075 | 0.134 | -0.369 |
γ9 | 1.041 | -0.024 | 0.321 | 0.042 | -0.O05 | 0.112 | -0.00 |
γ10 | 0.752 | 0 | 0.082 | 0.017 | 0.084 | 0.114 | -0.623 |
γ11 | 1.112 | -0.020 | -0.268 | -0.035 | 0.004 | 0.095 | 0 |
γ12 | 0.043 | 0.002 | 0.038 | -0.002 | 0.082 | 0.527 | 0.006 |
γ13 | 0.819 | 0 | -0.002 | 0 | -1.046 | 0.031 | 0.006 |
γ14 | 1.0 | 0 | -0.001 | 0.001 | 0.017 | 0.005 | -0.004 |
Sensitivity Matrix of Moduli with Respect to Estimated Parameters
Moduli | |||||||
---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | ||
Parameters | 0.991 | 0.871 | 0.816 | 0.184 | 0.624 | 0.251 | |
γ1 | 0.331 | -0.002 | 0.057 | -0.001 | -0.014 | 0.944 | 0 |
γ2 | 0.273 | 0.002 | 0.130 | 0.003 | -0.008 | 0.058 | 0 |
γ3 | 1.0 | 0 | 0 | 0 | 0 | 0 | 0 |
γ4 | 1.213 | 0 | 0.001 | 0 | 0.003 | 0.018 | 0 |
γ5 | 1.206 | 0 | 0.004 | -0.002 | -0.003 | 0.261 | 0.001 |
γ6 | 0.494 | -0.005 | -0.081 | 0.011 | 0 | 0.027 | 0 |
γ7 | 0.245 | -0.006 | -0.144 | 0.024 | 0 | 0.044 | 0 |
γ8 | 0.745 | 0 | -0.094 | -0.020 | -0.075 | 0.134 | -0.369 |
γ9 | 1.041 | -0.024 | 0.321 | 0.042 | -0.O05 | 0.112 | -0.00 |
γ10 | 0.752 | 0 | 0.082 | 0.017 | 0.084 | 0.114 | -0.623 |
γ11 | 1.112 | -0.020 | -0.268 | -0.035 | 0.004 | 0.095 | 0 |
γ12 | 0.043 | 0.002 | 0.038 | -0.002 | 0.082 | 0.527 | 0.006 |
γ13 | 0.819 | 0 | -0.002 | 0 | -1.046 | 0.031 | 0.006 |
γ14 | 1.0 | 0 | -0.001 | 0.001 | 0.017 | 0.005 | -0.004 |
APPENDIX V The Standard Money Demand Model
To highlight the simulation results of this model, its results have been compared in Section III with the path for inflation yielded by a standard dynamic money demand equation in which real income and the nominal stock of money are treated as exogenous. In the framework of this model, the stock of real money balances is assumed to adjust proportionately to the difference between the demand for real money balances and the actual stock in the previous period
where θ is the coefficient of adjustment, 0 ≤ θ ≤ 1.
Substituting for log md from equation (2) in the text, and solving for the level of real money balances, one obtains
Then, using the equation for the expected rate of inflation,
which, for γ14=1—the assumption used in the paper—implies that
one can write the estimating form as
With the addition of the country dummies, this model was estimated by the ordinary least-squares method on the same pooled sample as that used for the expanded model described in the text. The results were as follows:74
As can be seen, this equation does fit the data well, and all parameters have the correct signs and are significantly different from zero at the 1 percent level.
Using this estimated equation, along with the definitional equation for the price level
the path of prices was simulated for the various experiments described in Section III.
References
Aghevli, Bijan B., “Money, Prices and the Balance of Payments: Indonesia, 1968-73,” Journal of Development Studies, Vol. 13 (January 1977), pp. 37–57.
Aghevli, Bijan B., and Mohsin S. Khan, “Inflationary Finance and the Dynamics of Inflation: Indonesia, 1951-72,” American Economic Review, Vol. 67 (June 1977), pp. 390–403.
Aghevli, Bijan B., “Government Deficits and the Inflationary Process in Developing Countries,” Staff Papers, International Monetary Fund, Vol. 25 (September 1978), pp. 383–416.
Aghevli, Bijan B., “Credit Policy and the Balance of Payments in Developing Countries,” in Money and Monetary Policy in Less Developed Countries; A Survey of Issues and Evidence, ed. by Warren L. Coats, Jr., and Deena R. Khatkhate (Oxford; New York: Pergamon Press, 1980), pp. 685–711.
Aghevli, Bijan B., P.R. Narvekar, and Brock K. Short, “Monetary Poiicy in Selected Asian Countries,” Staff Papers, International Monetary Fund, Vol. 26 (December 1979), pp. 775–824.
Barro, Robert J., “Unanticipated Money Growth and Unemployment in the United States,” American Economic Review, Vol. 67 (March 1977), pp. 101–15.
Barro, Robert J., “Unanticipated Money, Output, and the Price Level in the United States,” Journal of Political Economy, Vol. 86 (August 1978), pp. 549–80.
Basmann, R.L., “Letter to the Editor,” Econometrica, Vol. 30 (October 1962), pp. 824–26.
Blejer, Mario I, “The Short-Run Dynamics of Prices and the Balance of Payments,” American Economic Review, Vol. 67 (June 1977), pp. 419–28.
Blejer, Mario L, and Roque B. Fernandez, “The Effects of Unanticipated Money Growth on Prices and on Output and Its Composition in a Fixed-Exchange-Rate Open Economy,” Canadian Journal of Economics, Vol, 13 (February 1980), pp. 82–95.
Cagan, Phillip, “The Monetary Dynamics of Hyperinflation,” in Studies in the Quantity Theory of Money, ed. by Milton Friedman (Chicago: University of Chicago Press, 1956), pp. 25–117.
Clements, Kenneth W., “A General Equilibrium Econometric Model of the Open Economy,” International Economic Review, Vol. 21 (June 1980), pp. 469–88.
Clements, Kenneth W., and Peter D. Jonson, “Unanticipated Money, ‘Disequilibrium’ Modelling and Rational Expectations.” Economics Letters, Vol. 2 (No. 4, 1979), pp. 303–308.
Dornbusch, Rudiger, “Devaluation, Money, and Nontraded Goods,” American Economic Review, Vol. 63 (December 1973), pp. 871–80.
Dutton, Dean S., “A Model of Self-Generating Inflation: The Argentine Case,” Journal of Money, Credit and Banking, Vol. 3 (May) 1971, Part 1, pp. 245–62.
Frenkel, Jacob A., and Harry G. Johnson, eds.,The Monetary Approach to the Balance of Payments(London: Allen & Unwin, 1976.
Goldman, Steven M., “Hyperinflation and the Rate of Growth in the Money Supply,” Journal of Economic Theory, Vol. 5 (October 1972), pp. 250–57.
Harberger, Arnold C, “A Primer on Inflation,” Journal of Money, Credit and Banking, Vol. 10 (November 1978), pp. 505–21.
International Monetary Fund,The Monetary Approach to the Balance of Payments (Washington: International Monetary Fund, 1977).
Keller, Peter M., “Implications of Credit Policies for Output and the Balance of Payments,” Staff Papers, International Monetary Fund, Vol. 27 (September 1980), pp. 451–77.
Khan, Mohsin S., “A Monetary Model of Balance of Payments: The Case of Venezuela,” Journal of Monetary Economics, Vol. 2 (July 1976), pp. 311–32.
Khan, Mohsin S., “Variable Expectations and the Demand for Money in High-Inflation Countries,” Manchester School of Economics and Social Studies, Vol. 45 (September 1977), pp. 270–93.
Khan, Mohsin S., “Monetary Shocks and the Dynamics of Inflation,” Staff Papers, International Monetary Fund, Vol. 27 (June 1980), pp. 250–84.
Knight, Malcolm D., and Clifford R. Wymer, “A Macroeconomic Model of the United Kingdom,” Staff Papers, International Monetary Fund, Vol. 25 (December 1978), pp. 742–78.
Knight, Malcolm D., and Donald J. Mathieson, “Economic Change and Policy Response in Canada under Fixed and Flexible Exchange Rates,” in Economic Interdependence and Flexible Exchange Rates, ed. by Jagdeep S. Bhandari and Bluford H. Putnam (Cambridge, Massachusetts: MIT Press, 1983).
Laidler, David E.W., The Demand for Money: Theories and Evidence (New York: Dun-Donnelley, >2nd ed., 1977).
Laidler, David E.W., “The Demand for Money in the United States—Yet Again,” in On the State of Macro-Economics, ed. by Karl Brunner and Allan H. Meltzer, Carnegie -Rochester Conference Series on Public Policy, A Supplementary Series to the Journal of Monetary Economics, Vol. 12(1980), pp. 219–71.
Laidler, David E.W., and Patrick O’Shea, “An Empirical Macro-model of an Open Economy under Fixed Exchange Rates: The United Kingdom, 1954—1970,” Economica, Vol. 47 (May 1980), pp. 141–58.
Magee, Stephen P., “The Empirical Evidence on the Monetary Approach to the Balance of Payments and Exchange Rates,” American Economic Review, Vol. 66 (May 1976), pp. 163–70.
Mathieson, Donald J., “Interest Rates and Monetary Aggregates During a Financial Reform” (unpublished; International Monetary Fund, December 26, 1979).
Olivera, Julio H.G., “Money, Prices and Fiscal Lags: A Note on the Dynamics of Inflation,” Banca Nazi on ale del Lavoro, Quarterly Review, Vol. 20 (September 1967), pp. 258–67.
Otani, Ichiro, and Yung Chul Park, “A Monetary Model of the Korean Economy,” Staff Papers, International Monetary Fund, Vol. 23 (March 1976), pp, 164–99.
Polak, J.J., “Monetary Analysis of Income Formation and Payments Problems,” Staff Papers, International Monetary Fund, Vol. 6 (November 1957), pp. 1–50.
Reichmann, Thomas M., and Richard T. Stillson, “Experience with Balance of Payments Adjustment: Stand-By Arrangements in the Higher Credit Tranches, 1963-72,” Staff Papers, International Monetary Fund, Vol. 25 (June 1978), pp. 293–309.
Robichek, E. Walter, “Financial Programming Exercises of the International Monetary Fund in Latin America,” address to a seminar of Brazilian professors of economics. Rio de Janeiro, September 20, 1967.
Robichek, E. Walter, “Financial Programming, Stand-By Arrangements, and Stabilization Programs” (unpublished; International Monetary Fund, 1971).
Sargent, Thomas J., and Neil Wallace. “Rational Expectations and the Dynamics of Hyperinflation,” International Economic Review, Vol. 14 (June 1973), pp. 328–50.
Tanzi, Vito, “Inflation, Real Tax Revenue, and the Case for Inflationary Finance: Theory with an Application to Argentina,” Staff Papers, International Monetary Fund. Vol. 25 (September 1978), pp. 417–51.
Vogel, Robert C, “The Dynamics of Inflation in Latin America, 1950-1969,” American Economic Review, Vol. 64 (March 1974), pp. 102–14.
von Furstenberg, George M., “The Effect of the Changing Size and Composition of Government Purchases on Potential Output,” Review of Economics and Statistics, Vol. 62 (February 1980), pp. 74–80.
White, William H., “The Importance of ‘Blocked’ Compensating Deposit Balances for Setting Stand-By Credit Ceilings and Monetary Targets in LDCs” (unpublished; International Monetary Fund, February 27, 1980).
Williamson, John, “Economic Theory and International Monetary Fund Policies,” in Monetary Institutions and the Policy Process, ed. by Karl Brunner and Allan H. Meltzer, Carnegie -Rochester Conference Series on Public Policy, A Supplementary Series to the Journal of Monetary Economics, Vol, 13(1980), pp. 255–78.
Wymer, Clifford R., “Continuous Time Models in Macro-Economics: Specification and Estimation,” paper presented at SSRC-Ford Foundation Conference on Macroeconomic Policy and Adjustment in Open Economies, Ware, England, April 28-May 1, 1976.
This paper was published in Staff Papers, International Monetary Fund, Vol. 28 (March 1981), pp. 1-53. Apart from colleagues in the IMF, the authors are grateful to Vittorio Corbo, Rudiger Dornbusch, Michael Mussa, and John Williamson for helpful comments on the paper. The views expressed are the authors’ and do not necessarily represent those of the IMF
See Vogel (1974), Aghevli and Khan (1978), Harberger (1978), Aghevli and others (1979), and Khan (1980).
See the papers contained in Frenkel and Johnson (1976) and International Monetary Fund (1977). Magee (1976) also provides a useful survey of this topic.
See, for example, Blejer and Fernandez (1980).
One previous study, by Aghevli and Khan (1980), does try to generalize by using the same model for eight developing countries.
Robichek (1967), pp. 1-2.
See, for example, Williamson (1980).
Such stickiness could arise from the existence of prior contracts, or simply because inflation is driven by expectations that are slow to be revised.
See Frenkel and Johnson (1976) and International Monetary Fund (1977). Such generalizations have been made by Blejer (1977) for Mexico, Knight and Wymer (1978) for the United Kingdom, and Knight and Mathieson (1983) for Canada.
See Polak (1957) and Robichek (1967). For a view of the Fund approach from the outside, see Williamson (1980).
Careful study of the individual characteristics of a specific country would obviously be required before a stabilization program could be tailored to its particular circumstances.
This does not mean that the exchange rate cannot be altered, but only that it is policy determined. Indeed, an exchange rate change may well be one of the policy actions included in a stabilization package. It is recognized that a few developing countries have a floating exchange rate, and that a larger number follow some variation of a crawling-peg system. The present model explicitly allows an exchange rate change to affect domestic prices, and it could easily be extended to a crawling-peg regime.
Financial reform policies in recent years in some Latin American and Asian countries have resulted in a fairly rapid development of domestic capital and financial markets. See, for example, Mathieson (1979). For such countries, it is possible that this model would involve some misspecification. However, the experience of these countries is quite recent and not representative of developing countries in general.
In the steady state, λ1 = γ1γ4g, where γ4 is the income elasticity of the demand for money and g ¡0 is the rate of growth of capacity output. This ensures that domestic prices are at their equilibrium level relative to foreign prices.
Here, μ0 represents the equilibrium ratio of domestic prices to prices in the rest of the world. This ratio depends on such factors as domestic and foreign tastes and levels of productivity.
This assumption of γ= 1 is, of course, one of the main features of the monetary approach to the balance of payments.
For a discussion of the role of this variable in the inflationary process, see Knight and Mathieson (1983)).
See Khan (1980).
For a survey of money demand functions, see Laidler (1977; 1980).
Furthermore, because of controls imposed by the authorities, the interest rates that are available show very little variation over time. This makes it difficult to detect empirically any systematic relationship between money holdings and interest rates. For a discussion of the interest rate data that are available for some developing countries, see White (1980).
This has to be done because such valuation changes do not affect the domestic money stock or the excess demand for money. If F is the stock of reserves valued in foreign currency, then R =ϵ F. It is Δlog F that is related to the excess demand for money, and Δlog F = Δlog R -Δlogϵ.
The demand for nominal money balances is simply logMd = logmd + logP.
See also Olivera (1967), Dutton (1971). Aghevli and Khan (1977), and Tanzi (1978).
The Aghevli-Khan expenditure function is cast in real terms, that is, desired real expenditure is related to the level of real income. In combination with a nominal revenue formulation, this implies asymmetric behavior in the components of the deficit. In the model here, both expenditure and revenues are written in nominal terms.
If government expenditure and revenues both grow at the same rate as nominal income in the long run, then it would imply that γ9 = γ11= 1. Starting from an equilibrium position, this would ensure a balanced budget in the steady state. In the short run, however, even with the condition being satisfied that the income elasticities equal unity, one could observe a divergence between expenditure and revenues that would result from differences in the values of the adjustment parameters γ8and γ10.
See Knight and Wymer (1978) for an example of such a model. In a recent paper, Keller (1980) also examines theoretically the relationship between monetary factors and the supply side of the economy in developing countries.
In principle, one would also like to include the effects of changes in fiscal policy and relative prices on the flows of real aggregate demand and output. To determine the direct impact on output of a change in the relation between domestic and foreign prices, the term
To catch the stimulative effect of an increase in real government spending on output,
In a growing economy where g ≠0, λ 2 must be equal to
At the same time, the increase in the expected rate of inflation would tend to lower demand.
The countries are Argentina, Brazil, Chile, Colombia, the Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti, Honduras, Nicaragua, Panama, Paraguay, Uruguay, Jamaica, Jordan, Sri Lanka, India, the Republic of Korea, Malaysia, Nepal, the Philippines, Singapore, Thailand, Burundi, Ghana, Kenya, Malawi, and Zambia.
In each estimated equation, the constant term is a combination of an adjustment parameter and the basic ratios, such as those of domestic to foreign prices, money to income, and government spending and taxes to income. The dummy variables that are introduced into each equation standardize the sample. Thus, allowance is made for the fact that these ratios differ widely from one country to another, while adjustment parameters and the elasticities are still restricted to be the same across countries.
The computer program employed to calculate the estimates is entitled RESIMUL and was written by Clifford R. Wymer.
This approximation method automatically adjusts the domestic currency value of international reserves to offset the valuation effects of exchange rate changes. See Appendix III.
Because of the presence of these dummies, equations (22) and (23) were treated as stochastic, although with the values for γ15to γ23from the linearization imposed. This effectively leaves the identity for real money balances as the sole non-stochastic equation.
Since none of the countries in our sample have freely floating exchange rates, Δlog ϵ is generally zero, and this variable was omitted from equation (1) during estimation. Since log ϵ-1 is present in the term multiplied by γ2, however, devaluations or revaluations can still affect the domestic price level via this term.
These are denoted as “t-values,” even though, strictly speaking, this ratio has an asymptotic normal distribution. Hypothesis testing would thus have to be based on the normal distribution rather than the r-distribution. In this case, with a sample size of 232 observations, however, there is obviously no distinction.
In simultaneous estimation, the R2is bounded (-7∞, 1) and not bounded (0, 1). See Basmann (1962).
See, for example, the results reported by Aghevli and others (1979) and Khan (1980).
The existence of a wide variety of close financial substitutes for money in those countries also allows for economies of scale in holding money, thereby giving rise to the possibility that the income elasticity may be less than unity. See Laidler (1977).
Khan’s (1980) estimates for 11 developing countries tend to cluster around a value of two, using quarterly data.
It would have made it more nonlinear in parameters. The alternative of searching for γ14, since it is bounded (0,1), would have been fairly time consuming. The estimation model has 208 parameters (including those of the country dummies) and 232 observations, so that even a single FIML estimate takes a considerable amount of computer time.
The elasticity is significantly greater than unity, which would mean that government revenues rise secularly as a proportion of nominal income. Of course, since the revenue data have not been adjusted for discretionary tax changes, the elasticity could be biased upward.
This result can be shown to be similar to that emerging from models of the rational expectations variety, where only unanticipated monetary changes affect output. For a discussion of the relationship between the specification in this paper and the models of Sargent and Wallace (1973)) and Barro (1977; 1978), see Clements and Jonson (1979)).
The equation can be interpreted in a partial-adjustment framework where the rate or growth of real income responds proportionally to the difference between suppliers’ “desired” level of real output, as represented by thecapacity level, and actual real output of the previous period. In this case, the parameter γ13 would represent the coefficient of adjustment, with the expression l/γ13 measuring the average time lag.
A value of unity would mean that in the absence of any monetary disequilibrium, real income would always be at its trend value.
Such a model could be derived by relating domestic expenditure to the excess demand for real balances and utilizing the national income identity. See Laidler and O’Shea (1980)).
This result is consistent with the well-known empirical observation that the elasticities of export supply and import demand differ widely from one developing country to another. Furthermore, the inclusion of this variable affected other parameters in the system. For these reasons, the current specification was chosen.
The expression [1 -(MSE/σ2γ)] is bounded from above at unity and may be treated, heuristically, like an R2.
While the coefficients in the domestic credit and money supply equations were imposed by the linearization procedure, as mentioned earlier, these equations were actually treated as stochastic in the estimation.
Even though the theoretical model can be considered linear in logarithms, the size of the matrix of endogenous variables makes it impossible to evaluate the stability of the model analytically. It is thus necessary to determine stability through numerical means.
See Appendix IV.
The effect of a change in the relative price of domestic versus foreign goods (for example, an oil-price shock) could also be simulated by changing the level of μ, but that experiment is somewhat tangential to the present paper.
As already noted, the fact that no direct effect of government spending on real output could be detected empirically means that only the monetary effects of changes in fiscal policy are taken into account in the model. In these circumstances, it makes no difference whether the present stochastic shock initially hits domestic credit, government spending, or taxes, since the time path traced by the model will be the same in all cases.
The model is constructed in such a way that the reserve leakages associated with the balance of payments deficit begin in the period after the monetary expansion that induces it. Thus, the monetary shock is assumed to take place right at the end of period one, and the reserve leakages start from the beginning of period two. Since no leakages occur during the first period, the money stock initially rises by the full extent of the increase in domestic credit.
The simple model assumes that a partial-adjustment process determines holdings of real money balances and imposes the same restrictions on the expectations mechanism as the model in the paper. The reduced form of the simple model was estimated for the pooled sample by the ordinary least-squares method. For details, see Appendix V.
The short-run responses to a change in nominal income are estimated from the sample as γ8γ9=0.775 for nominal government expenditure, and γ10γ11=0.837 for tax revenue.
As noted earlier, λ= γ1γ4g and λ2= (1 + γ4γ12-γ13)g.
The analysis could also be cast in terms of a flow target: namely, a desired balance of payments position, or target reserves-to-imports ratio. While making the simulations somewhat different, the results should be qualitatively similar to the experiment performed here.
For this analysis, the target level of international reserves is taken as exogenous; this means that one equation must be eliminated from the model. We have chosen to drop the government-spending equation, so that the level of public expenditure in nominal terms is now determined by the domestic credit identity.
It is certainly possible that with a target that is defined in flow terms, the pattern for domestic credit would be quite different. This possibility should be kept in mind when examining the present simulations.
The fluctuating pattern is essentially the consequence of the built-in inertia in the model. We are indebted to Michael Mussa for pointing this out.
The costs of underemployment are universally acknowledged, but overemployment may also cause serious externalities in a developing economy, especially when it is concentrated in the industrial sector. Even if job prospects in this sector are only temporary, overemployment may induce a significant increase in migration to the cities, exacerbating urban problems and stretching public services beyond capacity.
In principle, a benevolent government that was certain to retain office forever would apply a social discount rate of zero. Obviously, most real-world governments can be presumed to have a shorter time horizon, so that they regard immediate employment losses as more costly than future losses.
For a theoretical analysis of the relation between government fiscal policy and potential output in a developed economy, see von Furstenberg (1980).
This is in the context of a model where real capacity output is constant over time. An analogous restriction in a growth context would obviously be the rate of domestic credit expansion.
Obviously, the concept of real income is complicated when the relative price of traded a net nontraded goods is allowed to change. However, this problem is of only limited relevance to the empirical results. The theoretical and empirical aspects of this problem are discussed in more detail in Knight and Mathieson (1983).
See Clements (1980).
If the supply of nontraded goods were made a function of capacity rather than current output, a term representing a Phillips curve effect would appear in the price equation (equation 40). We are indebted to Rudiger Dornbusch for emphasizing this point. This specification, however, did not prove to be empirically robust.
If αj is complex, that is, αj = a + bi, then the modulus of αj is given by
This technique is described in Wymer (1976).
Since γ3 is a coefficient solely of an exogenous variable, its partial derivatives are all necessarily zero.
The t-values are reported in parentheses below the coefficients; R2 is the adjusted coefficient of determination; and SEE is the standard error of the estimated equation.